G-ECON - Yale University

advertisement
G-ECON
Creation and Analysis of a Geophysically Scaled Economic Data Set
May 2001
William Nordhaus
Yale University
I. Background
The last two decades have witnessed a dramatic growth in the
availability and analysis, at both national and international levels, of a wide
array of socioeconomic processes in parallel with development of data and
analysis of geophysical data. A major gap in current analysis is the virtually
complete disjunction of socioeconomic and geophysical analyses. In part, the
lack of intersection of the research programs has been due to the disparate
interests and descriptions of the two sets of groups working in these two areas.
Analysis of socioeconomic data has been largely undertaken by economists
and other social scientists concerned with “wealth of nations,” comparative
growth rates, and national economic policies. These analyses have naturally
relied upon socioeconomic data constructed from accounts largely free of any
geophysical dimension and sorted by political boundaries.1
The other group, largely consisting of natural scientists, has concentrated
on understanding geophysical processes. These have increasingly relied upon
direct earth or satellite observation of a wide variety of geophysical
phenomena such as weather or climate, ecological features, glaciology, and so
forth. The units of analysis in these geophysical studies are generally physical,
or “gridded” by latitude or longitude. Prominent example of such
geophysically-based studies are ones that attempt to project the climatic
consequences of the accumulation of radiatively active atmospheric gases.
There have been several attempts to link geophysical studies with
socioeconomic studies. Perhaps the most intensive have been the “integrated
The major exceptions are forestry and agriculture, which have sometimes combined geophysical
and political boundaries.
1
1
assessment” models developed in the climate-change community. (See
Nordhaus [1994] for an early version. A full review of existing integrated
assessment models as of around 1995 is contained in IPCC [1996].) These
analyses have linked the geophysical and socioeconomic modules by rescaling
the geophysical variables (such as climate and ecological impacts) into political
boundaries. For example, in the RICE-99 model, Nordhaus and Boyer [2000]
take the outputs of geophysically scaled climate variables and rescale them to 8
or 13 politically-defined regions.
For the most part, there have been no attempts to rescale from political to
geophysical definitions. This gap has occurred partly because the major
impacts being examined were economic ones along with national economic or
energy policies. Even more important, however, has been the complete lack of
an integrated socioeconomic data base scaled on a geophysical level. To put it
simply, it would be impossible to conduct an economic analysis as a geophysical basis
today because the underlying data do not currently exist.
The purpose of this project is to close this gap by developing a
geophysically scaled economic data set. The end product will be a global
Geophysically Scaled Economic Data Set (called G-ECON). The G-ECON
project will be a gridded data set at a 1 degree longitude by 1 degree latitude
resolution. This is approximately 100 km by 100 km, which is somewhat
smaller than the size of the major subnational political entities for most large
countries (e.g., states in the United States, Länder in Germany, or oblasts in
Russia). The main effort of this research will be to create data on gross regional
product (explained in detail below). In addition, we will merge the economic
data with other important demographic and geophysical data such as climate,
physical attributes, location indicators, population, and luminosity. More
details about the construction of G-ECON, the data sources, and the methods
are introduced below.
II. Utility of the Research
A. Value of the Data Set
The major point to emphasize is that geophysical scaling is a radically
different conception of socioeconomic data. Up to now, virtually all socioeconomic
data have been viewed outside of their geophysical context. As a result,
current macroeconomic research for the most part ignores the spacial and
2
geophysical dimensions of economic activity. G-ECON will be useful for a
wide variety of purposes. Some can be foreseen at the present time, while
others are either conjectural or cannot be forecast at present. This section
outlines a few of the potential uses of G-ECON.2
One research area is to examine the impact of geography on economic
activity. This research would consider the impact of climatic variables,
altitude, distance from coast, proximity to markets, and similar variables on
the density of economic activity. Because this research project is envisioned as
part of this research, we postpone a full discussion of this topic to part IV. This
topic has been of great interest for many decades and will be greatly informed
by G-ECON.
Another research area, which is likely to be undertaken by natural
scientists, will consider the impact of economic activity on important physical
activities. One important area will be to link environmental variables to
economic activity. Many environmental variables – such as deforestation,
pollution, or visibility – are conveniently observed on a geophysical scale, such
as by satellite data. The availability of G-ECON will allow statistical and other
analyses to trace the relationship between the economic and geophysical or
environmental phenomena. For example, it will be possible to trace the impact
of economic activity on pollution, deforestation, coastal degradation, and other
important environmental phenomena.3
More generally, G-ECON will facilitate a deepening of our
understanding of anthropogenic environmental phenomena. Until recently,
there has been very little spacial analysis of the relationship between economic
activity and environmental variables. For example, work on the
“environmental Kuznets curve” has uniformly relied on city or national
economic and environmental data.4 Particularly for local pollution,
geophysically based data are clearly preferable.
