The Commitment to Development Index:
2004 Edition
David Roodman
Center for Global Development
April 2004
Roodman, The Commitment to Development Index: 2004 Edition
Contents
Introduction................................................................................................................................................... 1
I.
Scaling and weighting .......................................................................................................................... 2
II. The Seven Components........................................................................................................................ 7
Aid............................................................................................................................................... 7
Trade ......................................................................................................................................... 17
Investment................................................................................................................................. 28
Migration................................................................................................................................... 32
Environment.............................................................................................................................. 36
Security ..................................................................................................................................... 40
Technology ............................................................................................................................... 44
III.
Overall results ................................................................................................................................ 46
Appendix. Aggregation of Protection Levels across Sectors...................................................................... 54
Bibliography ............................................................................................................................................... 57
Tables and figures
Table 1. Computation of selectivity weights .............................................................................................. 12
Table 2. Quality-adjusted aid quantity by donor, bilateral or multilateral.................................................. 15
Table 3. Calculation of policy-induced charitable giving........................................................................... 16
Table 4. Quality-adjusted aid quantity with multilateral aid allocated back to bilaterals ........................... 17
Table 5. Calculation of uniform ad valorem tariff-equivalent of tariff regimes against agricultural
commodities, 1997–98 (percent)........................................................................................................ 21
Table 6. Calculation of uniform ad valorem tariff-equivalent of tariff regimes against other agriculture,
1997–98 (percent) .............................................................................................................................. 22
Table 7. Calculations of production-distorting agricultural subsidies for CDI and of Total Support
Estimate of OECD, 2000–02.............................................................................................................. 23
Table 8. Overview of estimation of ad valorem tariff-equivalent of subsidies........................................... 24
Table 9. Protection in ad valorem tariff equivalents, by sector (%) ........................................................... 25
Table 10. Computation of Aggregate Measure of Protection (AMP) ......................................................... 26
Table 11. Revealed openness, 2002 ............................................................................................................ 27
Table 12. Computation of overall trade score............................................................................................. 27
Table 13. Investment component scoring system ....................................................................................... 30
Table 14. Summary of Investment Component .......................................................................................... 31
Table 15. Comparison of 2003 and 2004 CDI investment ranks ................................................................ 32
Table 16. Summary of migration component ............................................................................................. 35
Table 17. Comparison of 2003 and 2004 CDI migration ranks.................................................................. 36
Table 18. Indicators used in environment component ................................................................................ 39
Table 19. Summary of environment component......................................................................................... 40
Table 20. Non–U.N.-run military operations counted in CDI security component. ................................... 42
Table 21. Summary of security component—contributions as a share of GDP (%) .................................. 43
Table 22. Comparison of 2003 and 2004 CDI security/peacekeeping ranks .............................................. 44
Table 23. Calculation of technology scores ................................................................................................ 46
Table 24. The Commitment to Development Index: Scores....................................................................... 49
Table 25. Detailed confrontation of 2003 and 2004 scores ........................................................................ 50
Table 26. Confrontation of 2003 and 2004 standings using 2004 methodology ........................................ 51
Figure 1. Correlation of standard CDI with versions with higher weight placed on one component......... 52
Figure 2. Average absolute change in CDI rank when higher weight placed on one component .............. 53
Figure 3. Supply and demand for an import from a developing country .................................................... 54
Roodman, The Commitment to Development Index: 2004 Edition
Introduction
One year ago, the Center for Global Development and Foreign Policy magazine introduced the
Commitment to Development Index (Birdsall and Roodman 2003; CGD and FP 2003).1 The superficial purpose was to rate rich countries based on how much their government policies hurt or
help development in poorer countries. But “ranking the rich” was a means to more important
ends: to draw media attention to the many ways that rich-country governments affect development, to provoke debate on which policies matter and how to measure them, to highlight gaps in
current knowledge, to stimulate data collection and other research, to educate the public and
policymakers, and, ultimately, to prod policy reform.
By these standards, the initiative has been reasonably successful in its first year in the
limelight. The index garnered coverage from the New York Times, the Economist, and the BBC,
among many other outlets, and provoked commentary in the American National Review and the
Australian Financial Review.2 Experts are writing papers on how to improve the CDI. (Picciotto
2003; Chowdhury and Squire 2003; Sawada et al. 2004) Professors are teaching with the Foreign
Policy article. And the Dutch government has adopted the CDI as a performance indicator of its
“policy coherence” for development.
The first edition of the CDI embodied 12 months of rapid, creative work. As noted in
Birdsall and Roodman (2003), the richest components of the first edition of CDI were those for
policy areas with the best data sets and traditions of scholarly inquiry, namely foreign aid and
trade. Other components were cruder for lack of such data and analysis, sometimes using measures of outcomes rather than rich-country policies—for example, the amount of foreign direct
investment in developing countries, which is largely a function of developing-country policies,
as well as private decisions. We nevertheless included the cruder components in 2003 in part to
send the message that “helping is about more than aid” (investment, migration, etc., matter too),
in part to invite commentary about how to better measure policy in those areas. Indeed, in the
months after the release, CGD and FP were showered with constructive criticism of the CDI.
Thus it was clear when the 2003 edition was released that the design work needed to continue. Improving the CDI would increase the meaningfulness of the index, boost its credibility,
and help it serve its ultimate purposes more effectively.
Guided by our own understanding of the shortcomings of the original edition and the
helpful suggestions we received, CGD undertook or commissioned substantial data gathering and
analysis in order to revise the CDI methodology. The collaborators for the 2004 edition were
Theodore Moran of the Georgetown University School of Foreign Service (on investment),
Elizabeth Grieco and Kimberly Hamilton of the Migration Policy Institute (migration), and Michael O’Hanlon and Adriana Lins de Albuquerque of the Brookings Institution (security). Together with CGD researchers, the team refined the aid, trade, and environment components,
added a technology component, and overhauled the sections on investment, migration, and secu1
The Commitment to Development Index is a collective effort. I am grateful to the Rockefeller Foundation for its
support; to the outside collaborators for technical work on components; to Alicia Bannon for excellent research assistance, and to CGD President Nancy Birdsall and the rest of the Advisory Board for guidance. The design of the
2004 CDI does not necessarily represent the views of any of the outside collaborators.
2
Jeff Madrick, “Looking beyond Free Trade,” New York Times, June 12, 2003; “Economics Focus: Gauging Generosity,” Economist, May 1, 2003; Jim Lacey, “We’re Number Twenty?!” National Review, June 30, 2003; Helen
Hughes, “Faulty Index A Poor Guide to Real Benefits,” Australian Financial Review, May 20, 2003.
1
Roodman, The Commitment to Development Index: 2004 Edition
rity (formerly called peacekeeping). (The 2004 final design departs in places from the recommendations of non-CGD authors. Final responsibility rests solely with CGD.)
One thing that did not change is the basic concept of the CDI. It still ranks 21 countries:
all the members of the Development Assistance Committee (DAC) of the OECD save Luxembourg. Each country gets a separate score in each of a set of major policy areas that were selected
because they affect development and have relevant data available for all 21 countries. This year
policy domains are aid, trade, investment, migration environment, security, and (newly added)
technology. A country’s overall score is the average of its seven component scores. The CDI
rates countries in ways that allow normative comparisons, which usually means adjusting for
size. Denmark cannot be expected to give as much foreign aid as Japan, which has an economy
26 times as big, but Japan could be asked to give as much as Denmark as a share of its gross domestic product, and that is how the index gauges aid quantity. Switzerland cannot be expect to
import as much from developing countries as the Unite States, but it could have trade barriers as
low, which is what the trade component looks for.
And as in 2003, the CDI aims to assess policies today. In practice, because it takes time
for official data to surface, most information used is for 2001 and 2002. The trade data are one
major exception, being for 1997. Another is the data on contributions to internationally approved
security operations, which cover the 1993–2002 decade. Since these troop levels committed can
vary dramatically from year to year, a historical average is a better indicator of current policy
stance than a single year of data.
This paper describes the full CDI methodology as it stands now. Section I describes how
we revised our handling of what is likely to become a perennial issue, how to scale and weight
the components. Section II reviews the index component by component. It builds on background
papers written for each of the seven policy areas (Roodman 2004a, 2004b; Cline 2004; Moran
2004; Grieco and Hamilton 2004; O’Hanlon and de Albuquerque 2003; Bannon and Roodman
2004), while making explicit where the final CDI departs from their recommendations. Section
III presents the overall results, compares them to last year’s, and analyzes their sensitivity to
changes in component weights.
I.
Scaling and weighting
The CDI combines readings on dozens of indicators. Since the indicators are not perfectly correlated, countries’ standings on the final results are affected by the relative importance the CDI
formulas give to the various indicators. In mathematical terms, the results are affected by choices
of both functional form and choice of coefficients. Both the CDI designers and commentators
have naturally asked whether the CDI makes the best choices.
In some parts of the CDI, the way in which indicators are combined is grounded in a clear
conceptual framework and calibrated to available evidence. For example, the aid component
combines donors’ aid-giving totals with information on the extent to which they tie their aid (requiring recipients to spend it on donor-country goods and services) by referring to a finding that
tying raises project costs 15–30%. Tied aid is discounted 20% (see below for the rationale), and
the result is a figure, tying-discounted aid, that still has real-world meaning. Other examples are
the theoretically-grounded method that is used to express agricultural subsidies in tariffequivalent terms, which allows them to be combined with actual tariffs; and the reasonable but
2
Roodman, The Commitment to Development Index: 2004 Edition
coarse assumption that the marginal cost of deploying personnel in international security operations is $10,000/month/person, which allows personnel and financial contributions to such operations to be combined in a bottom-line dollar figure. All these techniques use theory and evidence
to reduce, though certainly not eliminate, arbitrariness in the CDI design.
But where theory and evidence are thinner, we have not found such solid ways to reduce
arbitrariness. When needed to combine indicators in a sort of conceptual vacuum, we restricted
ourselves to taking linear combinations, as a first step toward managing the complexity. This
happened in the trade, investment, migration, and environment components, and in each of these
cases the CDI designers chose to weight some indicators more than others. The weights are of
course open to challenge, but are backed by years of experience in the relevant fields.
At the top level of the CDI hierarchy, however, where the seven CDI components merge
into a single index, we equally weighted the components. Because of the prominence of this
choice and its importance for the final results (section III examines how important it really is),
and because we have weighted equally, this decision has provoked many challenges. I will focus
on it for the rest of the section.
Intuitively, taking linear combinations happens in two steps: mapping each variable to be
combined onto a standard scale, which may involve scaling and translation (shifting up or
down); then taking a weighted average. Both steps—standardizing and weighting—raise tough
conceptual questions. Consider the challenges of standardizing first. To prepare the scores on the
seven CDI components for being combined into an overall score, the standardizing system
should arguably have the following properties:
1. Standardized scores should all fall within some intuitive scale, say 0–10.
2. For components that measure “goods” (in the 2004 CDI, they are aid, investment, migration, security, and technology), zero should map to zero. That is, if a country gives
no aid (more precisely, if its aid program is deemed valueless after adjusting for quality), its final aid score should be 0—not, say, 3. For components that measure “bads”
(environment and trade, which mainly assess environmental harm and trade barriers)
a perfect absence of the thing assessed should translate into an intuitive maximum
score, such as 10. All this is nearly equivalent to requiring that the coefficient of
variation (standard deviation divided by the mean) be preserved. For the “good”
components, it also means that the transformation should be a simple rescaling, with
no translation.
3. The standardized averages on each component, at least in some base year, should be
the same—say, 5. Then one can immediately tell by looking at a country’s aid, environment, or other score whether it is above or below the base-year average. And one
can tell whether a country’s score in one component is better than its score in another,
by the standards of the country’s peers. Last year’s scoring system (described below)
did not have this property. The average trade score (6.4) was twice the average aid
score (3.2). As a result, when Switzerland scored 4.0 on trade and 3.3 on aid, it appeared to a lay reader to be better on trade than aid when in fact it was below average
on trade and above average on aid.
3
Roodman, The Commitment to Development Index: 2004 Edition
4. The standard deviation of standardized scores should be the same for each component—as they would be if they were z scores (number of standard deviations from the
mean). In other words, countries should be “graded on a curve” for each component.
If they are not—if, instead, standardized scores on one component are relatively clustered—this effectively under-weights that component because differences between
countries on the component will have relatively little effect on the overall results.
Since we have restricted ourselves to linear transformations, two free parameters—slope
and intercept—determine how the results from each component are standardized. With seven
components, that yields 14 degrees of freedom. The above constraints together would consume
far more than 14 degrees of freedom. The first imposes what we can call 14 inequalities3, and the
other three impose 6 equalities each, for a total of 18. Thus only by luck could all four conditions
be satisfied. If one drops the requirement that standard deviations are equal, there is more hope
(12 equalities and 14 inequalities imposed on 14 parameters), but it still would take luck.
In the first two years of the CDI, luck has not been with us. As a result, we have faced
trade-offs, trade-offs that are tricky because they involve mathematical principles, our (limited)
understanding of rich world-poor world linkages, and the imperatives of effective mass communication. For example, last year’s standardized investment scores had an average of 3.0. To force
those scores to have an intuitive mean of 5, we would have had to add 2 to every country’s standardized investment score, which would have raised Portugal to 11 and given a “no investment”
country 2 points out of 10. Or we would have had to multiply all the scores by 5/3, which would
have raised Portugal to 15. Thus, enforcing condition 3 would have led to violations of condition
1 and perhaps 2.
We made different choices in 2003 and 2004. In 2003, we did enforce conditions 1 and 2.
Scores for the “good” components were scaled so that the best country got a 9 (and a country that
gave no aid or made no contribution to peacekeeping would have gotten a 0). Nine was chosen to
allow some room for improvement in future years within the 10-point scale.4 For the “bad” components, scores were scales so that the worst country got a –9 (and the country that did no harm
to the global environment or erected no trade barriers would have gotten a 0); then 10 was added
to bring the scores within the 0–10 frame. This system resulted in different components having
different means.
In 2004, we gave up on condition 1 in favor of condition 3. Scores on each component
now average 5 by fiat; as a result, so do the overall CDI scores. But the boundaries of 0 and 10
are no longer inviolable. Countries whose aid programs, say, are deemed more than twice as
good as average score above 10. And countries with trade barriers or rates of environmental
harm more than twice the average score below 0. In fact, 6 of the 147 component scores this year
exceed 10 and one is negative.5 These few transgression of the intuitive range seem worth the
greater ease of comparison within and across components. For example, in a reversal from last
3
Technically the first condition imposes 21×7×2=294 inequalities: each country’s score on each component should
be ≥0 and ≤10. The “14 inequalities” apply to the maximum and minimum scores on each component.
4
The plan was to stay with the 2003 scaling factors, so that the best would no longer automatically get a 9 in future
years. Note that this meant that scores might have some day transgressed outside the 0–10 range, so that compliance
with condition 1 was not strictly guaranteed.
5
See section 3. Denmark, the Netherlands, Norway, and Sweden score over 10 on aid, as do Canada and the United
States on migration; and Norway’s trade score is negative.
4
Roodman, The Commitment to Development Index: 2004 Edition
year, Switzerland now scores higher on aid than trade—5.8 versus 0.3—which makes more sense
for a country that is above the average of its peers on aid and well below it on trade.
An astute reader will have noticed in the discussion of condition 4, which demands equal
standard deviations, that the question of weighting crept into the discussion of scaling. Using a
linear transformation to double the range or standard deviation of a component has exactly the
same effect on overall standings as doubling its weight.
Nevertheless, for the lay reader, weighting is a distinct concept, and raises distinct concerns. Indeed, among the most oft-voiced criticisms of the CDI is that it is “equal weighted,”
even though some policy domains, it is argued, clearly matter more than others. The accusation
of equal weighting is true in a sense: in 2004, as in 2003, a country’s overall CDI score is the
simple average of its component scores.
Before examining the criticism, it is worth noting that “equal weighting” is a not a welldefined concept, for all it is voiced. Consider that allowing trade scores to range more widely in
2004 (Norway now has a negative score on trade) effectively happened to increase the effective
weight on trade. Yet the CDI is still “equal weighted.” Which year was trade really “equal
weighted”? Both, and neither. There are several reasonable ways to scale scores—characterized
in part by which of the above conditions are enforced—thus several possible rankings resulting
from “equal weighting.” Thus in choosing “equal weighting” for the CDI, we are not claiming to
truly give aid, trade, etc., equal weight. That would be meaningless. Rather, both this year and
last year, we have opted for what seems least arbitrary in the face of uncertainty.
Still, I agree with the attacks on “equal weighting” in the sense that the CDI almost certainly does not have the following property: any two CDI-measured policy changes in a given
country that have an equal effect on development have an equal effect on the CDI. We have not
striven for that ideal yet, out of several considerations. First, achieving it does not seem essential
for the CDI as a communications strategy and a goad to research, and it must be remembered that
these are the ultimate goals of the project, not scientific measurement. The CDI broadcasts the
basic message that many policy areas matter and that all countries have major room for improvement clearly as is. The success of the first edition in spotlighting issues despite being
“equal-weighted” is reassuring.
Second, a survey of expert opinion suggests that “equal weighting” is not unreasonable.
Shyamal Chowdhury and Lyn Squire (2003) surveyed members of the Global Development
Network, who are researchers in both rich and poor countries working on development issues. Of
the 200 solicited respondents in the stratified random sample, 105 completed the questionnaire.
They were asked to assign their own weights to each of the major issue areas then in the CDI.6
The authors computed the average chosen weight for each component. The results were statistically indistinguishable from equal weighting. The study corroborates my own experience. Of the
6
The survey was based on the first draft of the CDI, in which anti-corruption was a separate, seventh component
rather than being folded in to investment as it eventually was. The 2004 CDI also has a seventh component, technology, which was absent from this survey.
5
Roodman, The Commitment to Development Index: 2004 Edition
seven CDI issue areas, all but one has been nominated to me for extra weight by someone with
decades of experience in development.7
Of course, since “equal weighting” is not well defined, the precise meaning of this result
is unclear. In addition, although the survey instrument specifically mentioned “policy stance,” it
may not have highlighted the policy-flow distinction enough to avert misinterpretation. For example, a respondent might have given ample weight to investment because rich-country investment can indeed be a major factor in development—overlooking that rich-country policies that
influence investment are a smaller factor.
At any rate, there are more fundamental reasons to be cautious about departing from
“equal weighting.” One phrase in the ideal property enunciated above, “equal effect on development,” is, like “equal weighting,” not well defined. Different policies have different effects on
people in different times and places. Moral and philosophical conundrums arise about how one
should compare effects on people with different levels of poverty and opportunities; about which
discount rate to use; and about whether development is a something that happens to people or
countries.8 Huge uncertainties also loom about the actual long-term effects of trade barriers,
greenhouse gas policies, government R&D spending, humanitarian interventions, migration, etc.
Finally, it cannot be assumed that the proper mathematical form for combining the components into an overall score is linear. Especially for large donor nations, the policy areas may
interact significantly. For example, Thomas Hertel, head of the Global Trade Analysis Project,
has called for simultaneous computable general equilibrium modeling of trade and migration.9
To the extent policy areas interact, there can be no right weights in a linear framework.
None of this proves that it would be impossible in light of current knowledge, or especially with more research, to stick with the linear approach and yet find unequal weights that
would command a broader consensus than equal weighting does. One starting point might be estimates of global dollar flows of aid, trade, investment, remittances, and so on. Greenhouse gases
could be converted to the same dollar units via a fixed rate per ton, based on estimates of the
harm climate change could do to developing country economies. Picciotto (2003) suggests such
an approach along these lines.10
But from the point of view of the CDI, flows are merely intermediaries between richcountry policies on the one hand and poor-country development on the other, and it is the magnitude of the linkage between these latter variables that should determine ideal weights. In some
areas, these relationships are reasonably well understood. For example, several studies have estimated the economic effects of rich-country trade policies on poor-country development. (World
Bank 2001; Cline 2004) Cline estimates that complete rich-country liberalization would, after a
15-year adjustment, increase income in developing countries by $100 billion per year, which is
7
The exception is environment—and that is probably only because hardly any environmental experts have commented. Surely it can be argued that tinkering with the planet’s biogeochemical cycles is an issue of the first rank.
