May 7 + rev
May 16, 2014
Seminar on “Fiscal Procyclicality and Over-optimism in Official Forecasts”
Seoul, Korea, May 2014
Jeffrey Frankel
Harpel Professor of Capital Formation and Growth, Harvard Kennedy School
(1) "Fiscal Policy in Developing Countries: Escape from Procyclicality,” VoxEU, June 2011.
(Summary of: “On Graduation from Fiscal Procyclicality Procyclicality,”
with Carlos Végh & Guillermo Vuletin, Journal of Development Economics 100, no.1,
2013; NBER WP 17619. )
(2) “Over-Optimism in Official Budget Agencies' Forecasts,” in NBER Digest, Nov. 2011. ( Summary
of "Over-optimism in Forecasts by Official Budget Agencies and Its Implications," Oxford Review
of Economic Policy vol.27, no.4, 2011. )
(3) FULL PAPER: "Bias in Official Fiscal Forecasts: Can Private Forecasts Help?" with Jesse
Schreger, Harvard University, 1st draft, June 2013. Revised, May 2014.
Fiscal policy is distressingly procyclical in many countries, expanding during upswings and
contracting during down swings. One reason for this may be the tendency for governments to
think that booms will go on forever and thus that they can afford tax cuts and spending increases.
Government forecasts of GDP growth and budget balances are generally overly optimistic,
especially in booms. Rules such as limits on the budget deficit do not solve the problem and in fact
make the forecast bias worse. Our most recent research finds that official forecasts are also more
over-optimistic than private sector forecasts, in a sample of 31 countries. When official forecasts
are especially optimistic relative to private forecasts ex ante, they are more likely to be overoptimistic relative to realizations ex post. Euro area governments during the period 1999-2007
assiduously and inaccurately avoided forecasting deficit levels that would exceed the 3% Stability
and Growth Pact threshold; private sector forecasters were not subject to this crude bias. Three
important implications. First, the severe recessions that began in 2008 do not explain the findings
of optimism bias. Second, the over-optimism behaves as if specifically designed so as to “game”
checks on budget deficits, rather than by politicians’ desire to “talk up” the economy in general.
Third, the budget-making process could probably be improved by using private-sector forecasts.
1
vox
Research-based policy analysis and commentary from leading economists
http://www.voxeu.org/article/how-developing-nations-escaped-procyclical-fiscal-policy
Fiscal policy in developing countries: Escape from procyclicality
Jeffrey Frankel, Carlos A. Vegh , Guillermo Vuletin, 23 June 2011
With the ongoing financial turmoil in Europe, many emerging market countries are now deemed less
risky than so-called “advanced” countries. This column examines why this is the case and finds that the
cyclicality of a country’s fiscal policy – a sign of its riskiness – is inversely correlated with the quality of
the country’s institutions.
Fiscal policy is taking centre stage. Among advanced countries, the news is bad; Europe’s periphery
teeters, the UK slashes, the US deadlocks, Japan muddles. But in the rest of the world there is good
news. In an historic reversal, many emerging market and developing countries have over the last
decade achieved a countercyclical fiscal policy.
In the past, developing countries tended to follow procyclical fiscal policy. They increased spending (or
cut taxes) during periods of expansion and cut spending (or raised taxes) during periods of recession.
Many authors have documented that fiscal policy has tended to be procyclical in developing countries, in
comparison with a pattern among industrialised countries that has been by and large countercyclical
(Gavin and Perotti 1997, Tornell and Lane 1999, Kaminsky et al. 2004, Talvi and Végh 2005, Mendoza
and Oviedo 2006, Alesina et al. 2008, and Ilzetski and Végh 2008).
Most studies look at the procyclicality of government spending, because tax receipts are particularly
endogenous with respect to the business cycle. Indeed, an important reason for procyclical spending is
precisely that government receipts from taxes or mineral royalties rise in booms, and the government
cannot resist the temptation or political pressure to increase spending proportionately, or more. We can
find a similar pattern on the tax side by focusing on tax rates rather than revenues, though cross-country
evidence is harder to come by.1
Figure 1 (which is a version of evidence presented in Kaminsky et al. 2004) depicts the correlation
between government spending and GDP for 94 countries over the period 1960-1999. More precisely, it
shows the correlation between the cyclical components of spending and GDP with the longer-term
trends taken out. The set includes 21 developed countries, which are represented by black bars, and 73
developing countries, represented by yellow bars. A positive correlation indicates government spending
that is procyclical, i.e. destabilising. A negative correlation indicates countercyclical spending, that is,
stabilising.