Several research areas on the interface between the environment and the social sciences are
reviewed in National Research Council [1992].
2
Studies of deforestation have begun to combine politically scaled and geophysically scaled
variables. See Meyer and Turner [1994], especially chapter 17.
3
4
See Grossman and Kreuger [1995] for an influential study.
3
One of the most important areas where G-ECON will be valuable is in
the investigation of climate change. Up to now, climate-change research has
taken gridded data and transformed it into politically scaled variables. (See
IPCC [1996] for a review of existing economic approaches. In most analyses of
impacts, climatic variables have been aggregated into national averages, which
then use conventional socioeconomic data.) In Mendelsohn, Nordhaus, and
Shaw [1994], the climatic data were converted to a county basis for the United
States. Also, Mendelsohn and Neumann [1999] estimated a variety of impacts
using politically scaled variables. However, once the conversion to politically
scaled variables takes place, there is no way to convert the results back into
geophysically scaled variables to link to ecological and other geophysical
models.
In working with G-ECON, two important characteristics of the data will
enhance its value relative to politically scaled data. The first is that the sheer
size of the data will allow much more refined analyses. While we have not
determined the exact number of relatively reliable data cells, our current
estimate is in the order of 20,000 land-based observations. This compares with
approximately 200 observations that are available with politically scaled
variables at the nation level. Such an increase in resolution is similar to the
availability of data from the Hubble telescope, which provided clear and crisp
answers to many previously difficult and fuzzily answered questions.
The second point is that gridded data are much more homogeneous in
scale than are politically scaled data. With political data, the largest 20
observations contain approximately 95 percent of either output or population,
and it makes little sense to conduct a statistical analysis with such disparate
elements as the United States (with 1995 GDP of $6176 billion) and Gabon
(with 1995 GDP of less than $1 billion).
As a final observation it should be emphasized that the data set proposed
here is likely to be the first rather than the last word. The history of innovative
data systems usually involves small-scale efforts by private researchers to
provide an example of how a particular data system might be constructed or
used. After the initial experience, if in fact the data set appears valuable, more
extensive and regular collection can be routinized and institutionalized within
government agencies.5 The creation of regular global and national G-ECON by
For example, the U.S. national income and product accounts were first developed by a private
researcher, Simon Kuznets, and were then lodged inside the government during the Great
5
4
statistics agencies of governments is entirely feasible once the general
principles are developed, and it could form part of the regular data collection
and processing activities of governments.
B. Relationship to New Techniques in Economics and Related Disciplines
The interaction between social and geophysical sciences has been
spurred by important developments in tools for analysis. New tools and
software—in particular, geographic information systems (GIS)—give the
opportunity to examine the relationship between geophysical variables such as
climate, location, and economic activity and to make more sophisticated
analyses of the complex relationships between geography and socioeconomic
variables.
GIS systems have three important characteristics. They process
geographically referenced systems; they relate locational data and
characteristics; and they are computerized.6 At present, GIS is widely used in
the geophysical sciences but is virtually unknown within economics. Although
GIS techniques are attractive for research at the intersection of economics and
the environment, there are many obstacles to such analyses. One set of
difficulties relates to the incompatible frameworks used by different scientists
to generate gridded data for use in GIS systems and, consequently, the
different data formats and the lack of a common framework.7
The most important obstacle, however, is the absence of available
economic data at adequate spatial resolution for use in GIS systems. Creation of
geophysically scaled databases that include the major socioeconomic variables at the
global and at national levels is an essential prerequisite for future analytical efforts.
Similar fruitful interactions can be made with “spatial statistics” and
“spatial econometrics,” which are critical for understanding the structure of
Depression when their value for understanding business cycles became clear. Similarly, private
efforts to develop environmental accounts by Nordhaus and Tobin [1972] and others (see Eisner
[1985]) were gradually developed by governments (see the recent report on environmental
accounting in Nordhaus and Kokkelenberg [1999]).
6
See Maguire, Goodchild, and Rhind [1991].
See Meyer and Turner [1994], pp. 412-413 for a discussion of different approaches to spatial
scaling in environmental and land-use models.
7
5
regional and geographical processes. Such techniques are widely used outside
of economics but are just beginning to creep into economics.8
III. Development of the Geophysically Scaled Economic Data Set (G-ECON)
This section describes the methods that will be used to construct
G-ECON. To date, a preliminary version of G-ECON has already been created.