8
This last distinction is important for migration. If someone moves permanently from a poor to a rich country, quadruples her income, and sends back no remittances, is that development?
9
Private communication between Thomas Hertel and Michael Clemens, CGD, October 2002.
10
But for trade, Picciotto suggests using estimates of the benefits, in producer surpluses, of complete rich-country
liberalization rather than current earnings on exports from developing to developed countries. This is not parallel to
current total aid, remittance, or investment flows.
6
Roodman, The Commitment to Development Index: 2004 Edition
approximately twice current concessional aid flows. Similar work is now being done on migration liberalization. CGE modeling by Walmsley and Winters (2003) suggests that an if rich countries increased their temporary migrant worker stocks by an amount equal to just 3% of their labor forces, global income would increase $150 billion, with most of that accruing to the temporary workers themselves. Complete liberalization could generate vastly larger gains for temporary workers.11
The trouble with unequal weighting is that one cannot do it halfway. As soon as one, say,
doubles trade’s weight relative to aid, one needs sound rationales for the choice of weights for
every other component. The links between policy and development in other policy domains are
much more uncertain or controversial. Most empirical evidence on the macro-level benefits of
aid is fragile (Easterly, Levine, and Roodman 2004; Roodman 2003, revised draft forthcoming).
There is little evidence on how investment-relevant policies in rich countries affect developing
countries. And it is far from clear how to weigh in security interventions and rich-country public
R&D investment.
For the time being then, we have stood by the humble choice of “equal weights.” We
hope that the CDI will increasingly spur research to speed the day when unequal weighting will
be more defensible.12 In the meantime, we find “equal weighting” serviceable.
II. The Seven Components
Aid
The aid component designed by Roodman (2004a) is a revised version of that used last year
(Roodman 2003). As it did last year, it modifies the traditional quantity measure of aid programs
(aid/donor GDP) to reflect several concerns, among them debt service received, tying, and selectivity.13 This year, it also incorporates a penalty for donors that raise the recipient administrative
costs of aid by funding many small projects.14 And it factors in private charitable giving to developing countries to the extent this can be credited to government fiscal policy. The component
is built largely on data from the Development Assistance Committee (DAC).
11
This does not automatically imply, however, that the migration component is currently underweighted relative to,
say, trade. On the current scale, conceivably, a country that completely liberalized temporary migration might earn a
score of 50 or 100—a score so high that it might actually exaggerate the benefits of migration. In other words, it is
possible with the current scaling that a 1-point increase in trade score still corresponds to more benefit than a 1-point
increase on migration.
12
In January 2003, in Cairo, at the annual Global Development Network conference, the Global Policy Project ran a
two-day workshop on “The Development Impact of Rich Countries’ Policies,” which identified such research gaps.
The summary is at
http://www.gdnet.org/pdf/Fourth_Annual_Conference/Workshop3/Workshop4_global_policies_project.pdf. In October, in Washington, D.C., CGD and the GDN sponsored a conference on “Quantifying the Impact of Rich Countries’ Policies on Poor Countries,” at which new research was presented. A full list of papers is at
http://www.cgdev.org/events/quantifying_impact.cfm.
13
Unlike in 2003, administrative costs are not subtracted from bilateral flows. Last year, Roodman (2003) argued
that “large donor-to-donor variations in their share of gross ODA may indicate inefficiencies. At any rate, netting
out administrative costs gives a truer picture of the amount of aid reaching recipients, which is a legitimate basis for
comparison.” But low administrative costs may also reflect lack of monitoring and evaluation, which can reduce aid
quality. I am indebted to Mark McGillivray for making this point.
14
I thank Robert Picciotto for suggesting this addition.
7
Roodman, The Commitment to Development Index: 2004 Edition
The calculations run broadly as follows. First all debt service received from developing
countries on concessional (low-interest) loans is subtracted from donors’ gross aid totals. Then
three aspects of quality are factored in as discounts or penalties: tying (requiring aid to be spent
only on donor-country goods and services), selectivity (the degree to which aid goes to countries
that are relatively rich or poorly governed), and proliferation (the tendency to fund many small
projects. Finally, a portion of the private charitable giving to developing countries is credited to
government tax policies.
More precisely:
•
The starting point is gross disbursements of grants and concessional loans by donor (bilateral or multilateral) and recipient. The data are the latest available, for 2002. Included
here is what DAC terms Official Development Assistance (ODA) and Official Assistance
(OA).15
•
Principal and interest payments are netted out, to more closely reflect net transfers to recipients. DAC’s standard “net ODA” statistic is net of principal payments only. The rationale for the DAC approach appears to be an analogy with net foreign direct investment.16 Only return of capital is netted out of net FDI, not repatriation of earnings. Similarly, only amortization is netted out of net ODA, not interest, which can be seen as the
donors’ “earnings” on aid investment. I find the analogy inapt. In the case of FDI, return
of capital can be expected to reduce the host country’s capital stock much more than repatriation of profits. Put otherwise, when foreign corporations locally reinvest profits,
they generally increase the host country’s capital stock—and net FDI treats it that way.
But when the government of Ghana writes a check to the government of Japan for $1 million, it hardly matters for either party whether it says “interest” or “principal” in the
check’s memo field. It seems unlikely that interest and principal payments have different
effects on Ghana’s development investments. For this reason, the CDI treats all debt service uniformly.
•
For bilaterals, tied aid is discounted 20%. The base for this discounting is gross disbursements. Studies suggest that tying raises aid project costs 15–30% (Jepma 1991),
which translates into a reduction in aid value of 13–23%.17 20% is a round figure toward
the top of this range. “Partially untied”18 aid is discounted 10%. The tying figures this
year come directly from project-level data in DAC’s Creditor Reporting System database.
But like last year, the available tying data are for aid commitments rather than disbursements. Rates of tying are assumed to be the same for commitments and disbursements.
15
OA is like ODA except that it goes to “Part II” countries, which include most European states that emerged out of
the Soviet bloc and richer non-DAC members such as Israel and Singapore. DAC excludes OA from its most frequently cited statistic (net ODA), but we included it in our quality-adjusted aid measure because many Part II countries are in need and receiving aid. Some, such as Ukraine, are poorer than many Part I countries. Aid to relatively
rich countries, such as Israel, will be heavily discounted in a subsequent step.
16
I base this on an excellent exchange with Simon Scott, Principal Administrator of the Statistics and Monitoring
Division of the OECD’s Development Co-operation Directorate, who I cannot assume was representing institutional
views, but only his own views.
17
A 15-percent cost increase lowers the purchasing power of aid by 1–1/1.15 = 13 percent. Similarly, a 30-percent
costs increase cuts the value of aid 23 percent.
18
Aid that must be spent on goods and services from the donor nation or developing countries; or aid that must be
spent on goods and services from developing countries only.
8
Roodman, The Commitment to Development Index: 2004 Edition
Unlike last year, technical assistance is not automatically assumed tied, and is only
treated as tied if reported as tied in the CRS.19
•
The resulting tying-discounted net transfers are split into two kinds of aid, project and
program, based on the detailed classification by purpose in the CRS. “Project aid” is
meant to include funding for activities over which donors try to exercise detailed, ongoing control. “Program aid” is meant to include budget support, adjustment loans, debt relief, and sector-wide action programs (SWAPs)—aid over which the recipient ideally has
near-complete control.20 In the splitting, all debt service is treated as (negative) program
aid since its opportunity cost should be a reduction in recipient governments’ discretionary spending.
•
For each triplet of donor, recipient, and aid type (project or program), the tyingdiscounted net transfer is multiplied by a “selectivity weight” that is meant to reflect the
country’s appropriateness for aid. The selectivity weight is now the product of two factors.21 The first is linearly related to the country’s Kaufmann-Kraay composite governance score. (Kaufmann, Kraay, and Mastruzzi 2003) For project aid, the governance factor ranges between 0.25 and 0.75. The country with the lowest governance score, the
Democratic Republic of Congo, defines the bottom of that range, getting a 0.25, while
Chile anchors the top. For program aid, the scale is stretched to 0.0–1.0. As a result, program aid to the best-governed countries is rewarded more than project aid to them, while
the reverse is true for countries with the weakest governance. This distinction is meant to
capture the trade-offs between local and foreign “ownership” of aid-financed activities. In
Chile, for example, detailed donor management of aid is deemed to be counterproductive,
so project aid gets a 0.75 multiplier while program aid gets a 1.0. But in the DRC, governance is so weak that more donor control is treated as beneficial (weight of 0.25 instead
of 0.0). Overall, though, both types of aid treated as more effective the better governed
the recipient is.
Of course because it is poor, the DRC cannot be expected to have very good governance. So the second selectivity multiplier reflects the country’s poverty. It is linearly
related to the country’s log real 2002 GDP/capita, with Hong Kong (GDP/capita of
$23,200), getting a 0 and Sierra Leone, the poorest country with data (real GDP/capita of
$450), defining the upper end. Table 1 shows the resulting weights.22
Emergency aid is exempted from selectivity discounting, to acknowledge, in a
way that is practical given the available data, that some forms of aid may be more valuable in countries with the worst governance.
19
Technical assistance may deserve a discount far higher than 20% since foreign experts are often an order of magnitude more expensive than local ones. Most studies of the costs of tying have looked at tied goods rather than services.
20
In practice, the CRS purpose classification was not designed to make this distinction, so the splitting is imperfect.
See Roodman (2004a) for details.
21
In 2003, the selectivity weight was a linear function of two factors, which were added rather than multiplied: log
real GDP/capita and KK governance controlling for GDP/capita. The new system seems more intuitive and allows
the new distinction between program and project aid.
22
The maximum of this scale, Sierra Leone’s value, is 2.27. The intention in choosing this number was to arrange
for the maximum combined selectivity factor (poverty factor × governance factor) for any country for either program or project aid would be 1.00. However, because of a programming error discovered after the 2004 numbers
were frozen, it turns out that the maximum combined factor is 1.16, for program aid to Madagascar.
9
Roodman, The Commitment to Development Index: 2004 Edition
•
For each bilateral and multilateral donor, the resulting selectivity-weighted aid figures are
summed across recipients to obtain a single figure for each donor, whether bilateral or
multilateral.
•
Each donor is assessed a penalty for “project proliferation,” the phenomenon of funding
many small aid projects. Project proliferation is thought to overburden recipient governments with administrative and reporting responsibilities, and lure the most talented workers out of government and into the employ of the donors, thus undermining the effectiveness of aid projects, and government administration in general. (Cassen 1994; Brown et
al., 2001; Knack and Rahman 2004). The penalty is calculated as the fraction of aid dollar
commitments made in quanta of $100,000 or less, times 4, times gross disbursements.23
The factor of 4 was derived from a more sophisticated model of proliferation costs in
Roodman (2004c).24 Thus a country that committed 1% of its aid in 2002 in amounts less
than $100,000, would be assessed a penalty of 4% of its gross disbursements. The penalty
is treated as already having been weighted for selectivity, as if proliferation cost is already appropriately discounted for countries where aid is predicted to be less productive
anyway.
•
The result is a “quality-adjusted aid quantity” for each bilateral or multilateral donor.
(See Table 2.) The quality-adjusted aid totals of multilaterals are then allocated back to
bilaterals in proportion to the bilaterals’ contributions to the multilaterals during 2002.
For example, since France accounted for 5.96 percent of 2002 contributions to the World
Bank’s concessional lending arm, the International Development Association (IDA), it
receives credit for 5.5 percent of the IDA’s quality-adjusted aid of $2.8 billion, or $166
million.
•
The final performance measure for government aid is bilaterals’ total quality-adjusted aid
as a share of GDP. (See Table 3.)
This year, the aid component also rewards policies that encourage private charitable giving to development organizations. Private giving is encouraged by specific tax incentives that
lower the “price” of giving. And it is encouraged by a low tax/GDP ratio, which leaves citizens
and corporations with more after-tax income to spend on charitable giving. The approach taken
in the CDI is to estimate the proportional increase in giving caused by each country’s fiscal policies, compare that to actual giving, then work backwards to estimate how much actual giving is a
credit to policy. (See Table 3.) Specifically:
•
An estimate is made of the increase in charitable giving brought about by tax incentives.
All but three CDI countries—Austria, Finland, and Sweden—offer income tax deductions
23
With two exceptions. New Zealand did not report to the CRS database. Lacking commitments data, the full
model’s estimate, based on a large set of interpolations of underlying variables, was used. For multilaterals without
CRS data, the extrapolations seemed too tenuous because multilaterals are more diverse in purpose and giving patterns, and so multilaterals without CRS data were assigned a zero proliferation penalty.
24
Because of the complexity of the model in Roodman (2004c) and the minimal empirical evidence to test and calibrate it, it seemed best to use something much simpler into the CDI. So the results from the full approach for the
selectivity-weighted proliferation costs by donor were regressed on a set of simple indicators of proliferation. The
share of aid commitment dollars made in amounts below $100,000 turned out to have a correlation of 0.78 with the
full model’s results, and other indicators could explain little additional variation. See Roodman (2004a).
10
Roodman, The Commitment to Development Index: 2004 Edition
or credits for charitable giving. We treat all such provisions as deductions; that is, we
treat them as reducing the price of giving by “the” marginal income tax rate. Because tax
codes are complex and varied, different taxpayers within countries pay different marginal
rates, and the codes are extremely difficult to summarize and compare internationally.
We therefore use a crude but available proxy for the marginal income tax rate, namely the
share of gross wage earnings going to income tax for workers at 167% of the income
level of the average production worker (OECD 200x). Thus, for example, this statistic is
22.6% for the United States, so incentives in the U.S. tax code are treated as reducing the
price of charitable giving by 22.6%. In other words, thanks to the deduction, a U.S. citizen can effectively spend 77.4¢ to get $1 worth of charity, on average. Based on a survey
of the literature, we set the price elasticity of charitable giving at –0.5. In the United
States, this implies that income tax incentives increase charitable giving by 13.7%.25
•
An estimate is also made of how much having lower taxes increases giving. The benchmark against which “lowness” is measured is Sweden’s 2001 tax revenue/GDP ratio of
51.4%, the highest among the 21 countries. The United States, to continue the example, is
treated as having reduced its total tax burden from Sweden’s 51.4% to the actual 28.9%.
This raises the privately claimed share of GDP from 48.6% to 71.1%, an increase of
46.3% over the base.26 Again drawing on the literature, we take the income elasticity of
giving to be 1.1: charitable giving increases somewhat more than proportionally with private income. As a result, the lower U.S. tax burden is estimated to raise charity 52.0%.27
•
The price and income effects are then combined. For the United States, the 13.7% and
52.0% increases compound for a 72.7% increase.
•
DAC data on actual private giving to developing countries is then used to estimate what
giving would have been in the absence of these policies, and what credit should be given
to policy. This statistic counts all giving by individuals and foundations, but leaves out
official aid that is channeled through foreign NGOs. In the U.S. case, charitable giving is
reported at $5.66 billion for 2002. The CDI estimates this would have been $3.27 billion
before the policy-induced 72.7% increase, and attributes the difference, $2.38 billion to
favorable fiscal policies.
•
The policy-induced increases in charitable giving are then discounted for quality and
added to the official quality-adjusted aid quantities. Private giving too can go to countries
that are more or less appropriate for aid, and can contribute to the problems of project
proliferation, for example by siphoning off talented administrators from government service. As a rough adjustment, the CDI discounts policy-induced private giving by the simple average of the quality discounts for bilaterals’ own aid programs, which is 37%. To
complete the U.S. example, we credit the country for $2.38 billion × (1–37%) = $1.50
billion in quality adjusted aid. Added to its $7.295 billion in official quality-adjusted aid,
25
The precise calculation is (1–0.226)–0.5–1=0.137.
Some share of the revenue funds transfer payments, which the recipients may use as they please, including for
charitable giving. The precise effect of this redistribution on giving is difficult to know because it depends on such
factors as the difference in income between the typical taxpayer and typical transfer recipient. This seems like too
complicated an issue to address.
27
The precise calculation is ((1–0.289)/(1–0.514))1.1–1=0.52.
26
11
Roodman, The Commitment to Development Index: 2004 Edition
this raises its CDI aid score to 1.9, from what would have been 1.6 had charitable contributions not been considered in the CDI in 2004.
The treatment of charitable giving involves a number of coarse assumptions. It models
taxpayers with a single representative agent, simplifies complex tax provisions, uses rough but
ready approximations for the appropriate tax rates, assumes certain fixed elasticities, and assumes that the elasticities are the same for charitable giving to developing countries as they are
for charitable giving in general. Its methodological sophistication, such as it is, should not be
confused with precision. But given the modest impact the adjustment has for the United States—
the country most known for private generosity and public stinginess when it comes to aid—it
seems unlikely that refinements in methodology, though desirable, would affect results much.
Overall, despite the quality adjustments and the incorporation of private giving, what
most distinguishes donors from each other in the CDI is still the raw quantity of official aid they
disburse. Denmark, the Netherlands, Norway, Sweden are large donors by DAC’s quantity
measure (net ODA), and they score highest on the CDI aid measure too—as they did last year.
The nearly 8-fold variation between the most generous donor by DAC’s measure (Denmark, at
0.96% of GDP) and the least generous (the United States, at 0.13%) dominates differences in
quality, which probably does not vary across donors so much. They also dominate private giving.
That said, the new proliferation penalty does have a noticeable effect at the bottom of standings,
pulling down Greece, Ireland, and New Zealand. These small donors appear to spread their aid
thinly. As a result, the United States and Japan no longer occupy the bottom two slots of the aid
ranking, despite being lowest in net concessional transfers to developing countries.
The final column of Table 4 compares the CDI index of official aid performance to the
standard DAC measure. This summarizes all of the “adjustments” in quality-adjusted aid. New
Zealand suffers most from the adjustments because of a very high project proliferation penalty28.
Greece, Ireland, and Spain also experience heavy discounts because of project proliferation. And
Japan is pulled low because of its high debt service receipts on ODA loans. Of course, aid quality as defined here leaves out many factors that influence aid effectiveness, such as appropriateness of project designs, appropriateness and enforcement of loan conditions, and so on. The CDI
aid component is hardly the final word on measuring aid performance.
Table 1. Computation of selectivity weights
Country name
Formula:
Madagascar
Malawi
Tanzania
Mali
Benin
Burkina Faso
Mongolia
Zambia
Niger
28
A. Real
GDP/capita, B. Log real
2002 ($)
GDP/capita
650
519
493
777
913
896
1,462
713
685
Log A
6.48
6.25
6.20
6.66
6.82
6.80
7.29
6.57
6.53
D. KaufE. Governmann-Kraay ance selecC. GDP
composite tivity multiselectivity governance plier, project
multiplier score, 2002
aid
(linear map
(linear map
of B to 0–
of D to 0.25–
2.27 range)
0.75 range)
2.06
–0.07
0.53
2.19
–0.42
0.47
2.22
–0.54
0.46
1.96
–0.35
0.49
1.86
–0.26
0.50
1.88
–0.31
0.49
1.59
0.21
0.58
2.01
–0.57
0.45
2.03
–0.64
0.44
F. Governance selectivity multiplier, program aid
(linear map
of D to 0–1
range)
0.56
0.45
0.41
0.47
0.50
0.49
0.65
0.40
0.38
H. ComG. Com- bined sebined se- lectivity
lectivity multiplier,
multiplier, program
project aid
aid
C×E
1.09
1.04
1.01
0.95
0.93
0.92
0.92
0.90
0.89
C×F
1.16
0.98
0.92
0.92
0.94
0.91
1.04
0.80
0.77
This penalty had to be extrapolated because the country no longer reports the requisite data to DAC.