2
There is no missing the message. Yellow bars lie overwhelmingly on the right-hand side. More than 90%
of developing countries show positive correlations (procyclical spending). Black bars dominate the left
hand side. Around 80% of industrial countries show negative correlations (countercyclical spending).
Figure 1. Correlations between government spending and GDP, 1960-1999
Why would policymakers pursue procyclical fiscal policy? One does not have to believe in “fine tuning”
to see the undesirability of a pattern under which government response exacerbates the amplitude of
the business cycle. The most convincing explanations in the literature entail either imperfect access to
credit or political distortions.
The historic shift in cyclicality
Over the last decade there has been a historic shift in the cyclical behaviour of fiscal policy in the
developing world. Figure 2 updates the statistics, showing the period 2000-2009. The number of yellow
bars on the left side of the graph (negative correlations) has greatly increased. Around 35% of
developing countries 26 out of 73 now show a countercyclical fiscal policy, more than quadruple the
share during the earlier period.
3
Figure 2. Correlations between government spending and GDP, 2000-2009
Figure 3 presents a scatter plot with the 1960-1999 correlation on the horizontal axis and the 2000-2009
correlation on the vertical axis. The lower right quadrant shows the graduates from procyclical to
countercyclical fiscal policy. The star performers include Chile and Botswana; but 24 developing
countries altogether (out of 73) have made this historic shift.
Figure 3. Correlations between government spending and GDP, 1960-1999 vs. 2000-2009
4
The evidence of countercyclicality among many emerging-market and developing countries matches up
with other criteria for judging maturity in the conduct of fiscal policy, such as debt/GDP ratios, rankings
by rating agencies, and sovereign spreads. Low income and emerging market countries in the
aggregate have achieved debt/GDP levels around 40% of GDP over the last four years. The IMF
estimates the 2011 ratio at 43% among emerging market countries and 35% among low-income
countries. This is the same period during which debt in advanced countries rose from about 70% of GDP
to 102%.
The financial markets have ratified the historic turnaround. Spreads are now lower for many emerging
markets than for some “advanced countries.” Rating agencies rank Singapore as more creditworthy than
Belgium, Korea ahead of Portugal, Mexico ahead of Iceland, and just about everybody ahead of
Greece. Euromoney ranks Chile as less risky than Japan, Korea less risky than Italy, Malaysia less risky
than Spain, and Brazil less risky than Portugal.
Largely as a result of their improved fiscal situations during the period 2000-2007, many emerging
markets were able to bounce back from the 2008-2009 global financial crisis more quickly than
advanced countries (Didier et al. 2011).
How did they do it?
What explains the ability of some countries, particularly emerging-market and developing countries, to
escape the trap of procyclical fiscal policy? We believe that the main story concerns institutions.2 In our
new research (Frankel et al. 2011), we find that the cyclicality of a country’s fiscal policy is inversely
correlated with the country’s institutional quality which includes measures of law and order, bureaucracy
quality, corruption, and other risks to investment.
Although one thinks of institutions as slow-moving, they can change over time. Chile’s institutional
quality has risen strongly since the early 1980s, during which time its fiscal policy has turned from
procyclical to countercyclical. A country with good institutional quality in the general sense of rule of law
can help lock in countercyclical fiscal policy through specific budget institutions. Frankel (2011a)
explains how Chile did it, with the structural budget reforms of 2000 and 2006. Chile’s approach could
be emulated by others.
Fiscal rules, such as the Eurozone’s Stability and Growth Pact, may accomplish little in themselves.
Rules can actually worsen the tendency of governments to make overly optimistic forecasts for
economic growth and budget balance (Frankel 2011b). Chile’s key innovation was to give responsibility
for forecasting to independent expert commissions, insulated from politicians’ wishful thinking.
Even advanced countries have something to learn about countercyclical fiscal policy from Chile and
others to the South. Saving during expansions such as 2001 to 2006 is critical for weathering the storm
in recessions such as 2008-09. Otherwise there may be no way out but to adjust at the worst possible
time.
5
References
Alesina, Alberto, Filipe Campante, and Guido Tabellini (2008), “Why is Fiscal Policy Often Procyclical?”, Journal of the European Economic
Association, 6(5):1006-1036.
Alesina, Alberto and Roberto Perotti (1996), “Fiscal Discipline and the Budget Process ,” American Economic Review, 86(2):401-407.
Calderón, César, and Klaus Schmidt-Hebbel (2008), “Business Cycles and Fiscal Policies: The Role of Institutions and Financial Markets”,
Working Paper 481, Central Bank of Chile.
Didier, Tatiana, Constantino Hevia, and Sergio Schmukler (2011), “How Resilient Were Emerging Economies to the Global Crisis?”, World
Bank Policy Research WP 5637, April.