The purpose of the current research project is to develop, test, and validate the
methods used to construct G-ECON (discussed in this section) and to use
G-ECON to investigate a number of economy-geography relationships
(discussed in the next section).
A. Analytical Background
In constructing a set of economic accounts, a primary question involves
the measures of output or income to be measured. G-ECON will be designed
to conform with standard principles of national accounting (see the SNA in
[1993]). The central measure will be gross regional product (GRP). This will
consist of the gross value added on a domestic product (not a national product)
basis. GRP consists of the output produced by the factors of production
located within the geophysical region. Under the principles of national
economic accounting, GRP will aggregate up across all cells within a country
to gross domestic product.
As in all accounting systems, we are faced with inaccurate and
incomplete data. The working assumption to be used in constructing G-ECON
is that per capita GRP is equalized within relatively small political area. We
thus calculate GRP by multiplying the population of a region times the
estimated per capita GRP of that region. The following comments will
elucidate the assumptions underlying the calculations.
1. One elemental component of the data set is the estimated population
of a region. These data are calculated from population censuses and are
generally relatively accurate (error rates of less than 3 percent in high-income
countries). Downscaling these to a gridded scale involves errors of attribution
to cells as well as seasonality. At present, we have no independent check of the
cell attribution, but it may be possible to validate the population estimates
For an example, see Fingleton [1999]. Spatial statistics are widely used in agriculture; see
http://www.dina.dk/~ml/dina/itcoord/.
8
6
when alternative measures of gridded population become available. This area
is not the major focus of this project, however.
2. Gross regional product consists of labor compensation and non-labor
incomes (the latter consisting of rents, interest, profits, and depreciation). In
most countries, labor’s share between 55 and 70 percent of gross output. The
analytical basis for the assumption of equal per capita GRP across cells is that
within regions people are likely to migrate in a manner that equalizes real
income or “utilities.” Utilities will be proportional to per capita GRP when (1)
labor’s shares is a constant fraction of GRP, (2) hours of work per capita are
equal, (3) amenities, transfers, and non-wage attractiveness do not differ
among regions, and (4) individuals migrate to equalize utilities in different
regions. None of these assumptions holds perfectly, but empirical research
indicates that they are likely to be reasonable close approximations to reality
within relatively small jurisdictions (such as within states of the U.S. or within
small countries like Belgium).
3. The major source of error in estimating GRP is likely to occur when
non-labor items are a particularly high fraction of output. Labor’s share is
particularly low in highly capital intensive industries such as electric utilities,
transportation, and communications. The procedure here would be vulnerable
to the assumptions if labor’s share were particularly low in geographically
sensitive industries like agriculture. In the United States, the share of
compensation in gross output of agriculture, forestry, and fishing in 1998 was
37 percent, as compared to 57 percent for gross domestic product. Note that the
share is even lower (25 percent) in finance, insurance, and real estate.
4. Given the high non-labor share in agriculture, it may be necessary to
consider adjustments for agriculture in regions that are heavily concentrated in
that sector. A further elaboration of G-ECON would include industrial
decomposition of output into the major sectors (agriculture, mining,
manufacturing, etc.). This decomposition is likely to be beyond the scope of the
present project but can be accomplished for the United States and some other
countries with existing data sets.
5. Several conceptual issues arise in developing G-ECON, although
these are common ones in constructing economic accounts. One interesting
issue is the production of fisheries. Even though fishing activities take place
over or in water, the output is attributed to land, usually to ports. A similar
7
issue arises for offshore oil and gas production. The numbers here are by and
large too small to be of concern for most high-income countries, however. For
the United States, the share of offshore oil and gas production is less than ½
percent of GDP, while fisheries is even smaller. These considerations may also
need to be considered in developing accounts for Mideast oil producing
regions. The issues raises by fisheries, minerals, and agriculture would be
largely resolved if G-ECON were developed on a major industry basis.
A further set of issues arise in countries with relatively weak statistical
systems. Many low-income countries have no regional economic data, so it will
be necessary to develop proxy data or to develop rudimentary regional data
for rescaling.9
A more significant issue involves subsistence agriculture and similar
activities in developing countries. These activities raise thorny conceptual
issues because the “non-market” activities are conceptually excluded from the
national accounts (because they are non-market). Many countries (including
the U.S.) do in fact include some adjustments for subsistence agriculture in
their national accounts because they are such close substitutes for market
activity and because their omission would distort growth-rate estimates. Our
G-ECON will follow standard procedures in these different areas.