12
Roodman, The Commitment to Development Index: 2004 Edition
Country name
Mauritania
Mozambique
Senegal
Ethiopia
Guinea-Bissau
Eritrea
Sierra Leone
Moldova
Yemen, Rep.
Kenya
Ghana
Nepal
Uganda
Gambia, The
Lesotho
Cambodia
Togo
Nigeria
Comoros
Chad
Burundi
India
Tajikistan
Bolivia
Rwanda
Congo, Rep.
Vietnam
Kyrgyz Republic
Bangladesh
Central African Republic
Papua New Guinea
Djibouti
Honduras
Cameroon
Sri Lanka
Cote d’Ivoire
Pakistan
Armenia
Jamaica
Dominica
Morocco
Lao PDR
Jordan
Cape Verde
Guinea
Uzbekistan
Philippines
Georgia
Egypt, Arab Rep.
Congo, Dem. Rep.
Guyana
Syrian Arab Republic
Haiti
Ecuador
El Salvador
Guatemala
Albania
China
Indonesia
Belize
A. Real
GDP/capita, B. Log real
2002 ($)
GDP/capita
1,337
7.20
1,009
6.92
1,359
7.21
641
6.46
689
6.54
848
6.74
451
6.11
1,267
7.14
693
6.54
878
6.78
1,815
7.50
1,171
7.07
1,199
7.09
1,525
7.33
2,011
7.61
1,459
7.29
1,291
7.16
753
6.62
1,452
7.28
892
6.79
543
6.30
2,276
7.73
811
6.70
2,089
7.64
1,081
6.99
856
6.75
1,982
7.59
1,391
7.24
1,537
7.34
1,061
6.97
1,895
7.55
1,795
7.49
2,231
7.71
1,515
7.32
3,052
8.02
1,328
7.19
1,783
7.49
2,618
7.87
3,341
8.11
4,660
8.45
3,335
8.11
1,485
7.30
3,634
8.20
4,238
8.35
1,793
7.49
1,426
7.26
3,560
8.18
1,939
7.57
3,276
8.09
537
6.29
3,617
8.19
2,996
8.01
1,397
7.24
3,050
8.02
4,138
8.33
3,476
8.15
3,517
8.17
3,961
8.28
2,778
7.93
5,229
8.56
D. Kaufmann-Kraay
C. GDP
composite
selectivity governance
multiplier score, 2002
1.64
–0.08
1.81
–0.40
1.64
–0.16
2.07
–0.84
2.03
–0.84
1.91
–0.73
2.27
–1.16
1.68
–0.43
2.02
–0.94
1.89
–0.81
1.47
–0.15
1.72
–0.66
1.71
–0.73
1.57
–0.53
1.41
–0.21
1.59
–0.59
1.66
–0.72
1.98
–1.20
1.60
–0.69
1.88
–1.09
2.16
–1.40
1.34
–0.19
1.93
–1.17
1.39
–0.38
1.77
–1.02
1.90
–1.19
1.42
–0.48
1.62
–0.85
1.56
–0.78
1.78
–1.13
1.44
–0.64
1.47
–0.71
1.35
–0.49
1.57
–0.91
1.17
–0.12
1.65
–1.10
1.48
–0.84
1.26
–0.40
1.12
–0.03
0.93
0.65
1.12
–0.05
1.58
–1.03
1.07
–0.01
0.98
0.22
1.48
–0.98
1.61
–1.22
1.08
–0.22
1.43
–1.00
1.13
–0.36
2.17
–1.82
1.07
–0.24
1.18
–0.66
1.62
–1.40
1.17
–0.66
0.99
–0.18
1.09
–0.53
1.09
–0.52
1.02
–0.34
1.22
–0.84
0.86
0.22
E. Govern- F. Governance selec- ance selectivity multi- tivity multiplier, project plier, proaid
gram aid
0.53
0.56
0.48
0.46
0.52
0.53
0.41
0.32
0.41
0.32
0.43
0.35
0.36
0.21
0.47
0.45
0.39
0.28
0.41
0.32
0.52
0.54
0.44
0.37
0.42
0.35
0.46
0.42
0.51
0.52
0.45
0.39
0.43
0.35
0.35
0.20
0.43
0.36
0.37
0.23
0.32
0.14
0.51
0.53
0.35
0.21
0.48
0.46
0.38
0.26
0.35
0.20
0.46
0.43
0.41
0.31
0.42
0.34
0.36
0.22
0.44
0.38
0.43
0.36
0.46
0.43
0.40
0.29
0.52
0.55
0.37
0.23
0.41
0.31
0.48
0.46
0.54
0.58
0.65
0.80
0.54
0.57
0.38
0.25
0.54
0.58
0.58
0.66
0.38
0.27
0.35
0.19
0.51
0.52
0.38
0.26
0.48
0.47
0.25
0.00
0.50
0.51
0.44
0.37
0.32
0.13
0.44
0.37
0.51
0.53
0.46
0.42
0.46
0.42
0.49
0.48
0.41
0.31
0.58
0.66
H. ComG. Com- bined sebined se- lectivity
lectivity multiplier,
multiplier, program
project aid
aid
0.87
0.92
0.86
0.82
0.84
0.87
0.84
0.65
0.83
0.64
0.81
0.67
0.81
0.48
0.79
0.75
0.79
0.57
0.78
0.61
0.76
0.79
0.75
0.64
0.72
0.60
0.72
0.65
0.72
0.73
0.71
0.63
0.71
0.59
0.69
0.39
0.69
0.58
0.69
0.44
0.69
0.29
0.69
0.70
0.69
0.40
0.67
0.64
0.67
0.45
0.67
0.38
0.66
0.61
0.66
0.50
0.65
0.52
0.64
0.40
0.63
0.55
0.63
0.53
0.63
0.58
0.62
0.46
0.61
0.64
0.60
0.38
0.60
0.46
0.60
0.57
0.60
0.64
0.60
0.74
0.60
0.64
0.60
0.40
0.58
0.62
0.57
0.64
0.57
0.40
0.56
0.31
0.55
0.56
0.55
0.38
0.55
0.53
0.54
0.00
0.54
0.54
0.52
0.44
0.51
0.22
0.51
0.44
0.51
0.52
0.50
0.45
0.50
0.45
0.50
0.49
0.50
0.38
0.50
0.57
13
Roodman, The Commitment to Development Index: 2004 Edition
Country name
Lebanon
Fiji
Peru
Panama
Namibia
Swaziland
Azerbaijan
Grenada
Angola
Sudan
Zimbabwe
Thailand
Bulgaria
Tunisia
Romania
Botswana
Ukraine
Costa Rica
Chile
Dominican Republic
Gabon
Latvia
Brazil
Paraguay
Macedonia, FYR
Malaysia
Bosnia and Herzegovina
Kazakhstan
Venezuela, RB
Algeria
Colombia
Trinidad and Tobago
Lithuania
Mexico
Belarus
Iran, Islamic Rep.
Mauritius
Antigua and Barbuda
Turkmenistan
South Africa
Croatia
Estonia
Russian Federation
Uruguay
Malta
Hungary
Argentina
Barbados
Saudi Arabia
Bahrain
Slovenia
Kuwait
Israel
Cyprus
Macao, China
Hong Kong, China
A. Real
GDP/capita, B. Log real
2002 ($)
GDP/capita
3,756
8.23
4,733
8.46
4,358
8.38
5,286
8.57
5,674
8.64
3,986
8.29
2,757
7.92
6,186
8.73
1,817
7.50
1,741
7.46
2,018
7.61
6,009
8.70
6,115
8.72
5,824
8.67
5,600
8.63
7,298
8.90
4,173
8.34
7,497
8.92
8,463
9.04
5,486
8.61
5,621
8.63
7,935
8.98
6,653
8.80
3,912
8.27
5,543
8.62
7,897
8.97
4,902
8.50
5,106
8.54
4,626
8.44
4,901
8.50
5,371
8.59
8,067
9.00
8,865
9.09
7,707
8.95
4,730
8.46
5,611
8.63
9,321
9.14
9,379
9.15
4,091
8.32
8,969
9.10
8,823
9.09
10,367
9.25
7,016
8.86
10,727
9.28
11,649
9.36
11,621
9.36
9,378
9.15
13,774
9.53
11,800
9.38
14,216
9.56
15,711
9.66
16,553
9.71
17,518
9.77
18,757
9.84
19,147
9.86
23,223
10.05
D. Kaufmann-Kraay
C. GDP
composite
selectivity governance
multiplier score, 2002
1.05
–0.44
0.92
–0.03
0.96
–0.22
0.85
0.16
0.81
0.31
1.02
–0.43
1.23
–0.96
0.76
0.50
1.47
–1.36
1.49
–1.40
1.41
–1.34
0.78
0.25
0.77
0.26
0.80
0.11
0.82
0.01
0.67
0.77
0.99
–0.59
0.65
0.81
0.58
1.28
0.83
–0.17
0.82
–0.28
0.62
0.64
0.72
0.02
1.03
–1.01
0.83
–0.47
0.62
0.45
0.90
–0.73
0.87
–0.67
0.93
–0.88
0.90
–0.81
0.84
–0.66
0.61
0.34
0.55
0.69
0.64
0.13
0.92
–0.98
0.82
–0.73
0.53
0.70
0.52
0.68
1.00
–1.30
0.55
0.38
0.56
0.29
0.46
0.94
0.69
–0.55
0.44
0.70
0.40
1.16
0.40
0.96
0.52
–0.58
0.30
1.24
0.39
–0.05
0.28
0.53
0.23
0.99
0.20
0.36
0.16
0.56
0.12
0.88
0.11
0.53
0.00
1.16
E. Govern- F. Governance selec- ance selectivity multi- tivity multiplier, project plier, proaid
gram aid
0.47
0.44
0.54
0.58
0.51
0.51
0.57
0.64
0.59
0.69
0.47
0.45
0.39
0.28
0.62
0.75
0.32
0.15
0.32
0.14
0.33
0.15
0.58
0.67
0.58
0.67
0.56
0.62
0.54
0.59
0.67
0.84
0.45
0.39
0.67
0.85
0.75
1.00
0.52
0.53
0.50
0.50
0.65
0.79
0.55
0.59
0.38
0.26
0.47
0.43
0.62
0.73
0.42
0.35
0.43
0.37
0.40
0.30
0.41
0.32
0.44
0.37
0.60
0.69
0.65
0.81
0.56
0.63
0.39
0.27
0.43
0.35
0.66
0.81
0.65
0.81
0.33
0.17
0.61
0.71
0.59
0.68
0.69
0.89
0.45
0.41
0.66
0.81
0.73
0.96
0.70
0.89
0.45
0.40
0.74
0.98
0.53
0.57
0.63
0.76
0.70
0.90
0.60
0.70
0.63
0.77
0.68
0.87
0.63
0.76
0.73
0.96
H. ComG. Com- bined sebined se- lectivity
lectivity multiplier,
multiplier, program
project aid
aid
0.50
0.47
0.49
0.53
0.49
0.50
0.49
0.54
0.48
0.56
0.48
0.45
0.48
0.34
0.48
0.57
0.47
0.22
0.47
0.20
0.46
0.22
0.45
0.52
0.45
0.51
0.45
0.50
0.45
0.48
0.45
0.56
0.44
0.39
0.44
0.55
0.44
0.58
0.43
0.44
0.41
0.40
0.40
0.49
0.39
0.43
0.39
0.27
0.38
0.36
0.38
0.45
0.38
0.31
0.38
0.32
0.37
0.28
0.37
0.29
0.37
0.31
0.36
0.42
0.36
0.45
0.36
0.40
0.35
0.25
0.35
0.29
0.34
0.43
0.34
0.42
0.33
0.17
0.33
0.39
0.33
0.38
0.32
0.41
0.31
0.28
0.29
0.36
0.29
0.38
0.28
0.36
0.24
0.21
0.22
0.30
0.21
0.22
0.18
0.21
0.16
0.20
0.12
0.14
0.10
0.12
0.08
0.11
0.07
0.08
0.00
0.00
14
Roodman, The Commitment to Development Index: 2004 Edition
Table 2. Quality-adjusted aid quantity by donor, bilateral or multilateral
A
Gross
aid
B
Amortization
C
D
E
Proliferation cost
F
Tying
cost
G
SelectivQualityInterest
Net aid
ity
adjusted aid
1
Formula:
A–B–C
(D–F)×G–E
Australia
777
0
0
777
113
70
0.63
332
Austria
509
3
2
504
83
31
0.59
198
Belgium
758
35
3
721
112
7
0.66
360
Canada
1,633
26
1
1,606
96
118
0.59
778
Denmark
1,178
45
9
1,124
181
35
0.70
580
Finland
291
7
0
283
31
17
0.66
145
France
5,624
947
0
4,678
309
348
0.51
1,886
Germany
4,538
945
356
3,237
150
119
0.63
1,826
Greece
123
0
0
123
46
22
0.47
1
Ireland
268
0
0
268
150
0
0.74
48
Italy
1,208
200
0
1,007
1
266
0.80
595
Japan
9,811
3,062
1,735
5,014
2
197
0.56
2,673
Netherlands
2,677
97
50
2,531
68
25
0.62
1,485
New Zealand
92
0
0
92
60
0
0.64
-1
Norway
1,193
5
0
1,188
365
2
0.71
482
Portugal
187
0
1
186
84
13
0.58
16
Spain
1,164
154
1
1,009
407
178
0.53
31
Sweden
1,351
1
0
1,350
143
12
0.72
817
Switzerland
826
4
0
822
27
2
0.70
549
United Kingdom
3,696
102
2
3,592
43
26
0.68
2,386
United States
13,847
1,033
443
12,371
140
1,997
0.55
5,597
AfDF
741
125
97
519
0
0
0.73
379
AsDF
1,168
262
156
750
0
0
0.67
506
CarDB
119
56
36
27
0
0
0.48
13
EBRD
72
0
0
72
0
0
0.44
32
EC
10,145
786
494
8,865
25
0
0.53
4,645
GEF
137
0
0
137
0
0
0.50
69
IDA
6,667
1,263
745
4,658
0
0
0.65
3,027
IDB Sp F
425
259
119
48
0
0
1.28
61
IFAD
250
102
34
114
0
0
0.75
85
Montreal Protocol
60
0
0
60
0
0
0.52
31
Nordic Dev.Fund
35
2
0
33
0
0
0.79
26
Other UN
630
0
0
630
0
0
0.50
318
SAF+ESAF(IMF)
2,936
1,984
0
951
0
0
0.91
868
UNDP
278
0
0
278
0
0
0.69
192
UNFPA
312
0
0
312
0
0
0.65
203
UNHCR
655
0
0
655
0
0
0.61
402
UNICEF
571
0
0
571
114
0
0.64
251
UNTA
478
0
0
478
0
0
0.60
287
WFP
352
0
0
352
0
0
0.68
240
1
The estimated proliferation cost already factors in selectivity, and so is not multiplied by the selectivity factor
in this formula.
15
Roodman, The Commitment to Development Index: 2004 Edition
Table 3. Calculation of policy-induced charitable giving
Country
Formula:
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Japan
Netherlands
N. Zealand
Norway
Portugal
Spain
Sweden
Switzerland
U.K.
U.S.
E. Giving
increase
B. Mar- C. Increase D. Tax
ginal in- in giving
reve- because of
come tax with incen- nue/GDP, smaller
1
rate (%)
tive (%) 2001 (%) gov’t (%)
((1–D)/(1–
(1–
51.4%))^
income
A×B)^price
3
4
elasticity–1
elasticity–1
1
32.0
21.3
30.1
49.2
0
17.3
0.0
45.4
13.7
1
34.4
23.5
45.8
12.7
1
24.8
15.3
35.1
37.5
1
41.5
30.7
49.8
3.6
0
32.8
0.0
46.1
12.1
1
17.9
10.4
45.0
14.6
1
29.5
19.1
36.8
33.5
1
7.6
4.0
36.9
33.3
1
22.2
13.4
29.9
49.6
1
24.2
14.9
42.0
21.5
1
9.4
5.1
27.3
55.7
1
21.2
12.7
39.5
27.2
1
25.7
16.0
33.8
40.5
1
28.4
18.2
43.3
18.5
1
12.2
6.7
33.5
41.2
1
17.8
10.3
35.2
37.2
0
30.8
0.0
51.4
0.0
1
14.6
8.2
30.6
48.0
1
18.2
10.6
37.3
32.3
1
22.6
13.7
28.9
52.0
A.
Tax
incentive?
G. Grants
F. Com- by NGOs
bined in- (million $,
2001
crease
2
prices)
(%)
(1+C)×
(1+E)–1
80.9
13.7
39.2
58.5
35.5
12.1
26.5
59.0
38.6
69.6
39.5
63.6
43.3
63.0
40.0
50.7
51.4
0.0
60.1
46.3
72.7
229
54
69
277
17
10
203
770
5
78
32
165
236
21
404
5
118
18
186
328
5,656
H. Giving in Giving
absence of attributed
to tax
favorable
tax policies policies
F/(1+G)
127
47
49
175
12
9
160
484
4
46
23
101
165
13
288
3
78
18
116
224
3,274
G–H
103
6
19
102
4
1
42
286
1
32
9
64
71
8
115
2
40
0
70
104
2,381
1
Income tax as % of gross wage earnings (2002)—single individual at 167% income level of the average
2
3
4
production worker. Data for latest available year. Price elasticity of giving taken to be –0.5. Income
elasticity of giving taken to be 1.1.
16
Roodman, The Commitment to Development Index: 2004 Edition
Table 4. Quality-adjusted aid quantity with multilateral aid allocated back to bilaterals
Bilateral
qualityadjusted
1
aid
Country
Australia
332
Austria
198
Belgium
360
Canada
778
Denmark
580
Finland
145
France
1,886
Germany
1,826
Greece
1
Ireland
48
Italy
595
Japan
2,673
Netherlands
1,485
New Zealand
-1
Norway
482
Portugal
16
Spain
31
Sweden
817
Switzerland
549
United Kingdom
2,386
United States
5,597
1
From Table 2.
QualityQualityMemo: OfTotal
adadjusted
ficial aid
qualityjusted Adjusted quality (Adaid allochari- (aid+chari justed
cated from adjusted
multilater- official Charita- table table giv- aid/Net
aid
ble giving giving ing)/GDP
als
ODA)
136
468
103
64
0.13%
44%
136
334
6
4
0.17%
44%
298
657
19
12
0.27%
52%
314
1,092
102
64
0.16%
48%
387
967
4
3
0.55%
59%
150
295
1
1
0.23%
53%
1,872
3,758
42
27
0.27%
52%
1,489
3,315
286
179
0.18%
51%
104
105
1
1
0.08%
35%
94
142
32
20
0.14%
29%
888
1,483
9
6
0.13%
60%
1,648
4,321
64
40
0.11%
45%
550
2,035
71
45
0.50%
54%
16
15
8
5
0.03%
59%
346
828
115
72
0.48%
45%
106
123
2
1
0.10%
32%
519
549
40
25
0.09%
31%
466
1,283
0
0
0.56%
58%
104
654
70
44
0.26%
61%
928
3,315
104
65
0.22%
57%
1,697
7,295
2,381 1,494
0.08%
45%
Trade
The focus of the trade component developed by William Cline (2002) is a measure of barriers in
rich-counties to goods exports from poorer ones. Cline’s index has two major parts. The first,
getting 75% weight, is an aggregate measure of protection (AMP), which estimates the combined
effect of tariffs, non-tariff measures, and domestic production subsidies on an ad valorem tariffequivalent basis. Out of concern that unmeasured (tacit) barriers may be an important factor in
reducing access of developing countries to rich country markets, especially in Japan, the remaining 25% weight goes to an indicator of “revealed openness,” which is essentially imports from
developing countries as a share of importer’s GDP. Cline has revised the index (Cline 2004, ch.