Frankel, Jeffrey (2011a), “A Solution to Fiscal Procyclicality: The Structural Budget Institutions Pioneered by Chile”, NBER WP 16945.
Frankel, Jeffrey (2011b), “A Lesson From the South for Fiscal Policy in the US and Other Advanced Countries,” forthcoming, Comparative
Economic Studies. HKS RWP11-014.
Frankel, Jeffrey, Carlos Végh, and Guillermo Vuletin (2011), “On Graduation from Procyclicality”, University of Maryland (in progress).
Gavin, Michael and Roberto Perotti (1997), “Fiscal Policy in Latin America”, NBER Macroeconomics Annual, 12:11-61.
Ilzetski, Ethan, and Carlos Vegh (2008), “Procyclical Fiscal Policy in Developing Countries: Truth or Fiction?”, NBER Working Paper 14191.
Kaminsky, Graciela, Carmen Reinhart, and Carlos Végh (2005), "When It Rains, It Pours: Procyclical Capital Flows and Macroeconomic
Policies", NBER Macroeconomics Annual 2004, 19:11-82.
Mendoza, Enrique and P Marcelo Oviedo (2006), “Fiscal Policy and Macroeconomic Uncertainty in Developing Countries: The Tale of the
Tormented Insurer”, NBER Working Paper 12586, October.
Persson, Torsten, and Guido Tabellini (2004), “Constitutional Rules and Fiscal Policy Outcomes”,American Economic Review, 94(1):25-45.
Poterba, James, and Jürgen von Hagen (1999) (eds.), Fiscal Institutions and Fiscal Performance, University of Chicago Press.
Talvi, Ernesto, and Carlos Végh (2005), “Tax Base Variability and Procyclicality of Fiscal Policy,”Journal of Development Economics, 78(1).
Tornell, Aaron, and Philip Lane (1999), “The Voracity Effect”, American Economic Review, 891:22-46.
Végh, Carlos, and Guillermo Vuletin (2011), “On the Cyclicality of Tax Rate Policy,” University of Maryland and Colby College (in progress).
von Hagen, Jürgen, and Ian Harden (1995), “Budget Processes and Commitment to Fiscal Discipline”,European Economic Review,39(3-4).
1 Végh and Vuletin (2011) find evidence that tax-rate policy has been mostly procyclical in developing countries,
and acyclical in industrialised countries.
2
In the case of fiscal policy, the importance of institutions has been emphasised by von Hagen and Harden (1995),
Alesina and Perotti (1996),Poterba and Von Hagen (1999),Persson and Tabellini (2004), and Calderón and
Schmidt-Hebbel (2008), among many others.
Jeffrey Frankel, Harvard U.
Carlos Vegh, SAIS, Johns Hopkins U.
Guillermo Vuletin, Brookings Inst.
6
Over-Optimism in Official Budget Agencies' Forecasts
NBER Digest, November 2011
Bringing budget deficits under control has been, and is, a difficult task for many
advanced countries. In Over-Optimism in Forecasts by Official Budget Agencies and Its
Implications (NBER Working Paper No. 17239), Jeffrey Frankel argues that overly optimistic
official forecasts of future budget balances have facilitated complacency and so have
contributed to tax cuts and increases in government spending, and therefore to realized budget
deficits, during the last decade.
Analyzing data for 33 countries, Frankel finds that the average upward bias in the official
forecast of the budget balance, relative to the realized balance, is 0.2 percent of GDP at the
one-year horizon, 0.8 percent at the two-year horizon, and 1.5 percent at the three-year
horizon. The longer the horizon, and the more genuine uncertainty there is, the more scope
there is for wishful thinking.
The forecast bias results are similar across nations. The bias is not larger for the
commodity producers in Frankel’s sample, or for the developing countries, than for others. Both
the U.S. and U.K. forecasts have shown positive biases , reaching around 3 percent of GDP at
the three-year horizon, which is approximately equal to their actual deficit on average. In other
words, on average the U.K. and U.S. forecasters repeatedly predicted a disappearance of their
deficits that never occurred.
Frankel argues that one likely reason for the optimism bias in official budget forecasts is
an optimism bias in forecasts of economic growth. He finds that a country's growth rate is an
important determinant of the budget balance at all three time horizons, so over-optimism in
predicting growth appears linked to over-optimism in predicting budget balances. On average,
the upward bias in growth forecasts is 0.4 percent when looking one year ahead, 1.1 percent at
the two-year horizon, and 1.8 percent at three years. The bias in growth forecasting appears in
the United States and most other industrialized countries, but not among the commodity
producing countries in the sample.