A final set of issues relates to country-specific questions. Among the
issues needing resolution are disputed areas, the treatment of refugee camps,
the treatment of military expenditures, estimation of nomadic activities, and,
as mentioned above, the treatment of economic activity in lakes and oceans.
B. Data Sources
Most of the relevant data have been located. We describe the major data
sets in this section.
Grid Cell Data
We rely on grid cell data prepared by Environment Canada and the
United Nations Environment Program, as part of a project to develop a GIS for
A preliminary project working with data from Morocco indicates that developing relatively fine-detail
regional data for countries without official regional economic statistics is possible.
9
8
world population.10 A critical element in the work is “Rate-in-Grid” (or
“RIG”), which is necessary for separating grid points between countries and
between land and water. Some inspection has determined that these data are
relatively inaccurate. We are consequently developing an alternative technique
for estimating RIG.
Population data
One of the key ingredients of G-ECON is population distribution. The
population data set is available on a 1 degree by 1 degree latitude/longitude
grid system. The database uses the most recent data available (1990). It is likely
that gridded population data will be available for multiple years, and this will
in the future make G-ECON available on a time-series basis. The population
data set was developed by Environment Canada and the United Nations
Environment Program. It is based on aggregating from cities of population of
50,000 or more and is adjusted using a “rural population factor.” For some
countries (U.S., Canada, China, and the former Soviet Union) national census
estimates are used.
Some testing has determined that the original data include some
mistakes and omissions, which were fixed. In addition, we have made
estimates of population for a number of cells which were missing (including
“unpopulated” areas) to obtain a complete sample.
Gross Domestic Product Data
As discussed above, we propose to estimate gross regional product
(GRP) for each cell. These use population and estimated gross domestic
product (GDP) per capita for the nation or subnational region in which the cell
is lodged. Data on 1990 GDP per capita were gathered at the national level
calculated both in terms of market exchange rates and purchasing power
parity (PPP). With a few exceptions, the data come from the World Development
Indicators, 1998.
The national estimates provide a first-order estimate of economic
activity. There are, however, major differences in per capita income and output
within countries. For example, within the United States, there is a variation of a
The original data file can be downloaded from http://grid2.cr.usgs.gov/globalpop and
description can be downloaded from
http://grid2.cr.usgs.gov/globalpop/1-degree/description.html.
10
9
factor of four in real wages across counties. Even larger differences are found
in Brazil and Russia.
The major statistical issue involved in creating G-ECON will be to
“downscale” national data to the gridded level. After surveying a number of
approaches, it appears that the best approach is to use per capita output at the
first or largest subnational level (which we call “states,” which is the
designation in the United States and in many other countries). In most
countries, it appears that state date are the only subnational level at which
gross output (gross value added) data are regularly prepared. Given the
difficulties of estimating regional output, it is dubious whether a lower level of
disaggregation below states would produce reliable data, and indeed there are
serious reservations about the quality of the state data. To date, we have
collected state data for about half of the terrestrial grid cells, including major
countries like Brazil, China, India, Russia, and United States. Because states
use the national currency, there are no ambiguities about the currency
measurement. There are potential issues about the cost of living, but these
apply less to real output measures, which we are constructing, than to real
income measures, which we are not.
Gross regional product in each grid cell is calculated (both in market
exchange rates and in PPP terms) by multiplying gridded population data by
GRP per capita data. These measures are insensitive to the RIG measures.
However, economic density measure (output per square kilometer) will be
extremely sensitive to the RIG for many cells.
Downscaling
One of the major issues that is unresolved is the downscaling or “fitting”
of the political boundaries to the gridded boundaries. Downscaling is a major
issue in attempting to merge data sets or methodologies (see the discussion
relative to integrated assessment models of climate change).11 We have tested
alternative approaches (regression, triangular interpolation, and “majority
rule”) and will need further tests before settling on a final procedure.12
11
For a discussion of some downscaling issues, see The Second International Conference on Climate
and Water, Espoo, Finland, 17-20 August 1998 at
http://www.water.hut.fi/wr/caw2/report.html .
Mendelsohn, Nordhaus, and Shaw [1994] used a weighted regression technique for
interpolating among climate stations. While not yet widely adopted, we believe this technique is
probably superior to many of the existing alternatives.