3), and I have made small changes in adapting it for the CDI.29
The CDI version of the AMP divides goods into six categories: agricultural commodities,
other agriculture (processed agricultural goods and fish), textiles, apparel, fuels, and other goods
(mainly manufactures). For all categories, estimates are gathered of the average ad valorem tariff-equivalent of tariffs (some tariffs are expressed in specific units such as tons). Subsidies (for
agricultural commodities only) and of quotas (textiles and apparel only) are also expressed in
29
One change cut the AMP for the European Union, Japan, and the United States by about half, but had relatively
little effect on relative standings on the component. In computing the average tariff within agriculture using world
production weights, 21 product categories were used, as described in text here, instead of the original 5.
17
Roodman, The Commitment to Development Index: 2004 Edition
tariff-equivalent terms. These are then combined into an overall estimate of protection in tariffequivalent terms. The EU is often treated as a single unit in trade data, so in some places, the 14
EU members that are scored in the CDI get identical numbers. But where the data permit, EU
members are differentiated.
The methodology computes the following variables in turn:
Agricultural tariffs on an ad valorem basis. For agricultural commodities and other agriculture, the simple tariff-equivalent of tariff regimes is computed as the weighted average of the
protection levels in major product categories such as wheat, rice, and cattle and sheep meat. The
weights are based on levels of world production. It may be theoretically preferable to derive the
weights from imports from developing countries rather than world production, but rich-country
protection in agriculture is so high that even a sophisticated model incorporating various estimated elasticities could be quite inaccurate in predicting imports in the absence of protection,
which is a necessary ingredient in calculating the harm that current protection does. Production
figures, in contrast, are not as endogenous to protection and give clear sense of which crops are
the major ones at the global level. All data are from the GTAP 5.0 database (Dimaranan and
McDougall 2002). (See Table 5 and Table 6.)
Dollar value of agricultural subsidies. It is often said that OECD governments spend
more than $300 billion a year subsidizing agricultural production. Although aid to rich-country
farmers is in fact copious, the $300 billion “fact,” so phrased, is wrong. Rather, OECD farmers
and food buyers receive support by virtue of government policy that is equivalent to more than
$300 billion in subsidies, as measured by the OECD’s Total Support Estimate (TSE). Much of
this benefit is actually delivered in the form of tariffs, which the OECD converts to subsidy
equivalents. Much of the rest includes “general services” such as agricultural education and
R&D, transfers to consumers rather than producers, and transfers to producers in ways that create
little incentive for additional production, thus little trade distortion.
Since the CDI aims to measure trade distortions, and handles tariffs separately, it uses a
narrower definition of subsidy, while still drawing on the OECD subsidy data. Table 7 shows the
full OECD agricultural subsidy typology, and how both the TSE and the CDI subsidy totals are
arrived at. The OECD lists three major kinds of support: support to producers, general services
such as agricultural extension and inspection services, and support to consumers. The first major
subcategory of producer support is Market Price Support (MPS, row B of the table), which is the
additional income accruing to producers because their farmgate prices are higher than world
prices. Governments maintain these price differentials with two kinds of border measures: barriers to imports (tariffs) and subsidies for exports. Import barriers account for the lion’s share of
MPS in OECD countries and, because they generate transfers from domestic consumers to domestic producers, they also show up as negative entries under consumer support (row T). Spending on export subsidies can be inferred by taking the algebraic sum of MPS and transfers from
consumers to producers, which carry a negative sign (see row X). The other subcategories of
OECD producer support are in fact subsidies in the sense of government expenditure.
The OECD’s TSE counts all producer support, including MPS, as well as general services and taxpayers subsidies to consumers—$300 billion/year average in 2000–02 for the 21
rated countries (row W). In contrast, the subsidy measure in the CDI consists purely of certain
subcategories of producer support, those that are true government expenditures that distort pro18
Roodman, The Commitment to Development Index: 2004 Edition
duction (rows X and Y). From the MPS it takes only export subsidies. It excludes payments
based on overall farming income since these should not distort production decisions. It also discounts payments based on historical entitlements by half. In theory, these subsidies too are decoupled from present production and shouldn’t distort it, but they are often administered in ways
that stimulate production. For example, the U.S. formally decoupled most support payments in
1996—but then disbursed an extra $8.6 billion/year in “emergency assistance” during 1998–
2001, and in 2002 allowed farmers to update the base figures for their “decoupled” subsidies.
And some EU payments are decoupled only at the national or regional level—allocation within
regions is still based on actual production. (de Gorter, Ingco, and Ignacio 2003) Throughout, averages for 2000–02 are used because subsidy levels are sensitive to volatile world prices. For the
21 scored countries, total trade-distorting subsidies are estimated at $79 billion/year for 2000–02.
EU subsidies disaggregated by member. Although agricultural subsidies are largely unified in the EU under the Common Agricultural Policy, the Policy itself can be thought of as the
outcome of a political process in which each country can be expected to have maximized its receipts by lobbying for subsidies on the products it is most able to produce. Moreover, EU members do offer additional subsidies to their farmers outside the CAP. Thus the amount of subsidy a
country’s farmers receive is in no small part a reflection of the country’s policy stance. For the
sake of differentiating EU members, the CDI therefore estimates subsidy levels for individual
EU members, following the method developed in Cline (2002). The CAP subsidy total is allocated to individual members based on the size of their contributions to the main Common Agricultural Policy fund, the European Agricultural Guidance and Guarantee Fund, and nationallevel, non-CAP subsidies are added in.
Ad valorem tariff equivalent of agricultural commodity subsidies. The methodology for
expressing agricultural subsidies in tariff-equivalent terms is too complex to summarize here.
Table 8 lays out the data and calculations as performed for the CDI. See Cline (2004, ch. 3).
Ad valorem tariff equivalent of protection in textiles, apparel, fuels, and other goods
(mainly manufactures). The CDI uses the values that Cline culls from the literature for each of
these categories. For textiles and apparel, the tariff-equivalents of tariffs and non-tariff measures
(quotas) are combined to give overall levels of protection in the sectors. See Table 9.
Aggregate Measure of Protection. For each rich country, the ad valorem tariff equivalents of protection in the six sectors are combined based on the value of imports in those categories from the non-DAC world. The precise formula used is different from Cline’s and arguably
more direct. (See Appendix.) Both approaches attempt to correct for the potential endogeneity of
imports to protection, which can bias straight import weighting (high protection reducing imports, thus underweighting the very categories where protection is highest). And both yield results that differ only modestly from straight import weighting.
Revealed openness is a weighted sum of imports from non-DAC countries over importer’s GDP for 2002. Imports from the least developed countries (LDCs) are double-weighted
to reflect the extra potential for trade to reduce poverty in countries where it is highest. Imports
of manufactures too are double-weighted because they seem more likely than, say, oil imports, to
be subject to the tacit barriers this component tries to detect (Cline 2004). As a result, manufac-
19
Roodman, The Commitment to Development Index: 2004 Edition
tures imports from LDCs are quadruple-weighted.30 All EU members are assigned the same revealed openness score.31 See Table 11.
Combining the AMP and revealed openness. The two main indicators have opposite
senses: lower AMP and higher openness should be rewarded. Because they are in effect separate
estimates of the same underlying variable, level of protection (they are correlated -0.77), they are
combined in a way that is unique within the 2004 CDI. The revealed openness scores are linearly
transformed to have the same mean, standard deviation, and sense as the AMP. Once they are on
the same scales, they are combined in a 75/25 ratio. (See Table 12.)
Not surprisingly given that most of the underlying data is the same as for last year, the
pattern of results has not changed much. Agricultural tariffs are still the dominant source of intercountry variation, giving Norway in particular a very low score, and Japan and Switzerland low
scores as well. Agricultural commodity subsidies, though harmful, only rival agricultural commodity tariffs where those tariffs are low, namely Australia, New Zealand, and the United States.
The source of the very high numbers for the EU, Norway, Switzerland, and Japan are tariff-rate
quotas (TRQs) enacted under the Uruguay Round agreement of the World Trade Organization.
Under this system, countries replaced import quotas with pairs of tariffs—lower tariffs on
within-quota imports higher ones on above-quota imports, in order to cap imports at the old
quota level, at least in theory.
These estimates of protection inherited several shortcomings from the GTAP tariff data
that dominate the results. Rather than using tariff rates actually applied, they use upper bound
rates committed to in the Uruguay Round of the General Agreement on Tariffs and Trade. The
estimates do not factor in the preferences (lower tariffs) that many rich countries grant developing countries, such as through the U.S. Africa Growth and Opportunity Act and the European
Everything But Arms initiative. They leave out contingent protection such as anti-dumping duties. And they are for 1997. In 2004, the World Bank (Kee, Nicita, and Olarreaga 2004) and the
Centre d’Etudes Prospectives et d’Informations Internationales (see Bouët et al. 2001) released
new sets of estimates which theoretically correct all these problems in the GTAP numbers. But
the results are already controversial among researchers. Kee, Nicita, and Olarreaga, for instance,
estimate Japan’s protection at just 3.9% tariff-equivalent with respect to all countries and 1.5%
with respect to low-income countries only—lower than their estimates for the EU, United States,
and Canada (World Bank 2004, ch. 10), and far less than the 13.0% figure with respect to low
and middle-income countries that is found here. Since the field was in ferment in early 2004, it
seemed best to stay with the GTAP 5 dataset which, whatever its theoretical limitations, has undergone peer review for use in a widely respected academic database. I hope that in 12 months
we will understand the key differences in assumptions and methods are that generating such differences in results. And by 2005, GTAP 6 should reach its own influential verdict.32
30
This is a small change from the 2003 CDI, in which they were triple-weighted. The new approach is more consistent and makes only a tiny difference since manufacturers imports from LDCs are low.
31
We experimented with computing revealed openness separately for each EU member, but found that it gave the
Netherlands and Belgium outsized scores, probably because they have small economies and are ports of entry for the
continent. The two probably ship a good share of their reported imports from developing countries on to other nations.
32
We seriously considered using the CEPII results in the 2004 CDI. I thank Lionel Fontagné and Yvan Decreux of
CEPII for their generous and responsive manner in which they provided us with their latest results, and Peter
20
Roodman, The Commitment to Development Index: 2004 Edition
Table 5. Calculation of uniform ad valorem tariff-equivalent of tariff regimes against agricultural commodities, 1997–98 (percent)
Greece
Germany
France
Finland
Denmark
Canada
Belgium
Austria
Australia
64.9
64.9
64.9
64.9
64.9
64.9
0.0
64.9
64.9
1.0
61.4
61.4
61.4
61.4
61.4
61.4
61.4
62.7
61.4
61.4
0.0
38.6
38.6
38.6
38.6
38.6
38.6
38.6
8.9
38.6
38.6
0.8
14.5
14.5
14.5
14.5
14.5
14.5
14.5
1.9
14.5
14.5
2.0
0.0 251.4
0.0 251.4
0.0 251.4
0.0 251.4
0.0 251.4
0.0 251.4
0.0 251.4
0.0
0.0 251.4
0.0 251.4
1.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.7
36.6
36.6
36.6
36.6
36.6
36.6
36.6
0.2
36.6
36.6
0.8
6.7
6.7
6.7
6.7
6.7
6.7
6.7
19.8
6.7
6.7
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
88.9
88.9
88.9
88.9
88.9
88.9
88.9
16.3
88.9
88.9
0.1
30.9
30.9
30.9
30.9
30.9
30.9
30.9
30.9
4.1
87.7
87.7
87.7
87.7
87.7
87.7
72.4 214.8
30.9
87.7
87.7
87.7
7.3
87.4
87.4
87.4
87.4
87.4
87.4
87.4
0.7
87.4
87.4
1.0
76.4
76.4
76.4
76.4
76.4
76.4
76.4
4.9
76.4
76.4
13.9
45.3
45.3
45.3
45.3
45.3
45.3
45.3
36.0
45.3
45.3
2.3
VegePlantAnimal
Cattle,
Dairy ProcOverall
Paddy
Other tables, Oil Sugar based Cattle, prod- Raw sheep Other prod- essed
weighted
rice Wheat grains fruit seeds cane fibers sheep ucts milk meat meat ucts
rice Sugar average
Ireland
64.9
5.0
0.0
36.4
30.9
87.7
87.4
76.4
58.2 287.0 409.0 116.1
3.1
45.3
118.8
0.0
Italy
0.0 149.1
88.9
0.0
0.0
0.0
0.0
76.4
6.7
11.2
44.9
36.6
11.1
20.2
0.0
3.9
198.1
Switzerland
Sweden
Spain
Portugal
4.9
64.9
64.9
64.9
2.6
61.4
35.8 120.0
64.9
102
61.4
61.4
61.4
141
0.6
38.6
38.6
38.6
4.7
14.5
98.9 101.2
38.6
439
14.5
94
17.7
48
0.7
0.0 251.4
97.4 152.8
47
9.7
0.0
205
1.1
36.6
291
0.6
6.7
202
0.0
0.0
246
5.3
88.9
275
3.6
30.9
285
42.5
87.7
155
5.3
87.4
98
53.4
76.4
2,776
9.4
45.3
137.3
155.4 333.9 195.3 109.4 129.9 205.3
409.0 249.2
Japan
0.0 251.4
0.0
53.1
45.3
14.5
0.4 351.3 368.2 316.8 107.2
76.4
45.3
38.6
0.9
87.4
76.4
45.3
61.4
0.0
87.7
87.4
76.4
64.9
0.0
30.9
87.7
87.4
Netherlands
0.0
88.9
30.9
87.7
0.0
0.0 319.3 103.1
0.0
88.9
30.9
2.5
6.7
0.0
88.9
1.2
36.6
6.7
0.0
0.0
0.0
36.6
6.7
0.0
0.0 251.4
0.0
36.6
33.0 131.0
New Zealand
14.5
0.0 251.4
0.0
0.4 197.7 296.3 223.9
United Kingdom
149
Norway
14.5
0.0 251.4
0.1 161.6 150.3
United States
Weight (World
production, billion $)
Timmer, Thomas Hertel, Bita Dimaranan, William Martin, Bernard Hoekman, and Paul Gibson for consultations on
the difficult decision not to use them in 2004.
21
Roodman, The Commitment to Development Index: 2004 Edition
Table 6. Calculation of uniform ad valorem tariff-equivalent of tariff regimes against other
agriculture, 1997–98 (percent)
Sweden
Spain
Portugal
Norway
New Zealand
Netherlands
Japan
Italy
Ireland
Greece
Germany
France
Finland
Denmark
Canada
Belgium
Austria
Australia
35.7
3.1
3.1
3.1
74.3
3.8
3.1
22.1
3.1
3.1
3.1
3.1
3.1
3.1
3.1
2.4
3.1
3.1
2.7
Other
crops
0.0
1.6
0.0
0.0
0.0
0.1
0.0
0.0
54.7
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.3
0.0
0.0
3.2
Wool
6.9
0.2
6.8
11.0
11.4
0.0
0.3
7.5
4.9
9.6
2.6
10.8
6.8
9.0
3.0
7.8
0.4
7.8
4.7
0.3
Fishing
11.4
114.5
11.4
11.4
11.4
94.4
2.2
11.4
6.6
11.4
11.4
11.4
11.4
11.4
11.4
11.4
8.6
11.4
11.4
2.8
Vegetable oils
28.8
111.2
28.8
28.8
28.8
157.2
12.5
28.8
38.3
28.8
28.8
28.8
28.8
28.8
28.8
28.8
14.1
28.8
28.8
5.6
Other
food
8.3
129.6
8.3
8.3
8.3
50.1
8.5
8.3
16.2
8.3
8.3
8.3
8.3
8.3
8.3
8.3
62.5
8.3
8.3
9.2
Beverages,
tobacco
18.4
102.5
18.4
18.6
18.7
105.6
9.1
18.4
26.8
18.5
18.1
18.6
18.4
18.5
18.1
18.4
24.9
18.4
18.2
5.8
Overall
weighted
average
Switzerland
3.1
United Kingdom
2,181
8.5
590
3.0
1,106
11.4
,341
4.3
142
0.6
23
0.9
160
21.5
United States
Weight (World
production, billion $)
22
Roodman, The Commitment to Development Index: 2004 Edition
Table 7. Calculations of production-distorting agricultural subsidies for CDI and of Total
Support Estimate of OECD, 2000–02
Australia Canada
National currency figures
A. Producer Support Estimate (PSE) 1,681
B.
Market Price Support (MPS)
43
C.
Payments based on output
50
D.
Payments based on area
37
planted/animal numbers
E.
“Counter cyclical payments”
0
F.
Payments based on historical
152
entitlements
G.
Payments based on input use 1,130
H.
Payments based on input
0
constraints
I.
Payments based on overall
269
farming income
J.
Miscellaneous payments
0
K. General Services Support
Estimate (GSSE)
L.
Research and development
M.
Agricultural schools
N.
Inspection services
O.
Infrastructure
P.
Marketing and promotion
Q.
Public stockholding
R.
Miscellaneous
1,032
678
0
88
236
10
0
21
EU
Japan
6,533 100,266
3,098 56,820
358
4,016
766 27,364
5,619
5,068
175
0
N. ZeaSwitzer- United
land Norway land
States Total ($)
147 20,065
101
8,464
0
2,852
0
3,241
7,666 46,972
4,510 16,630
355 7,345
882 2,461
0
845
0
606
0
0
0
0
0
0
0
1,269
584
8,527
506
0
7,998
3,812
259
117
44
0
4,350
670
336
115
7,264
1,917
903
0
0
2
488
0
2,244
57
–349
0
0
0
199
0
2,070
411
224
518
493
424
0
0
8,722
956
1,028
251
2,028
3,208
1,016
235
1,448
55
49
11
1105
26
46
155
212
121
6
59
25
0
0
1
1332
649
0
248
162
125
15
132
536 24,297
93 2,453
23
13
706
89 2,988
60 15,938
59
76
199
2137
S. Consumer Support Estimate
–194 –3,218 –49,127 –6,930 –112 –8,455 –5,302 3,800
(CSE)
T.
Transfers to producers from
–19 –3,097 –53,525 –5,067 –98 –9,575 –4,610 –16,627
consumers
Other transfers from consumers
0
–120
–251 –1,872 –13
–342 –1,079 –1,808
U.
Transfers to consumers from
–175
0
3,835
6
0
595
237 22,235
taxpayers
V.