Frankel also finds that over-optimism is more prominent, for both budget balances and
for economic growth, during economic booms. This is especially true as the horizon of the
forecast lengthens. This over-optimism in official forecasts can help to explain excessive budget
deficits, and especially the failure to run surpluses during periods of high output: if a boom is
expected to last indefinitely, then saving for a rainy day is unnecessary. Forecasters overestimate the permanence of booms, and also underestimate the persistence of busts. Frankel
thus finds evidence of over-optimism in downturns as well as booms.
7
Despite calls for the establishment of rules, such as a formal ceiling for the deficit,
Frankel finds that countries subject to a budget rule -- particularly the euroland’s Stability and
Growth Path -- make official forecasts of growth and budget deficits that are even more biased
and more correlated with booms than do other countries. Evidently when such governments
exceed the deficit limits set by the rules, they respond by adjusting their forecasts rather than by
adjusting their policies.
Finally, Frankel notes that while advanced economies in his sample have struggled with
balancing their budget since 2000, some countries in the South, notably Chile, have carried out
countercyclical policy during that time period, taking advantage of the 2002-07 boom years to
attain budget surpluses. As a result of budget institutions created in 2000, Chile’s official
forecasts of growth and of budget balance have not been overly optimistic, even in booms.
(Claire Brunel)
Quote: " Overly optimistic official forecasts of future budgets have facilitated complacency and so have contributed
to tax cuts and increases in government spending."
8
Bias in Official Fiscal Forecasts: Can Private Forecasts Help?
Jeffrey Frankel and Jesse Schreger
Harvard University
June 18, 2013 + Tbls&Figs + May 16, 2014
Merolapub
The authors wish to acknowledge capable research assistance from Marco Martinez del Angel; helpful
conversations with Philip Hubbard, publisher of Consensus Economics; and support from the Smith Richardson
Foundation and the Weatherhead Center for International Affairs. Views and findings are ours alone.
Abstract
Government forecasts of GDP growth and budget balances are generally more over-optimistic than
private sector forecasts, in a sample of 31 countries. When official forecasts are especially
optimistic relative to private forecasts ex ante, they are more likely also to be over-optimistic
relative to realizations ex post. Euro area governments during the period 1999-2007 assiduously
and inaccurately avoided forecasting deficit levels that would exceed the 3% Stability and Growth
Pact threshold; private sector forecasters were not subject to this crude bias. Three important
implications. First, the severe recessions that began in 2008 do not explain the findings of
optimism bias. Second, the over-optimism behaves as if specifically designed so as to “game”
checks on budget deficits, rather than by politicians’ desire to “talk up” the economy in general.
Third, the budget-making process could probably be improved by using private-sector forecasts.
JEL classification numbers: E62, F3
Keywords: budget, deficit, euro, fiscal, government, optimism, optimistic, forecast, official, SGP.
9
Bias in Official Fiscal Forecasts: Can Private Forecasts Help?
Introduction
Excessive public debts and deficits are among the most widely discussed of macroeconomic
problems. Why do countries find it so hard to get their deficits under control? There are many
theories, most having to do with short horizons on the part of politicians1 and dispersion of
political power.2 But we believe that systematic patterns in the errors that official budget agencies
make in their forecasts also play an important role.
To take an egregious example, the official Greek forecast in 2000 said that the fiscal deficit
would fall below 2% of GDP one year in the future, fall below 1% of GDP two years into the future,
and convert to a surplus three years into the future. (Figure 1). The subsequent budget deficit
actually fell into the range of 4-5%, well above the 3% of GDP ceiling required by the Maastricht
criterion and the Stability and Growth Pact. Did the Greek government adjust its overall fiscal
policies in response to missing the targets? No, it adjusted its forecasts, predicting steady progress
toward budget balance in the future. This pattern also describes the forecasts reported by the
government in 2001, and every year for the next ten years: always overly optimistic, and the more
so the longer the horizon.
The same pattern extends qualitatively to most other industrialized countries. Having a
fiscal rule like the Stability and Growth Pact does not seem to help. In fact it worsens the bias
toward overoptimistic forecasting. Most euro countries, even when they ran deficits that exceeded
For example, Alesina and Tabellini (1990a, b), Grilli, Masciandaro, and Tabellini (1991) and Roubini and Sachs
(1989a,b).
1
Lane (2003). Other explanations for budget deficits abound as well. Surveys include Alesina, Perotti and Tavares
(1998) and Persson and Tabellini (2002).
2
10
3% of GDP, continued to forecast future budget deficits that would fall under that threshold, every
year until the crisis of 2010 hit, and in some cases even then.