12
10
Climate data
The climate data were generated by Leemans and Cramer, (see Leemans
and Cramer [1991]).13 We have used the updated 1994 Leeman and Cramer
data set for our purposes. For generating observations, Leeman and Cramer
used the following interpolation method. One weather station was chosen as
the typical weather station in each cell of the grid, on the basis of data
coverage, centrality of location, assessed quality of the data, and
representativeness of the cell's altitude. As a result they obtained 6,280 records
for temperature and 6,090 records for precipitation. The data values of all
typical weather stations were adjusted for the altitude of the weather station,
subtracting 0.6 degrees Celsius for each 100 meters of relative increase in
altitude. The grid was then triangulated. Using the triangulation, Leeman and
Cramer estimated the temperature and precipitation for cells with missing
observations via a smoothing technique. This gave them the temperature and
precipitation for a central data point in that cell. The authors express less
confident in the interpolated values for precipitation than they are for
temperature.
Luminosity Data
A final data set that will be included is nighttime luminosity. This is the
only existing geophysically scaled data set that relates closely to economic
activity. The nighttime satellite imagery data set data were provided by the
Department of Geography and Environmental Science of the University of
Denver.14 The measure included is a true radiance calibrated measure of light
intensity data. The data were provided on a 1 km x 1 km scale and were
aggregated by us to the 1 degree by 1 degree grid used in G-ECON.
Grid Cell and Land Area Data
This database uses climatic “normals” for the period 1931-1960. The available data coverage
for Europe, U.S., Japan, India, and Africa is judged to be good. For other regions, climate data
come from a sparser network. In addition, many coastline cells are left out of the data set, which
covers only terrestrial cells having at least 60 percent land.
13
Dr. Paul Charles Sutton, Department of Geography & Environmental Science, University of
Denver.
14
11
The grid cell area data (in square miles) were gathered from software of
MapInfo Corporation and contain the area of all the latitude and longitude
grid points
C. Assessing the Reliability of the Data
Aside from actually constructing G-ECON, one of the major tasks
of the present research will be to assess the reliability of the data. Virtually
none of the data are compiled in a routine and automatic manner; in each case,
as we described above, there is considerable exercise of judgment. The
following is a priority list of the major concerns with the underlying data sets:
• The major task will be to collect and validate the economic and
population data at the second administrative (“state”) level for the major
countries where this information is available. This also involves fitting
the economic data into a longitude-latitude grid scale. As noted above,
this has been accomplished for five countries, and we are at present
working with data for Kenya, Madagascar, Mexico, South Africa, and
Tunisia.
• To calculate economic densities will require accurate RIG data
(percentage of the area of the grid cell occupied by land) and RIC data
(percentage of the area of a country contained in each grid cell). Current
data appear to have significant errors, and it may be necessary either to
recalculate the RIC and RIG, or to find alternative sources for
triangulation.
• There are serious questions about the accuracy of the current
luminosity data. Sparsely populated areas have zero luminosity even in
the United States. We have not determined the extent to which filtering
or nonlinear smoothing may be responsible for these errors.
• The different data sets usually have different geolocational algorithm
estimates. The documentation is often unclear what is the geographical
center of pixels.
• The population data sets rely upon a variety of extrapolation
techniques and are therefore prone to error, particularly for rural areas.
Population estimates may simply be wrong (Tobler et al. [1995]), but they
12
are also based on different methodology (Elvidge et al. [1997] and
Dobson et al. [1999]). We will attempt to compare different approaches to
determine the reliability of the underlying estimates.
IV. Examination of the Relationship Between the Economy and Geography
A. Background
From the age of Aristotle until early in this century, most economists and
scientists reflecting on the wealth of nations assumed that climate and
geography were the prime determinants of the differences among nations.
And in economies in which most of the population lived on farms, this
presumption was probably correct. In 1820, 72 percent of U.S. employment
was on farms (by 1999, the total was down to 2½ percent).
The geocentric approach was epitomized by one of the most influential
geographers of the early twentieth century, Yale social-Darwinist geographer
Ellsworth Huntington, who argued that “The climate of many countries seems
to be one of the great reasons why idleness, dishonesty, immorality, stupidity,
and weakness of will prevail.”15 His studies ranged widely, but his central
hypothesis, the climatic hypothesis of civilization, held that climate ranks “as
one of the three great factors in determining the conditions of civilization.”16
His studies were based on international comparisons, on analyses of the rise
and fall of civilizations, and on a detailed examination of data relating
individual achievement to weather conditions.
Although “geography” and “economic geography” were vibrant fields
of research a century ago, they came under attack and were generally
abolished in many major universities (including the PI’s) in the middle of the
last century. The methods used were often crude, the ethnocentrism was
morally offensive, while the subjective nature of the research limited its value.
As a result, sometime around the middle of the twentieth century, geography
virtually disappeared from many research universities. In economics,
geography was eclipsed by other “modern” factors, such as investment,
technological change, trade policies, and education. (See Todaro [1991] as an
example; important exceptions are Kamarck [1976] and Lee [1957].)