Excess feed cost
0
0
815
3
0
867
150
0
W. OECD Total Support Estimate
(A+K+U)
2,538
8,603 112,823
X. Export subsidies (B+T)
Y. Other direct trade-distorting
subsidies (C+D+E+F/2+G+H)
24
1,293
1
2,053
1.83
1.53
Z. Exchange rate/$
Dollar figures
AA. OECD Total Support Estimate
(W/Z)
AB. Export subsidies (X/Z)
AC. Other trade-distorting subsidies
(Y/Z)
Total trade-distorting subsidies
(AB+AC)
1,387
7,073
359 21,992
8,439 93,504
3,295
43,493
1
551
3 –1,111
44 11,113
–100
3
2,323 23,835
1.09
0.08
2.25
8.59
5,623 103,507 88,413
160
2,560
5,146 93,504 300,300
1.64
1.00
13
705
1
1,338
3,029
39,987
13
7,205
1
20
–129
1,293
–61
3
1,413 23,835
3,060
75,796
718
1,338
43,017
7,218
21
1,164
1,352 23,838
78,666
23
Roodman, The Commitment to Development Index: 2004 Edition
Table 8. Overview of estimation of ad valorem tariff-equivalent of subsidies
Tariff
Imports/ Imports/
Subsidy equiva- (produc- consumpTariff
tion at
tion at Import
Agricultural Estimated rate (sub- lent of
border farmgate world
commodity trade-dis- sidies/
price equivalent
prices)
prices elastic- of subsiproduction torting sub- produc- meas3
1
1
1
2
dies (%)
sidies ($) tion,%) ures (%)
(%)
(%)
($)
ity
12.4
Australia
19,866
718
3.6
2.3
3.5
5.2
3.4
17.5
Austria
4,021
1,496
37.0
45.3
10.0
13.6
3.1
15.6
Belgium
4,156
992
24.0
45.3
10.0
13.6
3.1
7.7
Canada
20,516
1,338
6.5
36.0
27.8
51.3
1.8
14.4
Denmark
6,132
1,127
18.0
45.3
10.0
13.6
3.1
20.1
Finland
1,796
1,649
92.0
45.3
10.0
13.6
3.1
15.1
France
47,260
10,000
21.0
45.3
10.0
13.6
3.1
15.7
Germany
27,149
6,575
24.0
45.3
10.0
13.6
3.1
13.8
Greece
13,576
2,226
16.0
45.3
10.0
13.6
3.1
17.1
Ireland
4,571
1,528
33.0
45.3
10.0
13.6
3.1
12.2
Italy
43,146
5,278
12.0
45.3
10.0
13.6
3.1
5.2
Japan
116,443
7,218
6.2
118.8
15.9
25.8
2.7
12.0
Netherlands
13,876
1,646
12.0
45.3
10.0
13.6
3.1
2.1
New Zealand
6,331
21
0.3
3.1
4.7
15.5
3.0
16.5
Norway
2,062
1,164
56.4
198.1
38.4
60.9
1.4
13.9
Portugal
4,793
799
17.0
45.3
10.0
13.6
3.1
13.1
Spain
35,650
5,133
14.0
45.3
10.0
13.6
3.1
18.1
Sweden
2,102
919
44.0
45.3
10.0
13.6
3.1
11.8
Switzerland
4,460
1,352
30.3
137.3
43.6
58.1
1.5
15.2
United Kingdom 16,676
3,650
22.0
45.3
10.0
13.6
3.1
13.8
United States
195,753
23,8938
12.2
9.4
9.3
10.7
3.2
1
Assumed the same for all European Union members.
2
Computed as β = σD(1–φM), where β is the import price elasticity,σD is the elasticity of substitution in demand between domestic goods and imports, assumed to be 3.6, and φM is imports/consumption at world
prices. See Cline (2004, ch. 3).
3
Computed as

1
τ = 1 +
 1 + ϕ M (1 + t ) 1 + 1
s

(
1
)
β
 − 1,


where ϕM is imports/(production at farmgate prices), t is the ad valorem tariff equivalent of border measures, and s is the subsidy rate. This is nearly equivalent to equation A10 of Cline (2004, ch. 3, appendix
3A–2), differing only in that the elasticity β enters here in a more theoretically correct way. Cline’s formula
is equivalent to
τ=
1
1
(
β 1 + ϕ M (1 + t ) 1 + 1
s
), which is the first term in a Taylor expansion of the above
formula.
24
Roodman, The Commitment to Development Index: 2004 Edition
Table 9. Protection in ad valorem tariff equivalents, by sector (%)
Agricultural commodities
SubsiTariffs
dies
Total
Australia
2.3
12.4
14.7
Austria
45.3
17.5
62.9
Belgium
45.3
15.6
61.0
Canada
36.0
7.7
43.8
Denmark
45.3
14.4
59.7
Finland
45.3
20.1
65.5
France
45.3
15.1
60.4
Germany
45.3
15.7
61.1
Greece
45.3
13.8
59.1
Ireland
45.3
17.1
62.5
Italy
45.3
12.2
57.5
Japan
118.8
5.2
124.1
Netherlands
45.3
12.0
57.3
New Zealand
3.1
2.1
5.2
Norway
198.1
16.5
214.9
Portugal
45.3
13.9
59.2
Spain
45.3
13.1
58.4
Sweden
45.3
18.1
63.5
Switzerland
137.3
11.8
149.3
United Kingdom
45.3
15.2
60.6
United States
9.4
13.8
23.2
Other
Textiles
Apparel
agriculture Tariffs Quotas Total Tariffs Quotas Total Fuels Other
5.8 17.0
7.1 25.3 29.3
8.3 40.0 0.0 13.4
18.2
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
24.9 15.7
9.1 26.2 21.2
11.4 35.0 0.0 5.4
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.1
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.5
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.6
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.1
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.5
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
26.8
8.5
0.0
8.5 12.5
0.0 12.5 0.0 3.6
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
9.1
8.6
7.1 16.3 25.0
8.3 35.4 0.0 12.3
105.6 13.8
7.1 21.9 17.5
8.3 27.3 0.0 3.6
18.7
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.6
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
102.5
0.7
7.1
7.8 17.5
8.3 27.3 0.0 1.9
18.4
9.7
5.1 15.3 12.1
5.2 17.9 0.0 5.0
8.5 11.2
9.1 21.3 13.3
11.4 26.2 0.0 4.6
25
Roodman, The Commitment to Development Index: 2004 Edition
Table 10. Computation of Aggregate Measure of Protection (AMP)
Portugal
Norway
New Zealand
Netherlands
Japan
Italy
Ireland
Greece
Germany
France
Finland
Denmark
Canada
Belgium
Austria
Australia
58.4
59.2
214.9
5.2
57.3
124.1
57.5
62.5
59.1
61.1
60.4
65.5
59.7
43.8
61.0
62.9
14.7
18.4
18.6
18.7
105.6
9.1
18.4
26.8
18.5
18.1
18.6
18.4
18.5
18.1
18.4
24.9
18.4
18.2
5.8
7.8
15.3
15.3
15.3
21.9
16.3
15.3
8.5
15.3
15.3
15.3
15.3
15.3
15.3
15.3
26.2
15.3
15.3
25.3
17.9
27.3
17.9
17.9
17.9
27.3
35.4
17.9
12.5
17.9
17.9
17.9
17.9
17.9
17.9
17.9
35.0
17.9
17.9
40.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
5.0
1.9
5.0
5.0
5.0
3.6
12.3
5.0
3.6
5.0
5.0
5.0
5.0
5.0
5.0
5.0
5.4
5.0
5.0
13.4
5.1
5.3
5.1
5.1
5.1
4.8
3.5
5.1
5.0
5.1
5.1
5.1
5.1
5.1
5.1
5.1
3.7
5.1
5.1
1.5
5.4
6.3
5.4
5.4
5.4
8.9
4.7
5.4
8.0
5.4
5.4
5.4
5.4
5.4
5.4
5.4
4.5
5.4
5.4
4.0
2.7
1.7
2.7
2.7
2.7
1.2
3.5
2.7
1.5
2.7
2.7
2.7
2.7
2.7
2.7
2.7
2.4
2.7
2.7
3.3
10.3
9.9
10.3
10.3
10.3
13.8
9.0
10.3
8.8
10.3
10.3
10.3
10.3
10.3
10.3
10.3
7.2
10.3
10.3
6.4
14.7
7.6
14.7
14.7
14.7
4.4
17.5
14.7
28.1
14.7
14.7
14.7
14.7
14.7
14.7
14.7
8.4
14.7
14.7
14.9
AgriculAgriculOther
Other
tural
tural
com- agriculcom- agricul- Textiles Apparel Fuels
modities ture Textiles Apparel Fuels Other modities ture
61.8
69.1
61.8
61.8
61.8
66.8
61.9
61.8
48.7
61.8
61.8
61.8
61.8
61.8
61.8
61.8
73.8
61.8
61.8
69.9
Other
9.8
22.2
10.0
9.7
9.8
32.0
12.1
9.6
13.0
9.7
9.9
9.8
9.9
9.8
10.1
9.8
10.2
9.9
10.0
13.4
Aggregate
Measure
of Protection (%)
Shares of imports, 2000 (%)1
Spain
63.5
102.5
15.3
Protection (tariff-equivalents, %)
Sweden
149.3
18.4
U.K.
4.6
2.2
3.5
1.2
11.1
11.7
70.3
7.4
Switzerland
60.6
United States
23.2
8.5
21.3
26.2
0.0
Assumed the same for all European Union members.
1
26
Roodman, The Commitment to Development Index: 2004 Edition
Table 11. Revealed openness, 2002
Australia
Canada
European Union
Japan
New Zealand
Norway
Switzerland
United States
Imports (billion $)
A
B
C
D
Least developed
All low and middle
Weighted
countries only
income
Total
Total
total
Manufactures imports Manufactures imports (A+B+C+D)
0.05
0.09
12.69
17.99
30.82
0.16
0.40
24.68
32.11
57.35
6.85
11.72
263.11
401.28
682.95
0.29
1.55
86.46
144.40
232.70
0.01
0.02
2.22
3.13
5.37
0.10
0.12
4.51
5.99
10.71
0.06
0.10
6.38
8.38
14.92
4.82
9.37
371.66
468.31
854.15
GDP
411
716
8,545
3,980
58
189
268
10,400
Weighted
imports/GDP
(%)
7.5
8.0
8.0
5.8
9.2
5.7
5.6
8.2
Table 12. Computation of overall trade score
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Japan
Netherlands
New Zealand
Norway
Portugal
Spain
Sweden
Switzerland
United Kingdom
United States
Average
Standard deviation
Measured Revealed openness (25%
of score)
protection
(75% of
Transformed to Compos- Standardized
score
score)
ite
Raw value protection scale
----------------------- (%)----------------------4.4
13.4
7.5
13.1
13.3
5.8
10.0
8.0
10.1
10.0
5.8
9.9
8.0
10.1
9.9
5.7
10.2
8.0
10.0
10.1
5.8
9.8
8.0
10.1
9.9
5.7
10.1
8.0
10.1
10.1
5.8
9.8
8.0
10.1
9.9
5.8
9.9
8.0
10.1
9.9
5.8
9.8
8.0
10.1
9.8
5.8
9.9
8.0
10.1
10.0
5.9
9.7
8.0
10.1
9.8
3.4
13.0
5.8
23.2
15.5
5.9
9.6
8.0
10.1
9.7
5.9
12.1
9.2
2.5
9.7
–2.7
32.0
5.7
24.3
30.1
5.8
9.8
8.0
10.1
9.8
5.8
9.7
8.0
10.1
9.8
5.8
10.0
8.0
10.1
10.0
0.3
22.2
5.6
24.9
22.9
5.8
9.8
8.0
10.1
9.9
6.7
7.4
8.2
8.7
7.7
11.8
5.3
7.7
0.9
11.8
5.3
11.8
5.1
5.0
2.2
27
Roodman, The Commitment to Development Index: 2004 Edition
Investment
Investment flows from abroad have long played a major role in economic development—from
the 19th century in the United States to the 21st century in China. Source-country policies can affect capital flows, and given the magnitude of the capital flows—net foreign direct investment
from DAC to non-DAC countries was $104 billion in 2002 (DAC 2004)—relatively small policy
changes on the source side could make a significant difference for countries on the receiving
side.
But incorporating investment into the CDI is difficult for two reasons. First, not all investment is good for development, or at least is as good as should be. Prime examples include oil
industry ventures in Nigeria and Angola, and foreign-financed factories with inhumane working
conditions.
Second, the role of rich-country policies in stimulating and guiding investment is subtle.
In 2003, a major CDI investment indicator was net FDI in non-DAC countries as a share of
source-country GDP. FDI, of course, is not a policy, but an outcome, and an outcome in which
poor-country policies and private decisions figure much more than rich-country polices. FDI was
nevertheless used for a lack of more direct measures of policy, and out of a belief that it was better to include an imperfect component than no component at all. It would still raise public
awareness about the policy-investment-development link, and highlight gaps in current knowledge.
For 2004, Theodore Moran (2004) has completely overhauled the investment component.
Moran’s approach, adopted without modification into the CDI, dispenses with FDI and puts in its
place a more qualitative survey of government policies—a checklist approach. Countries can
gain or lose points based on the answers to 18 distinct questions. A perfect score would be 100.
For example, countries get 20 points for having programs to insure nationals against political
risks for investment in developing countries. But they lose 4 if they do not screen for and monitor environmental, labor, and human rights problems.
Paraphrasing Moran (2004), the 18 questions fit into five categories, covering:
1)
provision of quasi-official political risk insurance (with careful screening
mechanisms regarding impact);
2)
procedures to prevent double taxation of profits earned abroad—taxation, that
is, in both source and receiving countries;
3)
actions to prevent bribery and other corrupt practices abroad;
4)
measures to support foreign investors moving to developing countries, including
help for developing countries creating investment promotion agencies, assistance with identifying investment opportunities, and appropriate mediation to
resolve investment disputes; but also factoring in whether countries advocate
against receiving countries applying labor, environmental, or human rights standards to FDI;
28
Roodman, The Commitment to Development Index: 2004 Edition
5)
policies that affect portfolio flows, namely, restrictions on pension fund investment in developing countries and—working in the opposite direction—support
developing countries designing securities institutions and regulations.
The first four categories, worth a total of 80 points, pertain to foreign direct investment.
The last, with 20 points, obviously relates to portfolio flows. (See Table 13 for the scoring system and Table 14 for the results.)
The shift since 2003 in focus from outcome to policy transformed the results. It lifted
Australia 18 spots, Germany 11, the United Kingdom 10, the United States and Italy 9, and Canada 8. On the inevitable flipside are Austria, Belgium, Spain, Sweden, and Switzerland. Spain,
for instance, has been a major investor in its former colonies in the Americas, and so scored extremely high last year. Moran’s data suggest that Spain’s policy support for such investment is a
bit below average (it scores 60, while the 2004 average is 66), with the historical linkage playing
a much bigger role. (See Table 15, which shows investment ranks for 2003 and 2004. Because of
the change in scaling system, it is not meaningful to directly compare scores from the two years.)
Ireland stands out at the bottom end of the standings, coming in last place with only 36 points.
Perhaps because until recently it had viewed itself as lagging economically, its policies are
strongly oriented toward keeping capital at home.
29
Roodman, The Commitment to Development Index: 2004 Edition
Table 13. Investment component scoring system
Factor
Official provision of political risk insurance
Member of multilateral insurance agencies?
Official national insurance agency?
Deduct if agency does not monitor environment/ labor/ human rights standards.
Deduct if investors in some sectors are ineligible.
Deduct if agency has inappropriate national economic interest tests.
Deduct if agency extends coverage to inefficient import substitution projects.
Deduct if agency restricts eligibility to firms majority-owned by nationals.
Prevention of double taxation
Avoids double taxation with a tax sparing agreement or foreign tax credit or bilateral
investment treaties?
Deduct if does not allow investors to enjoy developing country tax incentives o, rates, or
exemptions.
Prevention of corrupt practices
OECD anti-bribery convention -- Phase 1 completion/Phase 1 implementation/Phase 2
completion?
Participation in Extractive Industries Transparency Iinitiative workshop?
Bribe Payers Index score quintile
Other FDI facilitation measures
Official assistance in resolving investment disputes with host countries?
Official assistance in identifying investment opportunities?
Official assistance to developing country export promotion agencies?
Deduct if country advocated against labor, environmental or human rights standards
Facilitating portfolio investment
Official support for design of developing country securities institutions and regulations?
No restrictions on pension fund investment in emerging markets?
Maximum possible points
Maximum
addition or
deduction
10
20
–4
–3
–3
–2
–2
20
–6
14
2
4
4
4
2
–5
6
14
100
30
Roodman, The Commitment to Development Index: 2004 Edition
Table 14. Summary of Investment Component
Doesn't let investors enjoy developing country tax incentives?
OECD convention -- participation level?
Extractive Industries Transparency Initiative Conference
participation?
Bribe Payers Index Score
Quintile
Official assistance in resolving
investment disputes?
Help businesses identify investment opportunities?
Help developing countries set
up investment promotion
agencies?
Negative advocacy
Official support for design of
securities institutions and
regulations?
No restrictions on pension fund
investment in emerging markets?
Total
14
86
0
0
0
4
4
4
2
12
0
0
58
0
2
0
4
4
4
0
10
–6
0
57
0
0
0
4
0
3
2
12
–6
14
84
0
0
0
0
4
4
2
14
0
0
64
0
2
0
4
4
2
0
12
–6
0
68
0
2
0
4
0
2
0
14
0
0
62
0
0
0
0
4
2
2
14
–6
14
89
0
2
0
4
4
3
2
14
0
0
54
0
0
0
4
4
2
0
12
–6
14
36
0
0
0
0
0
2
0
10
0
14
71
0
0
0
4
0
1
2
12
0
0
61
0
0
–5
4
4
2
2
12
0
14
89
6
2
0
4
4
3
2
12
0
14
38
0
2
0
0
0
2
0
10
0
14
70
0
2
0
2
0
2
2
14
–6
14
75
0
2
0
4
0
2
0
12
0
0
60
0
2
0
4
4
2
0
10
–6
0
50
0
0
0
0
4
4
0
10
–6
14
62
0
0
0
4
4
4
0
10
–6
14
85
0
0
0
4
4
3
2
12
0
7
75
6
2
0
0
4
2
2
14
0
GerNeth New
SwitJa- er- Zea- Nor- Port- Spai Swe- zerAus- Aus- Bel- Can- Den- Fin- Fran man Gree IreFactor
tralia tria gium ada mark land ce
ce land Italy pan lands land way ugal n den land U.K. U.S.
y
10 10 10 10 10 10 10 10 10 10 10 10 10
5 10 10 10 10 10
10 10
Multilateral Insurance?
Official national agency?
20 20 20 20 20 20 20 20 20
0 20 20 20
0 20 20 20 20 20
20 20
Agency monitor environment/
labor/ human rights?
0 –2
0
0 –4
0
0 –2 –2
0
0
0
0
0 –2 –2
0
0 –4
0
0
Investors in all sectors eligible?
0 –3
0
0
0
0
0
0 –3
0 –3 –3
0
0
0 –3
0 –3 –3
0 –3
No inappropriate national eco0 –3
0
0
0
0
0
0 –3
0 –3 –3
0
0
0
0
0 –3 –3
0 –3
nomic interest tests?
Restrict extending coverage to
inefficient import-substituting
–2 –2 –2 –2 –2
0 –2 –2 –2
0 –2 –2 –2
0 –2 –2 –2 –2 –2
–2 –2
projects?
International companies with a
0
0
0
0
0
0
0
0
0 –2
0 –2
0
0
0
0 –2 –2 –2
–2 –2
significant presence in this
Avoids double taxation?
18 18 18 18 18 18 20 18 18
0 18 20 16
5 14 18 18 18 16
20 18
Political risk insurance
Corrupt practices Double taxation
Other FDI
Portfolio
31
Roodman, The Commitment to Development Index: 2004 Edition
Table 15. Comparison of 2003 and 2004 CDI investment ranks
Germany
Netherlands
Australia
United Kingdom
Canada
Portugal
United States
Italy
Norway
Finland
Denmark
France
Switzerland
Japan
Spain
Austria
Belgium
Greece
Sweden
New Zealand
Ireland
2003 rank 2004 rank
12
1
4
1
21
3
14
4
13
5
3
6
15
6
17
8
7
9
10
10
9
11
11
12
1
12
16
14
2
15
5
16
8
17
18
18
6
19
20
20
19
21
Migration
Migration is one of the thornier topics covered in the index. Though it is widely agreed that the
effects of migration and migration policy on development are great and could become much
greater, they have not been as extensively studied as those of aid and trade polices. There is no
widely accepted analytical framework from the perspective of development, and little empirical
evidence. In addition, there are data problems, including lack of comprehensive information on
remittances and illegal immigration, and a paucity of internationally comparable information on
rich countries’ migration policies.