Previous research (Frankel, 2011; Frankel and Schreger, 2013) studies forecasts of real
growth rates and budget balances made by official government agencies among 24 countries [or up
to 34 in some cases, mostly over the period 1999-2011]. In general, the forecasts are found: (i) to
have a positive average bias, (ii) to be more biased in booms, (iii) to be even more biased at the 3year horizon than at shorter horizons. This over-optimism in official forecasts can help explain
excessive budget deficits: if rapid growth is expected, retrenchment is treated as unnecessary.
Many believe that better fiscal policy can be obtained by means of rules such as ceilings for the
deficit or, better yet, the structural deficit. But we also find: (iv) countries subject to a budget rule,
in the form of euroland’s Stability and Growth Path, make official forecasts of growth and budget
deficits that are even more biased. This effect may help explain frequent violations of the SGP.
The question becomes how to overcome governments’ tendency to satisfy fiscal targets by
wishful thinking rather than by action. One possibility is a rule whereby the government must use
in its fiscal planning process forecasts coming from the private sector or some other independent
body that is insulated from political pressures or the temptations of wishful thinking.
(Chile instituted such a system in 2000. This presumably explains why its budget forecasts have
not been overly optimistic, even in booms.3)
When governments are overly optimistic, subsequent realized budget deficits turn out to be
larger than projected and realized surpluses are smaller. If no one used the forecasts for anything,
then these mistakes would not matter even though they are systematic. But the forecasts of the
Canada too has avoided optimism bias (and in fact has been too pessimistic on average), perhaps because it too uses
independent forecasts. With the years 2010 and 2011 added to the sample, several countries as well are no longer
over-optimistic on average at the one-year horizon, though they still are at the two-year horizon.
3
11
budget or economic agency within each government are used as the basis for fiscal planning. If the
forecast is for a strong budget, tax parameters and spending policies are likely to be eased
accordingly. Thus the excessive optimism in forecasts can help explain excessive deficits in
practice. (Returning to the case of Chile, its avoidance of overly optimistic forecasts may explain
why it was able to take advantage of the 2000-2008 boom to run a surplus, when so many other
countries did not.)
The present paper seeks further progress on these issues by expanding the data set in
crucial ways. Most importantly, it brings data on private sector forecasts together with the official
government forecasts. The resulting extension of the analysis helps answer two important
questions. First, might the earlier findings of over-optimism be explained by one major historical
event, the severe global financial crisis and recession that began in 2008 and which everyone
underestimated? More generally, when the time sample is short, results based on ex post
realizations can be too sensitive to particular historical outcomes. Private sector forecasts offer an
alternative standard by which to judge the performance of official forecasts and less sensitive to
historically volatile ex post outcomes. Second, if the reform proposal is that governments should
use in the budget-making process independent projections such as those by private forecasters,
what better way to test it than to see if private forecasters suffer from optimism bias as badly as
the government forecasters?
We have three main results, for a sample of 31 countries during sample periods that usually
go up to 2012. First, official forecasters are more over-optimistic than private sector forecasters
on average at the two-year horizon for budget balances and at the one- and two-year horizon for
real GDP forecasts. Second, the difference between the official forecast and the private forecast is
positively correlated with the difference between the official forecast and the ex post realization,
12
that is, the prediction error. Third, while euro area governments were very reluctant to forecast
violations of the 3% deficit/GDP cap in the Stability and Growth Pact during the period 1999-2007,
private sector forecasters were not. Together, these results suggest that incorporating private
sector forecasts into the budget process could help countries stick to fiscal rules by identifying
over-optimism ex ante rather than just ex post.
Related Literature
Our work builds on a moderately large literature that tries to understand bias in
government forecasts. Jonung and Larch (2006) demonstrate that European Union countries have
overly optimistic forecasts and propose independent budget forecasting as a remedy. Debrun et. al.
(2009) survey the literature on the performance of independent fiscal agencies. Merola and Perez
(2012) compare the forecasts of European governments in their Stability and Convergence
Programs (SCPs) to the forecasts made by the European Commission (EC); they find that the
forecasts made by the EC are no better than those made by national governments and are affected
similarly by political economy factors, such as over-optimism in budget years. Beetsma et. al.
(2012) also analyze the sources of national budget forecast errors from national SCPs and find that
political factors, specifically upcoming elections, play a role in explaining government overoptimism. Pina and Venes (2011) also point to the importance of upcoming elections in overoptimism. Beetsma et. al. break the budget forecasts into an implementation error, the difference
between the initial projection in the budget and the preliminary figures calculated at the end of
that fiscal year, and the revision error, or the difference between the preliminary measure and the
final statistic. The authors find that implementation errors are driven by overly optimistic
expenditure projections, whereas revision errors are driven by overly optimistic revenue forecasts.