International trade theory and practice today also are moving away from
geophysically based theories (an exception is Leamer [1984]).
15
Huntington [1915], p. 411.
16
Huntington [1915], p. 385.
13
In the last few years, however, geography and climate have reemerged
as new concerns surfaced and as geography took on a more scientific
approach. Particularly important was the growing interest in international
environmental issues such as global warming.17 Whereas in an earlier era,
geographers analyzed the influence of temperature, humidity, wind
movements, storminess, variability, and sunlight on comfort and economic
activity, today geographers and environmentalists worry about the role of
droughts, soils, sea-level rise, species depletion, vector-borne diseases, and
forest migration.
Geography is obviously important at high latitudes. Indeed there are no
important human habitations above 65º latitude in either hemisphere, and the
15 percent of the earth surface that is under permanent ice is essentially
worthless in the market place. The role of geography in tropical latitudes (or
climates) has long been thought crucial in leading to poor soils, unhealthy
conditions, and low productivity, although this is not firmly established.
Recent research within economics has returned to the earlier emphasis
on geography. Research on global climate change has examined the impact of
current climates and future climate change on economic activity (see the
summary in IPCC [1996]). Recent work by Gallup, Sachs, and Mellinger [1999]
analyzed the relationship between a number of economic variables and
geography on the basis of limited geophysical data. Mellinger, Sachs, and
Gallup [1999] undertook a similar exercise emphasizing water navigability.
These promising studies were hampered, however, by the lack of a reliable
data set on economic activity. Additionally, in most studies, climate is proxied
by a latitudinal variable rather than by actual climate data for different regions.
Current research on the economics of geography has suffered from three
major shortcomings. First, it has concentrated on politically scaled variables,
which are disparate in size and climatic distribution. For example, there are
only three major countries with significant regions with arctic climates while
there are more than twenty times that number with tropical or subtropical
climates. Second, the number of useful country observations is relatively
small, usually less than 100. Third, current studies have usually focused on
proxies for important geophysical variables; for example, virtually all studies
17
See particularly National Academy of Sciences [1992].
14
use latitude as a proxy for climate.18 While latitude is informative about
temperature, it is in fact a highly imperfect proxy variable.19 20
In the research using G-ECON, we can overcome each of these three
shortcomings exactly because most important geophysical variables are
recorded on the same scale as our economic variables in G-ECON. We can
therefore overcome each of the three shortcomings mentioned in the last
paragraph. First, because our unit of observation is the grid cell (1º latitude by
1º longitude), all land areas of the globe are well-represented. Second, the
number of observations is much larger than in typical economic studies, being
approximately 20,000 observations. Third, because we can merge our
G-ECON with standard geophysical data, we can use actual climate and other
geophysical data rather than proxy data.
We should note that the data are not without problems, and two in
particular should be noted. First, the output data are constructed from
national or subnational population and output data, and they will incorporate
any errors in the underlying data. As an example, we have relatively low
confidence in the output data for Russia and much of tropical Africa for the
period covered. Second, our data will cover only a single period (roughly
1990) and therefore are unable at this stage to be employed for dynamic or
growth analysis. We expect that if the effort to develop geophysically scaled
economic variables catches hold, it will be relatively straightforward to
develop time series for the major economic variables in G-ECON.
B. Estimation
Because we have not yet undertaken the research on estimating the
relationship between geography and the economy, we at this stage will only
sketch the basic outline of what is proposed.
18
See for example Hall and Jones [1996].
Most studies of economic geography examine at the impact of geography on per capita
income. Because migration tends to equalize per capita income within regions, we tend to
emphasize either economic density (output per unit area) or productivity (a measure of
economic density that adjusts for demographics) as more useful measures of the impact of
geography on economic activity.
19
For example, in the region between 35 degrees north and south, the R2 between mean
temperature and latitude is 0.21. In the tropic region between 15 degrees north and south, the
R2 between mean temperature and latitude is 0.004.
20
15
The basic idea is that the global economy is composed of a number of
different locations ( j = 1, ... , J ). Each location has a certain site-specific
productivity (A j ) which will be a function of its site-specific geophysical or
environmental characteristics (G j ), the population or labor force in that site (L j
), the capital stock of that site (K j ), plus the important institutional and policy
variables (P j ). These characteristics and inputs combine to produce the site's
output (Q j ):
(1)
Q = F(G, P, K, L)
where we have suppressed the superscripts. Assuming for concreteness that
(1) takes the Cobb-Douglas form, and that the other variables enter
multiplicatively, this yields:
(2)
Q = A(G, P) K a Lb
In this formulation we standardize on land area, so variables are averages per
unit land. Note that a + b < 1 to reflect diminishing returns with respect to
land. We have taken a simplified multiplicative formulation for expository
purposes, but this simplification will be relaxed in the actual empirical
estimates.