The CDI migration component is built on the conviction that migration advances development in source countries because it “provides immigrants with access to labor markets and
higher wages which, in turn, increase the potential for individual immigrants to remit money or
goods to the sending country…and enables migrants to establish migrant networks, which encourage continuous and expanding economic relations between sending and receiving countries.”
(Hamilton and Grieco, 2002)
In addition, freer flows of people, like freer flows of goods, should contribute to global
convergence in factor markets. The easier it is for a Vietnamese woman to get a job in Japan, the
more Nike will have to pay her to keep her sewing clothes in its Vietnam factories. And emigration of workers that are unskilled (by rich-world standards) should increase the wages of those
who do not leave by reducing labor supply. It should be said that while freer migration may directly benefit rich countries too, it can lower pay for nationals facing more intense competition
32
Roodman, The Commitment to Development Index: 2004 Edition
for their jobs. This is not a major consideration for the CDI, however, not because it doesn’t
worry us, but because the purpose of the CDI is to focus on effects on developing countries.
What happens when professionals leave—the so-called “brain drain”—is more complex.
The exodus of doctors and nurses from Ghana and South Africa has cost those countries. However, sometimes professionals gain skills abroad and then return home: Returned Indian expatriates are playing a big role in the software and services boom in Bangalore. Even when professionals remain abroad, they often retain links with industry and research at home. All this argues
at least for exploring how much rich countries discriminate in their immigration policies against
lower-skill workers. For the 2004 CDI, the Elizabeth Grieco and Kimberly A. Hamilton of the
Migration Policy Institution attempted to collect data from the 21 national governments on admissions by visa type, which might have allowed the CDI to make skills distinctions. (They also
hoped to gather data on whether entrants have legal access to the labor market, and for how
long.) But the task proved impossible. We hope that an international agency such as the OECD
will eventually step in to collect such information and make it internationally comparable.33
Grieco and Hamilton’s substantial data collection effort still led to major improvements
to the CDI migration component. In their paper (Grieco and Hamilton 2004), they proposed taking a weighted average of six indicators:
1) gross non-DAC immigrant inflow/receiving-country population;
2) gross non-DAC immigrant inflow/total immigrant inflow;
3) net migrant inflow over five years/receiving-country population—this includes inflows
from DAC countries too for lack of resolution in the data;
4) the difference between the unemployment rates for natives and immigrants, which is supposed to reflect barriers to immigrants entering the work force;
5) the share of foreign students that are from non-DAC countries; and
6) an index from the United Nations High Commissioner for Refugees (UNHCR) measuring
countries’ contributions to aiding refugees and asylum seekers.
The CDI adopts these recommendations, but with some substantial changes. Out of concern about conceptual overlap between the first three, it drops one—the first. The remaining two
it combines multiplicatively rather than additively. This seems to better capture the relationship
between them. As a thought experiment, if a country with a population of one million had a net
and gross immigration of 10 people, nine of whom were from developing countries, it would
score extremely well on indicator 2 because of the high non-DAC ratio (0.9) and extremely
poorly on 3 because its total flow/population would be essentially zero (0.00001). If the two indicators were put on a standard scale and averaged, the country itself would appear about average as far these two indicators went. If they were multiplied, then the resulting score would be
33
The IMF’s Balance of Payments Statistics Yearbook reports inflows and outflows of worker remittances by country and quarter. But the data are missing for many countries. And the inflow numbers are not disaggregated by
source, while the outflow numbers are not disaggregated by destination, making it hard to measure flows from individual rich counties to developing countries as a group.
33
Roodman, The Commitment to Development Index: 2004 Edition
essentially zero too—as seems appropriate given that only nine non-DAC immigrants entered a
country of a million people. The product of indicators 2 and 3—net inflow of migrants × DAC
share of gross inflow—can be seen as a rough estimate of the net inflow of non-DAC migrants.
The CDI also leaves out indicator 4, the unemployment rate difference. Higher unemployment among immigrants might actually reflect the greater attractiveness of a country’s labor
market to foreign workers. “Unemployment,” after all, is the state of not having a job, yet being
in the job market. If there are many immigrants “in the market for a job,” this could reflect policy
barriers to employment, which the CDI ought to penalize, or policies that facilitate entrance to
the market, which the CDI ought to reward. Because of this ambiguity in sign, it seems appropriate to leave this indicator aside until there is more evidence to validate it one way or the other.
The CDI adopts Grieco and Hamilton’s indicator 5, the share of the foreign student population that is non-DAC, without change. This deserves comment. Indicator 5 could mislead in
precisely the same way as 2: a country could admit almost no non-DAC students, yet have a high
non-DAC ratio if it admits even fewer DAC students. Japan is a case in point. Its 2001 non-DAC
student body was 60,687, which was 95% percent of its total foreign student body, the highest in
the sample. But that was only 0.05% of Japan’s population, which is barely above the 0.03% of
Italy and Portugal, which are lowest on this measure, and far behind Australia’s 0.47%. The essential question is, which indicator is more likely to capture differences in policy—non-DAC
students/total foreign students or non-DAC students/total population? For students much more
than unskilled workers, language is likely to be a major non-policy barrier, and probably does
much to explain Japan’s low foreign student numbers across the board. It seems more meaningful, then, to abstract from the predominantly non-policy factors that reduce the foreign student
body altogether, by taking foreign student population as the denominator.
Finally, as in 2003, the CDI uses a simplified version of the UNHCR index. The CGD
version is computed as total of three quantities, all taken over receiving-country GDP: the number of refugees hosted domestically; the number of other people “of concern” to UNHCR, such
as those internally displaced; and the number of asylum applications taken.34
Accepting the considered judgment of Grieco and Hamilton (2004), the student population measure gets 15% weight and the modified UNHCR index gets 20%. The remaining 65%
goes to the product of indicators 2 and 3. Before combining these three measures, each is rescaled to have a sample average of 5.0. Table 16 shows the calculations.
As with investment, overhauling the migration component rearranged the standings. Canada, the United States, and Australia, all historically nations of immigrants, have shot into the top
three spots, displacing New Zealand, Switzerland, and Germany. Switzerland in particular admits many people from developing countries—but almost as many are leaving. (Geneva diplomats and their nannies?) So it did well in 2003 when 90% of the migration score was based on
gross immigration from developing countries, but loses from the shift to net migration in 2004.
34
The UNHCR ranks all countries—not just rich countries—on the three indicators, averages the ranks, then reorders the countries and assigns final ranks.
34
Roodman, The Commitment to Development Index: 2004 Edition
Table 16. Summary of migration component
Refugee popu1
lation + asylum applications
Approximate net inflow of non-DAC Non-DAC stuimmigrants
dents
As % of
Net migrant
foreign StanNon-DAC inflow/1000
Per
Stanshare of population,
studard- billion $ dardStangross inized
PPP
ized
1995–
Prod dardized dents,
2
3
flow (%)
uct
score
2001 score GDP score Overall
2000
8.8
Australia
66
25.64 16.9
11.3
79
6.3 135.5
2.7
2.9
Austria
75
2.85 2.1
1.4
48
3.9 356.1
7.1
2.6
Belgium
48
6.08 2.9
1.9
47
3.8 199.0
3.9
11.2
Canada
92
23.86 22.0
14.6
58
4.7 246.4
4.9
6.1
Denmark
61
12.68 7.7
5.1
67
5.4 494.8
9.8
2.6
Finland
78
3.70 2.9
1.9
69
5.6 118.0
2.3
2.7
France
83
3.31 2.8
1.8
78
6.3 142.7
2.8
6.1
Germany
66
11.29 7.4
4.9
73
5.9 516.8
10.2
6.2
Greece
69
16.28 11.3
7.5
94
7.6
52.8
1.0
5.8
Ireland
44
24.27 10.6
7.1
26
2.1 228.7
4.5
3.6
Italy
93
7.03 6.5
4.3
59
4.8
10.7
0.2
1.9
Japan
81
2.22 1.8
1.2
95
7.7
1.1
0.0
5.9
Netherlands
74
10.25 7.6
5.0
49
4.0 511.8
10.1
5.0
New Zealand
78
10.61 8.3
5.5
81
6.6 109.4
2.2
4.9
Norway
58
10.21 5.9
3.9
57
4.6 414.1
8.2
2.8
Portugal
66
6.50 4.3
2.8
77
6.2
3.9
0.1
2.3
Spain
86
4.59 4.0
2.6
41
3.3
15.4
0.3
5.1
Sweden
67
5.06 3.4
2.3
47
3.8 775.3
15.4
3.6
Switzerland
50
3.06 1.5
1.0
34
2.7 644.7
12.8
4.4
United Kingdom
88
7.81 6.9
4.6
45
3.6 221.0
4.4
10.5
United States
93
22.89 21.3
14.1
77
6.2
99.9
2.0
5.0
Average
7.5
5.0
0.6
5.0 252.3
5.0
Weight
65%
15%
20%
1
2
”People of concern” to the U.N. High Commissioner for Refugees. Data for most recent
3
year available. Includes immigrants from other DAC countries.
35
Roodman, The Commitment to Development Index: 2004 Edition
Table 17. Comparison of 2003 and 2004 CDI migration ranks
Canada
United States
Australia
Greece
Germany
Denmark
Netherlands
Ireland
Sweden
New Zealand
Norway
United Kingdom
Switzerland
Italy
Austria
Portugal
France
Belgium
Finland
Spain
Japan
2003 rank 2004 rank
5
1
14
2
12
3
16
4
3
5
10
6
8
7
9
8
11
9
2
10
6
11
13
12
1
13
19
14
4
15
20
16
21
17
7
18
18
19
15
20
17
21
Environment
The environmental realm offers a wealth of potential indicators, but ones that are expressed in
various units. Considerations run from treaty ratifications to dollar amounts of subsidies to rates
of pollution. The approach I take in the environment component is therefore to organize indicators hierarchically—with groups and subgroups and sub-subgroups—put each indicator on a
standard mean-5 scale, and take weighted averages for input into the next level up. For example,
the climate stability subcomponent is a weighted average of greenhouse gas emissions per capita
and taxes as a share of gasoline prices. That in turn feeds into the depletion of share commons
component. In this component, as in the investment and migration components, weighting is often unequal.
At top level, the 2004 environment component is based on two components: depletion of
shared commons (67% weight) and contributions to international efforts—governmental cooperation (33%). Thus, it is divided between environmental impacts of rich-countries’ own activities on poorer countries, and rich countries’ efforts to help poorer achieve sustainable development. It takes the position that rich countries’ chief responsibility in this domain is to put their
own houses in order, as it were, because their use of shared environmental resources is vastly out
of proportion to their populations, and because they can set an example for the developing world
of clean, efficient commerce.
Following is an outline of the component’s construction, with some commentary. Roodman (2004b) contains more detail. Percentages indicate each variable’s weight at the next level
up:
36
Roodman, The Commitment to Development Index: 2004 Edition
1) Depletion of shared commons (67%)
a) Climate stability (67%)
i) Greenhouse gas emissions per capita (75%). The risks of climate change bear particularly on developing countries in part because they have less capacity to adapt. Climate
change could affect agriculture and aid in the spread of diseases such as malaria and
cholera. (Gross 2002) Population rather than GDP is the denominator in order to
avoid sending the odd message that the richer a country is, the more acceptable it is
for it to harm shared resources.
Emissions, of course, are not a policy but an outcome. But policies ranging from
land use planning to utility regulation do affect emissions. Unfortunately, at this
point, we lack good, comprehensive measures of greenhouse gas–relevant policies.
The one exception—an indicator with available data and great relevance—appears
next.
ii) Taxes as a share of gasoline prices (25%). This is new for 2004. Gasoline taxes are
indicative of motor fuel taxes in general (the other major fuel being diesel). And there
is a strong negative correlation across the sample of 21 countries between motor fuel
taxes and motor fuel use. (Roodman 2004b) Road transportation accounted for 23%
of OECD carbon emissions from fossil fuel burning in 2001 (OECD 2003, p. II.93),
thus about 20% of total greenhouse gas emissions. Considering that this measure has
the virtue of capturing policy rather than outcome, a 25% weight seems appropriate.
b) Ozone layer—consumption per capita of ozone-depleting substances (17%). The Montreal Protocol is phasing down consumption of many chemicals, but ozone-depleting
chemicals are still being used in rich countries.
c) Marine fisheries—fishing subsidies per capita (17%). Rich country-fleets are contributing
to overfishing in many international fisheries reducing the available resource for developing countries.
2) Contributions to international efforts—governmental cooperation (33%)
a) Participation in multilateral environmental institutions (50%). One point each for ratification of:
i) Kyoto Protocol on climate change. Finalized in 1997, this is the most important international effort to date to prevent climate change. It set important precedents by setting
emissions targets for industrial countries, and opening the way for international trading in emissions rights.
ii) Beijing Amendment to the Montreal Protocol on ozone-depleting substances. The international community has adopted a series of amendments strengthening the Montreal Protocol on Substances that Deplete the Ozone Layer. The Beijing Amendment is
the latest. Countries that have ratified it ratified all earlier amendments.
37
Roodman, The Commitment to Development Index: 2004 Edition
iii) Convention on Biological Diversity. This is one of the major contributions to international law of the 1992 “Earth Summit” in Rio de Janeiro, and a first step toward international cooperation to protect biodiversity.
b) Financial contributions to such international environmental institutions (50%)
i) Contributions to GEF’s third replenishment (50%). The Global Environment Facility,
housed at the World Bank, is a conduit for grant funding for projects in developing
countries with global environmental benefits, in areas relating to biodiversity loss,
climate change, degradation of international waters, ozone depletion, land degradation, and persistent organic pollutants. The measure used is countries’ contributions to
its third funding cycle, or “replenishment,” which began in 2002. This is divided by
donor GDP.
ii) Contributions to the multilateral fund under the Montreal Protocol, as a share of GDP
(50%). The multilateral fund is meant to help developing countries phase out consumption of ozone-depleting substances. The measure is total 2001 and 2002 contributions as a share of GDP.
It is worth noting that 44% of the component’s weight goes to outcomes measures. However, consumption of ozone-depleting substances accounts for 11% of this share, and it is now
strongly determined by policy. On balance, it is fair to say that one-third of the component is
based on an outcome measure (greenhouse gas emissions) with clear non-policy elements.
Table 18 shows the values of all indicators in the environment component. All of these
indicators are eventually transformed onto the standard mean-5 scale as they are combined with
each other, but for clarity standardized scores are not shown. Table 19 does show the full calculations at the top level of the hierarchy. The pattern of results is similar to last year’s. Three large
and relatively sparsely populated countries, Australia, Canada and the United States do poorly
overall, largely because of their high greenhouse gas emissions, the most heavily weighted indicator. Switzerland retains its top spot because of low emissions.
38
Roodman, The Commitment to Development Index: 2004 Edition
Table 18. Indicators used in environment component
Greenhouse
gas emissions/capita,
2002 (tons
of CO2
equivalent)
Ozone-depleting
substances
Share of consumptaxes in tion/capita,
premium
1999
Fishing
subunleaded (ozonegasoline depleting- sides/ GEF conprice,
potential– person, tributions/
1
2002 (%)
grams) 1997 ($) GDP (%)
Montreal
Protocol
contributions/GDP Kyoto
(%)
Protocol
Australia
28.3
51.9
44.1
1.8
0.0031
0.0012
Austria
10.7
64.1
53.3
0.0
0.0037
0.0017
Belgium
Canada
14.7
23.8
69.2
40.0
53.3
29.4
0.3
25.6
0.0053
0.0043
0.0016
0.0007
Denmark
Finland
12.9
15.5
69.7
70.0
53.3
53.3
11.5
4.8
0.0061
0.0000
0.0014
0.0015
France
Germany
9.5
12.1
73.7
73.4
53.3
53.3
1.9
0.6
0.0036
0.0039
0.0000
0.0011
Greece
Ireland
13.3
18.3
55.5
64.2
53.3
53.3
3.7
25.3
0.0014
0.0016
0.0005
0.0007
Italy
Japan
9.6
10.9
68.4
53.7
53.3
54.0
1.1
23.3
0.0000
0.0026
0.0013
0.0016
Netherlands
New Zealand
14.1
21.0
70.9
46.3
53.3
18.6
1.9
10.7
0.0000
0.0037
0.0015
0.0015
Norway
Portugal
12.4
9.3
70.0
68.9
27.4
53.3
36.4
3.8
0.0049
0.0015
0.0012
0.0010
Spain
Sweden
10.2
8.0
62.4
69.6
53.3
53.3
4.3
4.9
0.0000
0.0000
0.0015
0.0014
Switzerland
U.K.
7.5
11.0
64.3
77.5
4.9
53.3
0.0
1.7
0.0074
0.0035
0.0016
0.0002
United States
25.1
24.8
92.2
1
For 95 RON (Research Octane Number) unleaded.
3.2
0.0011
0.0006
Ratifications
Beijing
Amend- Convention on
ment to
Montreal Biological
Protocol Diversity
39
Roodman, The Commitment to Development Index: 2004 Edition
Table 19. Summary of environment component
Depletion of
Contributions to
shared commons international efforts
Australia
3.1
3.8
Austria
6.4
5.5
Belgium
5.7
6.2
Canada
1.5
5.7
Denmark
4.9
7.3
Finland
5.1
4.6
France
6.5
4.6
Germany
6.2
5.9
Greece
5.4
3.1
Ireland
2.4
3.5
Italy
6.5
3.4
Japan
3.8
5.9
Netherlands
5.7
4.6
New Zealand
3.8
6.4
Norway
2.8
6.6
Portugal
6.3
3.7
Spain
6.0
4.6
Sweden
6.4
4.5
Switzerland
7.8
8.1
United Kingdom
6.3
4.8
United States
2.3
2.1
Average
Weight
5.0
5.0
67%
33%
Overall
score
3.3
6.1
5.9
2.9
5.7
5.0
5.9
6.1
4.7
2.8
5.5
4.5
5.3
4.7
4.0
5.4
5.5
5.8
7.9
5.8
2.3
5.0
Security
Security—internal stability and freedom from fear of external attack—is a prerequisite for development. Sometimes a nation’s security is enhanced by the actions of other nations. So security
should be in the CDI. But as recent events have made obvious, one person’s liberation is another’s destructive intervention, so choosing what to count and what to reward is inherently controversial.
The 2003 CDI restricted itself to contributions to international peacekeeping, in part for
reasons of data availability. The 2004 version, done again under the guidance of Michael
O’Hanlon, and with substantial new data collection (O’Hanlon and de Albuquerque 2003), expands the scope to forcible humanitarian inventions. Examples include the Australian-led intervention in East Timor in 1999 to halt Indonesian repression after the territory had voted for independence. The component also expands the time frame considered, from two years to ten.35 The
rationale for this long period is that total government contributions to such operations is a particularly volatile variable—Kosovo’s and East Timor’s and Rwanda’s do not come along that
often. A decade of history gives more insight than two years into a government’s current capacity and willingness to intervene.
35
After O’Hanlon and de Albuquerque’s wrote their report, which covers five years, we expanded the data set to ten.
40
Roodman, The Commitment to Development Index: 2004 Edition
Because of the inherent controversy in choosing which rich-country interventions to reward, it seems essential for validity, in considering the universe of interventions over the last
decade, to apply either a weighting system in counting interventions—analogous to the aid component’s weighting based on recipient poverty and governance—or a filter, which is actually an
extreme form of weighting. The CDI follows O’Hanlon’s advice for a filter: it only counts operations that have been endorsed by an international body such as the U.N. Security Council,
NATO, or the African Union.36
To be precise, the 2004 edition security component, counts five costs, all taken as a share
of rich-country GDP:
1) Dollar contributions to the U.N. peacekeeping budget. These are averaged over 1997–
2002. Data were not available for 1992–96.
2) The cost of maintaining capacity for contributing personnel to U.N.-run peacekeeping
operations. To estimate this, a country’s peak personnel contribution to such operations
during 1993–2002 as a share of its standing military forces is computed. This percentage
is then applied to its military budget for the year.