13
Frankel (2011) shows across a cross-section of countries that government forecasts are overly
optimistic in booms. Frankel and Schreger (2013) find that euro area countries are much less
likely to forecast a breach of the 3% limit but are no less likely actually to breach the limit. Pina and
Venes (2011) find a similar result.
Boylan (2008) examines the forecasting behavior of individual US state governments and
finds an election year bias. Bischoff and Gohout (2010) undertake a similar exercise for West
German States. Beetsma et. al. (2013) study budget forecasting in the Netherlands from 19582009 and find that the plans are on unbiased on average.
Muhleisen et. al. (2005) compare the forecasting of Canada to that of other developed
economies and find that countries with fiscal rules and strong budgetary institutions are more
successful in their budget forecasting. Poplawski-Ribeiro and Rulke (2011) use the same private
sector forecast data that we do, and examine whether adoption of the SGP leads private, national
and EC forecasts to converge and find mixed results.
Data on Private Forecasts
In this paper, we use three types of data: government forecasts from national budgets,
private sector forecasts, and realizations from international organizations and national sources. 4
The most important respect in which we seek to improve on earlier research is the use of the
private sector forecasts.
Our private sector forecasts come from Consensus Economics. Every month, Consensus
Economics surveys a number of private sector forecasters for their one- and two-year ahead
forecasts about a number of macroeconomic variables. Here, we examine forecasts of real GDP and
Realizations for budget balances come from Eurostat for European countries and the IMF or World Bank for the
others. Additional details on data sources can be found in the appendix.
4
14
the budget balance. Consensus polls private sector forecasters more commonly on Real GDP
growth than on the budget balance, limiting the size of our budget balance sample and leading to
different country and years coverage across the variables.
In addition, for many countries private forecasters forecast the size of the budget balance in
terms of local currency, rather than as a percentage of GDP. In these cases, we create a private
sector forecast for the deficit/GDP ratio by first estimating the implied private sector forecast for
the level of nominal GDP. We take the previous year’s nominal GDP and multiply it by (1+Real GDP
Growth Rate Forecast)*(1+Inflation rate forecast).
Unfortunately, GDP deflator forecasts are unavailable and so we use CPI inflation for the
inflation rate. For the two-year ahead nominal GDP forecasts, we multiply our one year ahead
estimate by one plus the two-year ahead Real GDP Growth Rate Forecast and one plus the two-year
ahead inflation rate. We then divide the budget balance forecast (in levels) by our estimated
nominal GDP forecasts. In the case of countries where the fiscal year differs from the calendar
year but only private sector calendar year GDP growth and inflation forecasts are available, we
collect official forecast data on the level of the budget deficit and divide both the official and private
sector forecasts by realized GDP. Although this means that these countries are treated slightly
differently than the others, within each country the treatment is identical across private and official
forecasts, minimizing the potential bias. The annual variation in the level of GDP is much smaller
than the variation in the level of the budget deficit. (The countries with fiscal years other than the
calendar year are Australia, Canada, New Zealand, the United States and the UK.)
In order to make sure our private and official forecasts are comparable, we match the two
forecasts by the date the forecast was made. Every month, Consensus releases two sets of forecasts
for the two upcoming years. The forecasts are not for a fixed horizon, but rather forecast a given
15
variable for a given year. For instance, in the January 2003 release of Consensus Economics,
respondents will forecast French real GDP growth for the 2003 and 2004 calendar years. In
February 2003, the same variables will be forecast, but now instead of being 12 and 24 month
forecasts, they are 11 and 23 month forecasts. In contrast, our government forecasts are made only
once a year in the official budget document. We match private and official forecasts so that there is
at most a 3 month gap between their forecast dates. 77.7% of our matches are exact to the month,
15.0% differ by one month, 6.4% differ by two months, and 0.9% differ by 3 months for budget
deficit forecasts. For real GDP growth rate forecasts, 71.4% of matches are exact, 19.2% differ by
one month, 8.8% differ by two months, and 0.6% differ by 3 months. Only one observation is used
per country-year, and no matches with a horizon difference greater than 3 months are used. There
still is the possibility that the forecasts were not made at the same time, as the official forecast date
is the date that the budget was released whereas the private forecast date is the date that
Consensus polled the firms. There is presumably a lag between when government forecasts are
actually made and when the budget is released.
Table 1 summarizes the dataset. Each column shows the first and last year we have a
matched private forecast, government forecast and realization. The data set is an unbalanced
panel with a different set of countries having Real GDP and budget balance forecasts.
What are the differences between private and official forecasts?