To make this approach operational, we need a few further assumptions.
We have no observations on capital. We therefore rely on the observation that
capital is mobile and assume that its net yield is equalized within countries.
Second, we will generally assume that labor is mobile within countries, so its
net wage is equalized within countries or regions. Solving these two
conditions yields:
(3)
Q = k ( ) A(G, P) s
Where k ( ) is a parameter that will depend upon countries or regional
groupings in which labor and capital is mobile, and s is a complicated
expression involving the parameters of the production function in (2). We can
estimate (3) by assuming that the site-specific productivity is a function of
observed geophysical characteristics G = (g1, g2, ...), which include climate,
altitude, etc. The policy and institutional variables will be treated either
explicitly (say for centrally planned v. market economies) or with the use of
national or regional dummy variables.
16
One end point of the analysis will be to estimate the “intrinsic”
site-specific productivity and its dependence upon geophysical, policy, and
institutional characteristics. Among the important questions we will address
are the following:
 One of the oldest questions in geography has involved the “optimal
climate.” What is the relationship between productivity and climate? Is
there an optimal climate from an economic point of view? What fraction
of the differences in productivity within nations or across nations is due
to climatic differences? Is climate more important in developing
countries or in high-income countries?
 One of the major issues that has dogged economic development for
decades has been the role of geography in economic activity. Do tropical
regions suffer an economic penalty because of their high temperature? Is
this offset in highly urbanized regions or in high-altitude tropical
regions? Once sectoral accounts are developed, it will be possible to
understand the relative importance of climate and geography in
agriculture and different sectors.
 Geography is often interpreted as climate. But other variables are
important. What is the role of coastal proximity, land-lockedness,
“rough” terrain, and insularity in economic productivity? Recent studies
have investigated the role of increasing returns, suggesting that
proximity to markets may be a crucial factor.
Undoubtedly, further questions will be investigated when the empirical
analysis is undertaken. The ability to examine these relationship with a data set
of approximately 20,000 regional observations, including a large number of
cells within major countries, will provide an enormous increase in the
resolution of our studies.
To provide reviewers an idea of the kinds of data and associations that
will be available with G-ECON, we show on the last page preliminary results
relating average temperature to GDP density for the U.S. Additionally, we
have made available 13 figures using preliminary data from the project. These
show scatter plots of GDP density, luminosity, temperature, precipitation,
altitude, and distance from coastline for the United States, India, and globally.
17
They are available at the web page http://www.econ.yale.edu/
~nordhaus/homepage/homepage.htm under “Geography and Economic
Activity.”
18
10000000000
1000000000
100000000
USA GDP density
10000000
1000000
100000
10000
1000
100
10
1
-20
-15
-10
-5
0
5
10
15
20
25
30
Average Temperature
Figure 1.
GDP density and average temperature for gridded data, United States
[Temperature in degrees C. GDP density is dollars per km2, 1990 prices]
19
References
Dobson, J.E., E. A. Bright, P. R. Coleman, R. C. Durfee, and B. A. Worley [1999].
“A Global Population Database for Estimating Populations at Risk,” Oak
Ridge National Laboratory, Oak Ridge, Draft Manuscript, 1999, accepted
for publication in Photogrammetric Engineering & Remote Sensing.
Eisner, Robert [1984]. “Extended Accounts for National Income and Product,”
Journal of Economic Literature, vol. 26, pp. 1611-1684.
Elvidge, C.D., K.E. Baugh, V.R. Hobson, E.A. Kihn, H.W. Kroehl, E.R.
Davis,and D. Cocero [1997]. “Satellite Inventory of Human Settlements
Using Nocturnal Radiation Emissions: A Contribution for the Global
Toolchest,” Global Change Biology, vol 3, pp. 387-395.
Fingleton, Bernard [1999]. “Economic Geography with Spatial Econometrics: A
‘Third Way’ to Analyze Economic Development and ‘Equilibrium’, with
Application to the EU Region,” EUI Working Paper ECO no. 99/21.
Food and Agricultural Organization [1995]. “The Digital Soils Map of the
World Version 3.5: Notes,” CDRom, Rome: FAO.
Gallup, John Luke and Jeffrey D. Sachs, with Andrew D. Mellinger [1999].
Geography and Economic Development, CID Working Paper no. 1,
Harvard University, March.
Grossman Gene M. and AlanB. Kreuger [1995]. “Economic Growth and the
Environment,” Quarterly Journal of Economics, vol. 112, pp. 353-378.