3) The cost of deploying personnel in U.N.-run peacekeeping operations. This is estimated
at $9,000/person/month. (The full cost is estimated at $10,000, but the U.N. reimburses
contributing countries at the rate of about $1,000/person/month.) This too is averaged
over 1993–2002.
4) The cost of maintaining capacity for contributing personnel to peacekeeping and forcible
humanitarian operations that are not U.N.-run but receive international approval. This is
calculated in the same way as item 2. (Table 20 lists operations counted.)
5) The cost of deploying personnel in such non-U.N. operations—calculated the same way
as item 3, except using $10,000/person/month.
Two more aspects of the methodology need to be explained. First, in a departure from
O’Hanlon and de Albuquerque, all the tabulations incorporate a discount rate of 7%/annum,
equivalent to 50%/decade, on the grounds that a recent contribution is substantially more indicative of present policy stance than an old one. Thus the averages described above are weighted
averages, with each year getting 7% less weight than the next. And the peaks are discounted too.
Absent a time discount, we would be faced next year with the choice of either dropping 1993
data as we shifted the time frame to 1994–2003, which could introduce unrealistic discontinuities; or expanding the time frame, to 1993–2003, a choice that, if perpetuated for many years,
would create absurdities as ancient events received as much weight as current. The discounting
will allow us to formally expand the time frame while smoothly phasing out old data.
Second, 2003 data are incomplete and so excluded—a fact of huge importance in light of
the U.S.-led invasion of Iraq. That intervention will probably never be rewarded in the CDI pre36
The component excludes a pair of operations that technically make it through the filter: the U.S. and French
peacekeeping interventions in Rwanda immediately after the genocide and revolution in 1994. These interventions
were approved by the U.N. Security Council, but the overall behavior of rich countries with respect to Rwanda during the genocide was totally contrary to the spirit of this component.
41
Roodman, The Commitment to Development Index: 2004 Edition
cisely because it lacked an international imprimatur. However, on October 16, 2003, the U.N.
Security Council did give its mandate to the postwar multinational military presence in Iraq.
Unless there are major changes in methodology, or some exception is made, this will dramatically increase the U.S. score in future years.37,38
Table 21 has the 2004 security results. Once again, major changes in method caused major changes in scoring. (See Table 22.) Australia and the United Kingdom shot up to nearly tie
Norway for first place—Australia on the strength of its intervention in East Timor, the U.K. because of interventions in the former Yugoslavia and Sierra Leone, among others. The United
States climbed from 18th to 11th—perhaps still a surprisingly low ranking given its major interventions in Kosovo, Haiti, and Somalia in the last decade. But perhaps for the simple reason that
Yugoslavia was at their doorstep, many European nations contributed troops at levels that, adjusting for size, rivaled the United States.
There is some continuity with last year’s results. Norway took the top spot this year
thanks to high contributions to both U.N.- and non–U.N.-run operations; it held second place last
year. Switzerland and Japan continued to occupy the bottom spots. Japan has a strong constitutional and cultural commitment to peaceful conflict resolution. And Switzerland has an ancient
tradition of neutrality. It only joined the United Nations in 2002.
Table 20. Non–U.N.-run military operations counted in CDI security component.
Where
Afghanistan (postwar)
Albania (aid for Kosovo refugees)
1
Bosnia
When
2001–present
1999
1996–present
Major participants
France, Germany, Spain, U.K.
Italy
Austria, Belgium, Canada, Denmark, Finland,
France, Germany, Greece, Italy, Netherlands,
Norway, Portugal, Spain, Sweden, U.K., U.S.
Bougainville, Papua New Guinea
1998–present
Australia, New Zealand
Côte d’Ivoire
2002–present
France
East Timor
1999–2000
Australia
Egypt and Israel
1982–present
Spain, U.S.
Haiti
1994–95
U.S.
Kosovo (air war)
1999
Belgium, France, Germany, Italy, Netherlands,
U.K., U.S.
2
Kosovo (postwar)
1999–present
Germany, Italy, Netherlands, Norway, Portugal,
Spain, Sweden, U.K.
Iraq (Northern no-fly zone)
1997–2003
U.K., U.S.
Sierra Leone
2000
U.K.
Somalia
1992–93
U.S.
1
Includes implementation force (IFOR), stabilization force (SFOR), and operation Deliberate Forge.
2
Includes operation Joint Guardian and Kosovo Force (KFOR).
37
O’Hanlon has argued for excluding all military and security operations in Iraq on the grounds that they are motivated primarily by national security rather than development interests.
38
The full impact would be felt in 2006, when 2004 data were included. 2004 will be the first year during which the
U.S. troop presence will be under U.N. mandate for a full 12 months.
42
Roodman, The Commitment to Development Index: 2004 Edition
Table 21. Summary of security component—contributions as a share of GDP (%)
U.N.-run peacekeeping operations and
Non–U.N.-run PKO and
humanitarian interventions
humanitarian interventions
Cost of
Cost of
Contributions
maintaining
maintaining
to U.N.
personnel
peacekeeping personnel
Cost of using
Cost of using
capacity
personnel
capacity
personnel
budget
Australia
0.008
0.054
0.019
0.012
0.174
Austria
0.009
0.014
0.039
0.017
0.014
Belgium
0.009
0.013
0.018
0.038
0.040
Canada
0.007
0.030
0.017
0.023
0.049
Denmark
0.008
0.048
0.025
0.051
0.076
Finland
0.008
0.033
0.067
0.044
0.046
France
0.011
0.027
0.014
0.041
0.071
Germany
0.010
0.002
0.002
0.027
0.044
Greece
0.007
0.001
0.003
0.075
0.031
Ireland
0.005
0.045
0.094
0.008
0.012
Italy
0.009
0.002
0.001
0.038
0.056
Japan
0.009
0.003
0.000
0.000
0.000
Netherlands
0.008
0.030
0.015
0.051
0.084
New Zealand
0.008
0.096
0.058
0.013
0.022
Norway
0.007
0.080
0.046
0.054
0.086
Portugal
0.007
0.035
0.053
0.040
0.020
Spain
0.008
0.001
0.002
0.027
0.021
Sweden
0.009
0.029
0.020
0.027
0.048
Switzerland
0.000
0.003
0.001
0.003
0.013
U.K.
0.009
0.035
0.012
0.049
0.165
United States
0.006
0.004
0.001
0.018
0.114
Total
0.266
0.093
0.119
0.127
0.209
0.198
0.165
0.085
0.118
0.164
0.106
0.012
0.189
0.198
0.274
0.155
0.059
0.132
0.020
0.270
0.144
43
Roodman, The Commitment to Development Index: 2004 Edition
Table 22. Comparison of 2003 and 2004 CDI security/peacekeeping ranks
Norway
United Kingdom
Australia
Denmark
Finland
New Zealand
Netherlands
France
Ireland
Portugal
United States
Sweden
Canada
Belgium
Greece
Italy
Austria
Germany
Spain
Switzerland
Japan
2003 rank
2
10
15
3
13
4
11
7
9
5
18
19
17
11
1
6
16
8
13
21
20
2004 rank
1
2
3
4
5
5
7
8
9
10
11
12
13
14
14
16
17
18
19
20
21
Technology
Technology is an essential factor in development. Innovations in medicine, communications, agriculture, and energy meet societal needs, improve quality of life, increase productivity, and facilitate industrialization in poorer countries. Taking the very long view, a fundamental reason
that China’s economy can grow at rates of 7% or more for many years is because the country is
taking up innovations developed elsewhere over the last century.39 Vaccines and antibiotics led
to major gains in life expectancy in Latin America and East Asia in the 20th century, achieving
in four decades improvements that took Europe almost 150 years. Cell phones have revolutionized communications in poor countries such as Nigeria. The Internet also helps developing countries access and disseminate information, form civil society movements, and do commerce with
rich-world economies.
Thus people in developing countries benefit from technological advances as both producers and consumers. Recognizing the link between technology and development, the second edition of the index introduces a new technology component, developed by Bannon and Roodman
(2004).
Technology policy can be compartmentalized into two areas: technology generation and
technology diffusion. We found it much more difficult to measure policies relating to diffusion,
and so the inaugural edition of the CDI technology component focuses on technology generation.
In particular, the index does not measure how rich countries limit the diffusion of technology,
39
I thank Steven Knack for making the this point forcefully.
44
Roodman, The Commitment to Development Index: 2004 Edition
such as through intellectual property rights (IPR) protection. The current system of IPR protection, governed primarily by the WTO’s TRIPS agreement, restricts generic imitations of patented
products in even the poorest countries. This keeps prices of new products high, reducing access
for poor countries and limiting their ability to imitate (and learn through imitation). The CDI
doesn’t score rich countries on their IPR approach because rich countries’ policies seem largely
indistinguishable from each other—all have signed on to TRIPS. Moreover, it is quite possible
that the TRIPS regime helps some developing countries on balance while hurting others. The
very poorest countries, which could consume technological products such as HIV drugs, but are
not in a position to produce them, are probably hurt by strong IPR. But more industrialized developing countries may benefit from strong IPR, since it attracts FDI and creates incentives for
innovation—innovation that could be particularly suited for developing countries needs.
As a result, the 2004 technology component simply tabulates government financial support for research and development, counting both direct government funding of R&D, and the
value of tax incentives for it. The component discounts government funding for defense R&D by
50%. While some military innovations have useful civilian spin-offs (including the Internet),
much military R&D does more to improve the destructive capacity of rich countries than the
productive capacity of poor ones. The component does not reward countries for supporting research that targets the immediate needs of the poorest countries, such as the development of a
malaria vaccine. This is primarily due to data constraints: detailed breakdowns of R&D expenditure are simply not available for the 21 CDI countries. More broadly, it can often be difficult to
assess a priori which innovations will be adaptable and most advantageous to the poor in the long
term.
The calculations are therefore straightforward. The OECD publishes a variant of the
standard “B-index” that measures the rate of tax subsidization for business expenditure on R&D.
We use the simple average of the rates for small and large companies. On the OECD variant, 1
indicates full subsidization, 0 indicates no subsidization or taxation, and negative values indicate
taxation. The benchmark is full expensing. That is, a 0 means that the tax code treats R&D as an
ordinary expense, allowing it to be fully deducted from taxable corporate income in the year the
expenditure is made. If governments do not allow immediate full deduction, this is considered
taxation. Tax treatment more favorable than simple expensing is a subsidy. This tax or subsidy
rate is multiplied by a country’s total business enterprise expenditure on R&D (BERD) to generate an estimate of government tax expenditures on R&D.
To this is added direct government spending on R&D, whether performed by public
agencies or by private parties on contract, with military R&D discounted by half. The total is
taken over GDP. (See Table 23.)
Overall, Austria scores tops because of high direct expenditure on R&D. The United
States would score first but much of its R&D is defense-related; the defense R&D discount pulls
it down to ninth.
45
Roodman, The Commitment to Development Index: 2004 Edition
Table 23. Calculation of technology scores
A
B
C
D
E
Tax subsidy
Direct
rate for R&D,
manufacturers Business
Total
Of
Tax
government
R&D
(average expenditure expenditure
which: government
on R&D/ on R&D/ GDP expenditure/ military support/GDP
small/large
1
1
(%)
(%)
(%)
GDP (%)
companies) GDP (%)
A×B
Formula:
C+D×(1–E/2)
Australia
19.9
0.78
0.16
0.71
7.3
0.84
Austria
11.7
1.13
0.13
0.78
0.0
0.91
Belgium
–0.8
1.60
-0.01
0.46
0.4
0.45
Canada
24.8
1.10
0.27
0.61
4.8
0.87
Denmark
–0.9
1.33
-0.01
0.67
1.2
0.65
Finland
–1.0
2.43
-0.02
0.87
2.9
0.83
France
6.1
1.41
0.09
0.82
24.3
0.81
Germany
–2.5
1.75
-0.04
0.80
6.7
0.73
Greece
–1.5
0.20
0.00
0.33
0.9
0.33
Ireland
0.0
0.80
0.00
0.26
0.0
0.26
Italy
20.9
0.56
0.12
0.51
4.0
0.62
Japan
6.5
2.26
0.15
0.57
4.1
0.71
Netherlands
5.0
1.10
0.05
0.68
1.9
0.73
New Zealand
–2.3
0.43
-0.01
0.55
0.7
0.54
Norway
10.7
0.96
0.10
0.64
6.9
0.72
Portugal
33.5
0.27
0.09
0.51
2.0
0.60
Spain
44.1
0.50
0.22
0.38
37.3
0.53
Sweden
–1.5
3.31
-0.05
0.90
22.2
0.75
Switzerland
–1.0
1.95
-0.02
0.61
0.7
0.59
United Kingdom
10.1
1.28
0.13
0.57
30.5
0.61
United States
6.6
2.00
0.13
0.82
55.1
0.73
1
A 0 indicates that R&D spending can be fully deducted like other business expenditures. Positive
values indicate active subsidization relative to this benchmark. Negative values indicate
businesses cannot fully deduct in the year of expenditure.
III. Overall results
As explained in section I, the overall scores from each of the seven components are rescaled to
have a sample average of five—and then averaged across components to yield final scores. Table
24 shows the final results.
On the overall 2004 Commitment to Development Index, most of the Nordics and the
Netherlands do well, buoyed by large aid flows, high contributions to security, and lower pollution rates. The exception is Norway, pulled down by huge agricultural tariffs. Japan comes in last
because of barriers to exports and workers from developing countries, relatively low net concessional aid transfers, and low contributions to internationally approved security interventions.
The final column of Table 24 shows some continuity with 2003, but also some major
changes in standing. The Netherlands continues to hold first place (now in a tie with former
number-two Denmark) and Japan remains in 21st. Meanwhile, Australia, Canada, and the United
States have shot up about a dozen places, while New Zealand, Portugal, Spain, and Switzerland
have sunk about as much.
46
Roodman, The Commitment to Development Index: 2004 Edition
Since one purpose of the CDI is to track policy change over time, it is important to investigate the extent to which changes in measurement, as opposed to changes in what is measured,
are responsible for the substantial rearrangement since 2003. The revisions since last year can be
thought of in four steps: switching to a new scaling system for the final scores on each component (so scores have a mean of 5 rather than a maximum of 9)40; making numerous changes to
the methodology of the six original components, but still applying it to the 2003 database41,42;
adding the technology component (applying it to the 2003 database); and, finally, updating to the
2004 dataset. Table 25 shows how the scores changed during each step, on each component and
overall.
At the bottom of Table 25 are two measures of how different the scores after each step
are from the final 2004 scores. The first measure is the root-mean-square difference, taken over
the 21 countries. The second is the correlation. Because of the change in scaling system, it is not
meaningful to expect the final scores for the two years to match. The 2004 scores have an average of 5, by decree, where the 2003 scores do not. But one could expect them to covary, making
correlation the better comparison across that methodological divide. But once scores are scaled
the same way, starting with step B, the RMS difference becomes more meaningful.
The bottom right of Table 25 makes clear that revisions to the original six components
largely explain the changes in standing since 2003. The correlation with the final 2004 scores
rises slightly after rescaling, from 0.27 to 0.29, then jumps to 0.96 after incorporating the revisions in methodology. Adding in technology raises the correlation to 0.99. Switching to 2004
data completes the journey. The RMS numbers tell the same story.
To be more concrete, Australia, Canada, and the United States clearly benefited from the
shift in the migration component from gross to net, as well as the shift from monitoring investment flows to investment policies. Australia especially was helped by the expansion of the security component, which now takes cognizance of the country’s 1999 intervention in East Timor.
The same reasons lie behind the declines of New Zealand, Portugal, Spain, and Switzerland. Portugal and Spain, for instance, have been major direct investors in their former colonies in the
Americas. They scored high on the 2003 investment component that emphasized FDI. But
Moran’s data suggest that their policy support for such investment is about average.
Table 26 distills a few numbers from the foregoing, showing what policy change the new
methodology detects between 2003 and 2004. It shows the standings for 2004, and what they
would have been had the new methodology been applied last year—steps C and D in Table 25.
The main source of change is the aid component, specifically the raw quantity of aid given. Most
donors increased concessional disbursements as a share of GDP between 2001 and 2002, the
years on which the 2003 and 2004 aid components are based. France, the country that climbs the
most in this table, increased its DAC-measured net ODA from 0.32% to 0.38% of GDP, a 19%
40
For the “bad” components trade and environment, scores were first scaled between –9 and 0; then 10 was added to
them.
41
In transforming indicators onto the standard “mean-5” scale, 2004 scale factors were still used in order to allow
valid comparisons between 2003 and 2004 results. Thus standardized scores often do not average precisely 5.0.
42
Not all of the underlying data could be “downdated” to 2003 in doing this back-calculation. For example, the most
important data for the trade component, from the GTAP database, was for 1996–97 in both editions of the CDI. The
investment component was taken to be identical both years. The net migration rate was still estimated for 1995–
2000, not 1995–1999. Security scores were computed for the nine years in 1993–2001 rather than the more ideal
1992–2001.
47
Roodman, The Commitment to Development Index: 2004 Edition
rise. The broad rise in aid giving may reflect some combination of pledges made at the Monterrey Financing for Development conference in the spring of 2001, and the altered geopolitics after
the attacks of September 11.
Another important question about the results is how sensitive they are to changes in those
controversial component weights. To investigate the effect of raising weights on individual components, I generated 63 non-standard versions of the CDI: first with the weight on aid raised to 2,
then 3, and so on up to 10 (while weights on the other components were held at 1), then the same
for trade, and then the other components. For each version I calculated the correlation of overall
scores with the standard CDI, and the average absolute change in rank.43 Figure 1 and Figure 2
show the results. For most of the components, the CDI proves reasonably stable despite large
overweighting. Except for trade and environment, even tenfold overweighting yields a score correlation of 0.66–0.82. The index proves more sensitive to the weighting on trade and environment because those components’ scores are more idiosyncratic. On trade, the United States ranks
first, and the 14 scored EU members are tightly bunched; the correlation after overweighting
trade tenfold is 0.45. In the case of environment, the divergence appears driven in substantial part
by Switzerland’s strong number-one performance on the component—its 7.8 puts it well above
the next finishers, Austria and Germany at 6.1—which clashes for its tie for 18th on the overall
CDI.
Rank turns out to be as sensitive to increasing weights on aid, migration, and security as
it is to higher weights on trade and environment. This is because aid, migration, and security all
have wider ranges. As one of those components gains weight, its high outliers climb rapidly in
overall rank while its low outliers fall.
43
I am indebted to Michael Clemens for this technique.
48
Roodman, The Commitment to Development Index: 2004 Edition
Table 24. The Commitment to Development Index: Scores
Aid
Trade
Country
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Japan
Netherlands
New Zealand
Norway
Portugal
Spain
Sweden
Switzerland
U.K.
United States
2.9
3.7
6.0
3.6
12.3
5.0
6.0
3.9
1.8
3.0
2.8
2.4
11.2
0.8
10.6
2.3
2.0
12.4
5.8
4.8
1.9
4.4
5.8
5.8
5.7
5.8
5.7
5.8
5.8
5.8
5.8
5.9
3.4
5.9
5.9
–2.7
5.8
5.8
5.8
0.3
5.8
6.7
Average
Standard dev.