Having constructed a dataset combining the private and official forecasts, we then turn to
understanding how the two sets of forecasts differ. Here, we only use the mean private forecast in
the Consensus dataset and compare it to the single official forecast. In Table 2, we present the
means, standard deviation and number of observations for dates where we were able to match the
16
Consensus and Official data, as well as having data on the realization. For several countries, the
2012 realizations were not available at the time of writing. [More information can be found in the
data appendix available on request.]
In Table 2A, we compare the one- and two-year ahead budget balance forecasts and actual
outcomes. At the one year horizon, mean actual budget deficits are roughly the same size as
Consensus and Official forecasts. At the two year horizon, however, budget balances show larger
deficits than forecast in either Consensus or Official forecasts, but Consensus forecasts are closer to
the realizations. The standard deviation of actual budget balances is larger than the Consensus or
Official forecasts, as one would expect. The same pattern is found for Real GDP growth rate
forecasts at both one- and two-year horizons in Table 2B.
In Table 3, we run two-sided t-tests for equality of means across the various forecasts and
realizations. We restrict all of our tests to a common sample by variable but the sample differs
across variables, as can be seen in the number of observations. In Table 3A, we see that the mean
one-year ahead ex post Consensus budget balance forecast error, ex post Official budget balance
forecast error, and the difference between the three forecasts are not significantly different from
zero.
At the two year horizon, however, both Official and Consensus forecasts are significantly
over-optimistic ex post. However, while Official forecasts were ex post over-optimistic by 0.66% of
GDP on average, official forecasts are on average 0.17% more optimistic than Consensus forecasts.
This criterion of over-optimism is even more highly significant statistically than judging official
forecasts against the standard of ex post outcomes. Given how bad were the ex post outcomes after
the global financial crisis, this is an important finding. Because the comparison of official forecasts
with private forecasts makes no use of ex post data, it is immune from the possibility that our
17
results are driven by one record-breaking recession. Intuitively, it is harder to excuse the
authorities from under-estimating the severity of the recession to the extent that private
forecasters did not make the same mistake.
Table 3B conducts the corresponding t-tests for Real GDP growth rate forecasts. At the one
year horizon, Consensus forecasts do not have statistically significant ex post errors (point estimate
of 0.18%), but Official forecasts have a mean over-optimism of 0.32%. In addition, official
forecasts are significantly more optimistic than private sector forecasts, with a mean difference of
0.14%. The differences are larger at the two-year horizon, with ex post official forecasts errors of
1.15%, private forecast errors of 0.99%, and a mean difference between the two of 0.16%, all of
which are significant at the 1% level.
Figures 2 and 3 plot the data for the example of Italy, comparing private and official
forecasts with the actual realizations. While private and official forecasts co-move strongly,
government forecasters consistently predict smaller deficits and faster growth. In sample, both
private and official forecasts proved themselves to be generally over-optimistic, but private
forecasts less so.
Can private forecasts be used to predict when official forecasts will prove to be over
optimistic?
In this section, we want to see if we can use private forecasts to improve on the
performance of predictions made by the official agencies. In other words, is the ex ante
discrepancy between private and official forecasts positively correlated with the discrepancy
between official forecasts and the realized outcome, that is, the ex post prediction error? In Table
4, we see that the answer is yes.
18
We define the Government-Consensus Budget Balance disagreement (Gov-Con) and the
government i-year ahead forecast error (BBEgt+i)
Gov-Con BBt+i =BB_Govt+i – BB_Cont+i
BBEgt+i =BB_Govt+i –BBt+i
where BB_Govt+i denotes the i-year ahead government budget balance forecast (as a % of GDP)
made in year t, BB_Govt+i denotes the i-year ahead Consensus budget balance forecast made in year
t, and BBt+i denotes the actual budget balance in year t+i. In Table 4, we regress one- and two-year ex
post official forecast errors on this variable and various combinations of country and year fixed
effects and find that contemporary disagreement between private and official forecasters is a
useful measure for predicting when government forecasts will prove over-optimistic. In our
baseline specification (columns 1 and 2) without any fixed effects, for every 1% that the
government forecasts are more optimistic than private forecasts at the one year horizon, we find
an ex post government error of 0.93%. At the two year horizon, the point estimate is 0.61%. One
way to read this is that when government forecasts are unusually optimistic relative to private
forecasters, government forecasts on average are also unusually optimistic relative to the ex post
outcome. With only a constant and the difference between private and official forecasts in the
equation, we have an R-squared of 18% at the one year horizon and an R-squared of 4% at the two
year horizon. When we include year fixed effects, the results are further strengthened (columns 5
and 6), meaning that in a given year, those countries that made forecasts relatively more optimistic
than private sector forecasts have larger ex post budget deficits than those countries that did not.