Hall, Robert E., and Charles I. Jones [1999]. “Why Do Some Countries Produce
so Much More Output per Worker than Others,” Quarterly Journal of
Economics, Vol. 114, pp. 83-116.
Huntington, Ellsworth [1915]. Civilization and Climate, Shoe String Press,
Hamden, CT.
IPCC [1996]. Intergovernmental Panel on Climate Change Climate Change 1995:
Economic and Cross-Cutting Issues. The Contribution of Working Group III to
20
the Second Assessment Report of the Intergovernmental Panel on Climate
Change, J.P. Bruce, H. Lee, and E.F. Haites, eds., Cambridge University
Press, Cambridge, UK.
Kamarck, Andrew N. [1976]. The Tropics and Economic Development: A
Provocative Inquiry into the Poverty of Nations, The Johns Hopkins
University Press, Baltimore, MD.
Leamer, Edward [1984]. Patterns of International Trade, MIT Press, Cambridge,
MA.
Lee, Douglas [1957]. Climate and Economic Development in the Tropics, New York,
Harper for the Council on Foreign Relations.
Leemans, R. and W.P. Cramer [1991]. The IIASA database for mean monthly values
of temperature, precipitation and cloudiness of a global terrestrial grid,
Research Report RR-91-18 November, International Institute of Applied
Systems Analyses, Laxenburg, 61pp.
(http://www.ngdc.noaa.gov/seg/eco/cdroms/gedii_a/datasets/a03/l
c.htm)
Maguire, D.J., Goodchild, M.F. and Rhind, D.W., eds. [1991]. Geographical
information systems : principles and applications, New York, Longman
Scientific and Technical.
Mellinger, Andrew D. and Jeffrey D. Sachs, with John L. Gallup [1999].
“Climate, Water Navigability, and Economic Development Geography
and Economic Development,” CID Working Paper no. 24, Harvard
University, September.
Mendelsohn, Robert, and James E. Neumann [1999]. The Impact of Climate
Change on the United States Economy, Cambridge University Press,
Cambridge, UK.
Mendelsohn, Robert, William D. Nordhaus, and Dai Gee Shaw [1994]. “The
Impact of Global Warming on Agriculture: A Ricardian Approach,”
American Economic Review, September, vol. 84, No. 4, pp. 753-771.
21
Meyer, William B. and B. L. Turner II [1994]. Changes in Land Use and Land
Cover: A Global Perspective, Cambridge University Press, Cambridge, UK.
National Academy of Sciences [1992]. Committee on Science, Engineering, and
Public Policy, Policy Implications of Greenhouse Warming: Mitigation,
Adaptation, and the Science Base, National Academy Press, Washington,
DC.
National Research Council [1992]. Global Environmental Change: Understanding
the Human Dimensions, National Academy Press, Washington, DC.
Nordhaus, William D. [1994]. Managing the Global Commons: The Economics
of the Greenhouse Effect, MIT Press, Cambridge, MA.
Nordhaus, William D. and Joseph Boyer [2000]. Warming the World: Economic
Modeling of Global Warming, with Joseph Boyer, forthcoming, MIT Press,
Cambridge, MA. A web version of the book is available at
http://www.econ.yale.edu/~nordhaus/homepage/dicemodels.htm .
Nordhaus, William D. and Edward C. Kokkelenberg, Editors [1999]. Nature's
Numbers: Expanding the National Economic Accounts to Include the
Environment, Report of the Panel on Integrated Environmental and
Economic Accounting, National Academy of Sciences Press,
Washington, D.C.
Nordhaus, William D., and James Tobin [1972]. “Is growth obsolete?” In
Economic Growth, 50th anniversary colloquium V. New York: Columbia
University Press for the National Bureau of Economic Research; also
published in M. Moss, ed., The Measurement of Economic and Social
Performance, Studies in Income and Wealth, vol. 38, 1973; New York,
Columbia University Press for the National Bureau of Economic
Research.
SNA [1993]. System of National Accounts, Commission of the European
Communities, International Monetary Fund, Organization for Economic
Cooperation and Development, United Nations, and World Bank,
Brussels and elsewhere.
22
Streeten, Paul [1972]. The Frontiers of Economic Development, Wiley, New York,
NY.
Tobler W., U. Deichman, J. Gottsegen, and K. Malloy [1995]. The Global
Demography Project Technical Report 1995-6, National Center for
Geographic Information and Analysis, Santa Barbara, CA.
Todaro, Michael P. [1991]. Economic Development in the Third World, Longman,
New York, NY.
23
Download