5.0
3.5
5.0
2.2
Invest- Migration Environ- Security Technol- Average 2004 2003
rank rank
ment
ment
ogy
5.9
6.5
8.8
3.3
9.0
6.4
4
19
4.7
4.4
2.9
6.1
3.1
6.9
12
9
4.6
4.3
2.6
5.9
4.0
3.4
13
12
5.8
6.3
11.2
2.9
4.3
6.6
6
18
6.7
4.8
6.1
5.7
7.1
5.0
1
2
5.2
5.1
2.6
5.0
6.7
6.3
11
17
5.3
4.7
2.7
5.9
5.6
6.1
7
14
5.3
6.7
6.1
6.1
2.9
5.6
7
6
4.1
4.1
6.2
4.7
4.0
2.5
17
13
3.9
2.7
5.8
2.8
5.5
2.0
18
15
4.5
5.3
3.6
5.5
3.6
4.7
14
15
3.2
4.6
1.9
4.5
0.4
5.4
21
21
6.7
6.7
5.9
5.3
6.4
5.5
1
1
4.3
2.9
5.0
4.7
6.7
4.1
16
4
5.3
5.3
4.9
4.0
9.3
5.5
7
10
4.5
5.6
2.8
5.4
5.2
4.5
14
3
3.7
4.5
2.3
5.5
2.0
4.0
20
6
6.1
3.8
5.1
5.8
4.5
5.7
3
8
3.9
4.7
3.6
7.9
0.7
4.5
18
5
5.9
6.4
4.4
5.8
9.1
4.7
4
11
5.3
5.6
10.5
2.3
4.9
5.5
7
20
5.0
1.1
5.0
2.5
5.0
1.3
5.0
2.4
5.0
1.3
5.0
0.9
49
Roodman, The Commitment to Development Index: 2004 Edition
Table 25. Detailed confrontation of 2003 and 2004 scores
Denmark
Canada
Belgium
Austria
Australia
3.1
3.0
3.5
2.8
1.7
A
3.3
4.9
4.6
2.6
5.4
4.4
2.7
B
1.5
1.4
3.5
3.8
4.1
2.8
4.0
2.8
2.7
3.0 6.6 5.4 5.4 5.8 2.3 3.9 2.7 2.7 4.5
1.8 6.7 5.4 5.5 5.8 0.0 0.0 4.1 4.1 1.6
6.0 6.8 5.6 5.5 5.8 1.7 2.9 4.7 4.7 0.8
5.0 6.8 5.6 5.4 5.7 1.7 2.9 5.1 5.1 1.3
3.6 6.6 5.3 6.0 5.7 2.1 3.6 6.3 6.3 6.1
6.0 6.7 5.5 5.4 5.8 1.4 2.4 4.3 4.3 4.5
3.7 6.8 5.6 5.4 5.8 2.6 4.3 4.4 4.4 6.5
2.9 7.2 6.1 4.7 4.4 1.6 2.6 6.5 6.5 3.7
5.6 6.7
2.0 6.6
3.9 6.8 5.6 5.4 5.8 1.4 2.3 6.7 6.7 8.1 10.3 6.3
9.0 14.0 14.8 12.3 6.8 5.6 5.5 5.8 1.0 1.7 4.8 4.8 4.4
1.7
2.3
1.0 2.5
1.6 2.6
5.6 6.1
8.2 2.3
4.7 8.3
5.8 1.6 2.5 2.6 2.8 3.7
6.1 6.0 6.4 5.8 6.1 3.8
2.7 4.9 5.4 5.4 5.9 5.2
2.6 5.4 5.9 5.6 5.0 2.9
6.1 5.0 5.5 5.0 5.7 7.1
2.6 4.5 5.1 5.6 5.9 3.5
2.9 5.4 5.9 5.9 6.1 2.6
8.8 1.8 2.7 3.2 3.3 2.8
6.7 3.0 3.6 4.7 4.7 3.6 4.1 4.1 4.2 4.5
4.7 5.9 5.5 2.0 2.0 3.6 4.3 4.1 3.8 3.9
6.2 4.6 5.1 4.7 4.7 9.0 11.4 3.5 4.0 2.5 2.5 3.9 4.4 4.3 4.0 4.1
7.7 11.0 11.2 1.7 2.6 3.0 2.9 2.4
5.6 3.2
3.6 5.3 5.8 5.6 5.5 5.3
4.8 2.9 2.9 5.5 5.6 4.7 5.4 5.1 5.2 5.3
6.6 5.0 5.6 6.2 6.1 3.8 4.4 4.5 4.7 5.3
3.7 7.0 6.7 6.4 6.3 3.5 4.0 5.0 5.2 5.2
9.0 7.4 7.1 5.0 5.0 5.5 6.9 7.3 6.9 6.7
3.0 4.0 4.3 6.5 6.6 3.4 4.1 5.5 5.7 5.8
4.4 3.8 4.0 3.4 3.4 4.0 4.7 4.4 4.3 4.6
3.2 3.1 3.1 7.0 6.9 4.4 5.3 4.0 4.4 4.7
3.6 9.6 9.0 6.4 6.4 3.2 3.7 5.8 5.9 5.9
Overall
B C' C D
Finland
2.1
4.0
1.3 3.4
TechEnvironment
Security
nology
A B C D A B C D C D A
France
1.5
2.8 7.0 5.8 5.5 5.9 1.5 2.5 5.3 5.3 1.1
Investment
Migration
A B C D A B C D
Germany
2.6
1.8
Trade
A B C D
Greece
2.2
D
Ireland
1.4
Aid
C
Italy
Sweden
Spain
Portugal
Norway
N. Zealand
3.3
2.2
3.7
3.4
6.6 10.3
1.7
5.1
7.0 10.8
2.4
4.7
5.0
5.7
1.7
2.3 6.9 5.8 5.5 5.8 9.0 15.0 5.6 5.6 1.0
9.7 10.6 1.0 –2.4 –2.3 –2.7 3.5 5.9 5.3 5.3 4.6
0.4
2.0 6.8 5.6 5.5 5.8 8.2 13.7 4.5 4.5 1.8
3.9 2.9
5.8 4.0 1.7 1.9 0.3 6.3 10.5 4.7 4.7 9.0 11.3 3.3
9.7 12.4 6.9 5.8 5.4 5.8 1.8 3.0 3.8 3.8 3.9
2.3
4.8 6.9 5.7 5.5 5.8 3.4 5.7 6.4 6.4 3.1
5.0 5.2
2.2 1.8
1.3 2.8
5.8 3.8
0.8 7.2 6.1 7.0 5.9 2.3 3.9 2.9 2.9 9.0 11.3 3.8
4.4 5.0 5.5 5.7 5.8 3.6
3.6 7.2 7.2 7.2 7.2 0.1
5.1 6.1 6.5 6.7 5.8 1.3
2.3 6.0 6.5 5.7 5.5 2.9
2.8 5.1 5.7 5.3 5.4 6.8
4.9 2.8 3.6 3.9 4.0 7.4
5.0 3.4 4.1 4.5 4.7 6.9
1.9 5.2 4.9 5.3 5.5 2.6 3.0 5.3 5.3 5.3
4.6 9.7 9.1 4.7 4.7 4.2 5.0 3.9 4.0 3.9
0.1 0.6 0.7 4.5 4.5 5.0 6.0 3.4 3.5 3.3
1.7 4.4 4.5 6.0 5.7 4.5 5.5 5.9 5.9 6.1
3.7 1.8 2.0 4.0 4.0 4.7 5.9 3.6 3.7 3.7
8.7 5.1 5.2 4.5 4.5 5.2 6.6 4.3 4.4 4.5
9.4 9.4 9.3 5.5 5.5 4.3 5.4 5.0 5.0 5.3
8.7 6.7 6.7 4.1 4.1 5.1 6.1 4.2 4.2 4.3
2.6
Switzerland
3.0
2.9 10.5 10.5 1.0 2.0 1.9 2.3 1.5
Japan
1.2 1.9 2.3 2.4 4.6 2.6 4.1 3.4 2.8 4.6 4.6 4.6 1.5 2.0 2.0 1.9 4.0 4.6 4.4 4.5 0.5 0.6 0.3 0.4 5.4 5.4 2.4 2.7 3.0 3.3 3.2
Netherlands 6.9 10.8 10.8 11.2 7.0 5.9 5.5 5.9 6.1 10.2 6.7 6.7 4.5 5.6 6.0 5.9 5.7 6.1 6.0 5.3 3.5 4.4 6.4 6.4 5.5 5.5 5.6 7.2 6.9 6.7 6.7
U.K.
1.9 7.7 6.8 7.2 6.7 2.0 3.4 5.6 5.6 2.3
1.4 1.3 0.3 0.2
.27 .29 .96 .99
1.6
0.1
1.00
1.3
2.7 2.9 0.3
.45 .45 .99
0.8
0.9 0.5 0.4
.94 .94 .95
U.S.
Comparisons with 2004 scores
Difference
(rms)
2.4
1.0
1.2
1.6
0.6
0.6
3.1
3.9
0.0
3.3 3.6 0.6
.96 .96 .96
.97 .97 .97
.15 .15 1.0
.23 .23 .98
Corr.
Key
A: Original 2003
B: Switch to new scaling: mean=5
C: Switch to new methology
C': Switch to new method, but exclude new component, technology
50
Roodman, The Commitment to Development Index: 2004 Edition
Table 26. Confrontation of 2003 and 2004 standings using 2004 methodology
1
2
3
3
5
5
7
8
8
10
11
12
12
14
15
15
17
18
19
20
21
2003 data
Denmark
Netherlands
Australia
Sweden
United Kingdom
Canada
United States
Finland
Germany
Norway
France
Austria
Portugal
Belgium
New Zealand
Italy
Greece
Ireland
Spain
Switzerland
Japan
6.9
6.7
5.9
5.9
5.7
5.7
5.3
5.2
5.2
5.0
4.7
4.4
4.4
4.3
4.2
4.2
4.0
3.8
3.7
3.5
3.3
1
1
3
4
4
6
7
7
7
7
11
12
13
14
14
16
17
18
18
20
21
2004 data
Netherlands
Denmark
Sweden
Australia
United Kingdom
Canada
United States
Germany
Norway
France
Finland
Austria
Belgium
Portugal
Italy
New Zealand
Greece
Ireland
Switzerland
Spain
Japan
6.7
6.7
6.1
5.9
5.9
5.8
5.3
5.3
5.3
5.3
5.2
4.7
4.6
4.5
4.5
4.3
4.1
3.9
3.9
3.7
3.2
51
Roodman, The Commitment to Development Index: 2004 Edition
Figure 1. Correlation of standard CDI with versions with higher weight placed on one
component
1.0
Correlation of scores
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
Tec
Secu hnolo gy
Envir rity
Migra onment
Inve s tion
tmen
Tra d
t
e
A id
0.0
1
2
3
4
Highe 5
6
r w ei g
ht
7
8
9
10
52
Roodman, The Commitment to Development Index: 2004 Edition
Figure 2. Average absolute change in CDI rank when higher weight placed on one component
Average absolute rank change
5.0
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
Tec
Envir hnolo gy
Secu onment
Migra rity
Inve s tion
tmen
Tra d
t
e
A id
0.0
1
2
3
4
Highe 5
r w ei g 6
ht
7
8
9
10
53
Roodman, The Commitment to Development Index: 2004 Edition
Appendix. Aggregation of Protection Levels across Sectors
In his proposal for an Aggregate Measure of Protection (AMP), Cline (2002) first estimates levels of protection in each of four major sectors, then aggregates the levels into a single
measure of protection. The CDI trade component draws heavily on Cline’s proposals, but departs
in some respects, including the method of aggregating across sectors. The method used in the
CDI is arguably more direct.
The starting point for our approach is Annex II of Cline (2002). There, Cline derives a
formula for the reduction in exporter’s producer surplus caused by a tariff. (See Figure A–1.)
Figure 3. Supply and demand for an import from a developing country
D
S
D’
e
t
b
f g
c d
a
D
D’
M
Q
S
In this diagram, line DD is the demand curve in the absence of protection, line D’D’ is
the demand curve in its presence, and SS is the supply curve. t is the ad valorem tariff and M is
the observed level of imports under protection. The increase in producer surplus, S, caused by
removal of protection is the sum of areas b, c, and d. Letting α and β be the absolute values of
the elasticities of demand and supply, Cline shows that this area can be estimated as:
S=M
(1)
tα
β tα
[1 +
].
β +α
2 β +α
This expression is quadratic in t, at base because high tariffs hurt exporters doubly, reducing both the price and quantity of their output. Cline notes that unless t is high, the quadratic term
54
Roodman, The Commitment to Development Index: 2004 Edition
is small. However, it turns out that for agriculture, levels of protection are quite high (reaching
215% for Norway in agricultural commodities), so this quadratic term can matter.
We generalize the above picture to multiple sectors. For simplicity, we assume that the
absolute elasticities of supply and demand, α and β, are the same for all sectors. In the generalization, there are protection levels ti, observed import levels Mi, and protection-induced losses in
producer surplus Si. In this partial equilibrium context, we estimate the total harm done by protection as the sum of the harms in each sector:
S = ∑ Si = ∑ M i
t iα
β t iα
[1 +
].
2 β +α
β +α
We then define the aggregate measure of protection to be the uniform tariff t that would
reduce exporters’ surplus by just as much as the observed set of protection levels. Our task then
is to solve for t in:
∑M
i
tα
tα
β tα
β t iα
[1 +
] = ∑ M i i [1 +
].
β +α
β +α
2 β +α
2 β +α
(The two sides of the equation differ only in that t is indexed on the right side.) Let
M
φi = i
M = ∑Mi
M . Dividing the above equation by M, canceling terms, and rearranging
and
gives:
t[1 +
1 tαβ
1 t iαβ
] = ∑ φ i t i [1 +
],
2 β +α
2 β +α
Now let Q = αβ/(α+ β), an expression that occurs on both sides of the equation.44 Substituting this in yields:
1
1
t[1 + tQ ] = ∑ φ i t i [1 + t i Q],
2
2
Using the quadratic equation to solve this for t, it works out that:
44
Q can also be written:
1
1
α
+
1
.
β
It is the reciprocal of the difference in slope between the demand and supply curves. (Recall that α is the absolute
value of the elasticity of demand, which is negative.) It represents the proportionate equilibrium change in exports
actually supplied with respect to the tariff level, as Cline shows: ∆M/M = tαβ/(α+β).
55
Roodman, The Commitment to Development Index: 2004 Edition
t=
1 
 Qt
1 + 2Q ∑ φ i t i 1 + i

2
Q

 
 −1 .
 
Using the Taylor expansion for 1 + x , it can be shown that at low tariff levels (as the
φt
ti→0), this formula converges to the simple weighted average of the tariffs, ∑ i i .
Now the ϕi and ti are observed variables. So to estimate t using this equation, we make
one more assumption, about the value of Q. We take it to be 0.5, which it is under the simple assumption that α = β = 1. Thus we calculate:

 t  
t = 2 1 + ∑ φ i t i 1 + i  − 1.

4  


This is the formula for the last column of Table 10.
56
Roodman, The Commitment to Development Index: 2004 Edition
Bibliography
Bannon, Alicia, and David Roodman (2004), “Technology and the Commitment to Development
Index,” Center for Global Development, Washington, DC, April.
Birdsall, Nancy, and David Roodman (2003), “The Commitment to Development Index: A
Scorecard of Rich-Country Policies,” Center for Global Development, Washington, DC,
April.
Center for Global Development and Foreign Policy (CGD and FP) (2003), “Ranking the Rich,”
Foreign Policy, May/June.
Bouët, Antoine, Lionel Fontagné, Mondher Mimouni, and Xavier Pichot (2001), “Market Access
Maps: A Bilateral and Disaggregated Measure of Market Access,” Working Paper 18,
Centre d’Etudes Prospectives et d’Informations Internationales, Paris, December.
Brown, Adrienne, Mick Foster, Andy Norton, and Felix Naschold, The Status of Sector Wide
Approaches. Overseas Development Institute. London. July 2001.
Cassen, Robert (1994), Does Aid Work? Report to an Intergovernmental Task Force, 2nd ed,
Clarendon Press, Oxford, U.K.
Chowdhury, Shyamal, and Lyn Squire (2003), “Setting Weights for Aggregate Indices: An Application to the Commitment to Development Index and Human Development Index,”
draft, Global Development Network, Washington, DC.
Cline, William (2002), “An Index of Industrial Country Trade Policy toward Developing Countries,” Working Paper 14, Center for Global Development, Washington, DC.
----------------- (2004), Trade Policy and Global Poverty, Center for Global Development and
Institute for International Economics, Washington, DC.
de Gorter, Harry, Melinda Ingco, and Laura Ignacio, “Domestic Support for Agriculture: Agricultural Policy Reform and Developing Countries,” Trade Note 7, World Bank, September 10, 2003.
Dimaranan, Betina V., and Robert A. McDougall (2002), Global Trade, Assistance, and Production: The GTAP 5 Data Base. Center for Global Trade Analysis, Purdue University
Easterly, William, Ross Levine, and David Roodman (2004), “Aid, Policies, and Growth: Comment.” American Economic Review, June, 94(3).
Grieco, Elizabeth M., and Kimberly A. Hamilton (2004), “Realizing the Potential of Migrant
“Earn, Learn, and Return” Strategies: Does Policy Matter?” Migration Policy Institute,
Washington, DC, February.
Jepma, Catrinus J. (1991), The Tying of Aid, OECD Development Centre, Paris.
57
Roodman, The Commitment to Development Index: 2004 Edition
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi (2003), Governance Matters III: Governance Indicators for 1996–2002, Policy Research Working Paper 3106, World Bank,
Washington, DC, August.
Kee, Hiau Looi, Alessandro Nicita, and Marcelo Olarreaga (2004), “Calculation of the Overall
Trade Restrictiveness Index,” World Bank, Washington, D.C. Processed.
Knack, Stephen, and Aminur Rahman (2004), “Donor Fragmentation and Bureaucratic Quality
in Aid Recipients,” Policy Research Working Paper 3186, World Bank, Washington, DC,
January.
Moran, Theodore (2004), “Rationale for Components of a Scoring System of Developed Country
Support for International Investment Flows to Developing Countries,” Center for Global
Development, Washington, DC, April.
O’Hanlon, Michael and Adriana Lins de Albuquerque (2004), “Note on the Security Component
of the 2004 CDI,” Center for Global Development, Washington, DC, April.
Organisation for Economic Co-operation and Development (OECD) (2003), Agricultural Policies in OECD Countries: Monitoring and Evaluation, Paris.
Picciotto, Robert (2003), “Giving Weight to the CGD Rankings: A Comment on the Commitment to Development Index,” Global Policy Project, London.
Roodman, David (2003), “An Index of Donor Aid Performance,” April, Center for Global Development, Washington, DC.
-------------------- (2004a), “An Index of Donor Performance: 2004 Edition,” Center for Global
Development, Washington, DC, April.
-------------------- (2004b), “An Index of Rich-Country Environmental Performance: 2004 Edition,” Center for Global Development, Washington, DC, April.
-------------------- (2004c), “Project Proliferation Costs: Modeling and Estimation,” draft, Center
for Global Development, Washington, DC.
Sawada, Yasuyuki, Hirohisa Kohama, Hisaki Kono, and Munenobu Ikegami (2004), “Commitment to Development Index (CDI): Critical Comments,” Discussion Paper on Development Assistance No.1, Foundation for Advanced Studies on International Development,
Tokyo, draft.
Walmsley, Terrie L., and L. Alan Winters (2003), “Relaxing the Restrictions on the Temporary
Movement of Natural Persons: A Simulation Analysis,” presented at conference on
“Quantifying the Impact of Rich Countries’ Policies on Poor Countries,” Center for
Global Development and Global Development Network, Washington, DC, October 23.
World Bank (2001), Global Economic Prospects and the Developing Countries 2002, Washington, DC.
58
Roodman, The Commitment to Development Index: 2004 Edition
--------------- (2004), Global Monitoring Report 2004: Policies and Actions for Achieving the
MDs and Related Outcomes, Washington, DC.
59