However, the result is not as strong within country, as the addition of country fixed effects reduces
the significance level at the one-year horizon and reduces the point estimate and all significance at
the two-year horizon.
19
This same relationship is illustrated graphically in Figures 4 and 5. In the scatter plot, each
marker indicates the mean of a country. However, the regression line is not a fitted line through
these country means. Rather, the regression line is weighted by country-year, as in the pooled OLS
in columns 1 and 2 of Table 4. Only countries with 4 or more observations are included.
In Table 5, we run similar regressions as in Table 4 but examine real GDP forecast errors.
We replace Gov-Con BB with an equivalent measure defined for real GDP growth rates, and
similarly for growth rate forecast errors GDPEg
Gov-Con GDPt+i =GDP_Govt+i – GDP_Cont+i .
GDPEgt+i =GDP_Govt+i –GDPt+i
Despite economically significant point estimates at the one-year horizon and occasional statistical
significance, the results are not as clear as for budget balances in Table 4.
Figures 6 and 7 plot the equivalent figures for Real GDP growth forecast disagreement and
ex post real GDP forecasts errors. The results are fairly inconclusive.
Can private forecasts be used to make fiscal rules move effective?
Finally, we turn to whether private forecast rules can potentially make fiscal rules more
effective. Frankel and Schreger (2013) demonstrate that the extra bias among euro area countries
took the specific form of rarely forecasting that their budget would breach the 3% threshold
enshrined in the Stability and Growth Pact during the period 1999-2007. (After 2007, the financial
crisis pushed them too far past the threshold to keep up the pretense.) This was in spite of the fact
that their actual budget balances displayed no such discontinuity at 3%.5
By only including forecasts made through 2007, the two-year ahead figure covers forecasts of budget deficits through
2009 and the one-year ahead figure covers forecasts through 2008. After the start of the financial crisis, budget
balances became much larger and European countries began deficits in excess of the 3% threshold.
5
20
In Figure 8, we replicate these results from Frankel and Schreger (2013) while checking
whether the private sector forecasters made a similar sort of error (in those country-years where
we also have Consensus data). In the middle panel, we reproduce the figure for official forecasts as
in our earlier work. In the upper left hand graph, we create the analogous figure for Consensus data
and find no discontinuity at 3%. In other words, if the national governments had been using
private sector forecasts to determine when their countries were likely to breach the 3%
requirements of the SGP, they could not have used wishful thinking to respond to warnings from
the EC. In Figure 9, we also produce the same figures at the one-year horizon. Although the euro
area forecasts do not have the dramatic discontinuity at 3%, one can still see a large cluster of
forecasts immediately to the right of the 3% threshold which is absent from the Consensus
Forecasts.
These results help sharpen the evidence on bias in official forecasts. They confirm that the
existence of a fiscal rule such as the Stability and Growth Pact does nothing to solve the problem of
over-optimism, and may even make it worse. They might even shed some light on the motives of
governments. The findings of over-optimism in the literature could have been explained in a
number of ways. One obvious hypothesis is that national leaders seek to convince their voters that
their country’s generic economic performance is good, either for the political purpose of winning
votes or for the economic purpose of boosting consumer and business confidence. There may be
psychological explanations as well. But the finding of a threshold of 3% among the euro country
forecasts suggests a narrower explanation, that they were in effect gaming the rules of the
European Commission’s excessive deficit procedure under the SGP.
21
Conclusion
In this paper we document three main findings. First, official forecasts are not only overoptimistic, but are more over-optimistic than private forecasts. Second, the ex ante discrepancy
between private and official forecasts is positively correlated with the discrepancy between official
forecasts and the realized outcome, that is, with the ex post prediction error. Private sector
forecasts can improve on official forecasts (though less so when we add fixed effects for countries).
Third, private sector forecasts predicted euro area countries would breach the 3% limit in the
Stability and Growth Pact, usually accurately, even though government agencies did not forecast
breaching the limit.
The evidence is mounting that over-optimism in official forecasts may play a key role in
delivering excessive budget deficits. In the first place, we see that the severe recessions that began
in 2008 are not driving the finding by themselves. In the second place, we see that euro
governments refrained from forecasting deficits over the 3 per cent SGP threshold, even when
private forecasts and actual deficits were both above this level. Apparently the optimism bias is
not just an example of politicians’ general tendency to look on the bright side.
There is an important possible implication for reform proposals: Tightening fiscal rules will
not help limit budget deficits, if forecasts remain subject to gaming the rules or (more charitably)
to wishful thinking. Giving independence to the agencies that make the forecasts used in the
budget-making process or even using leading private forecasts directly may be more likely to help.
22
References
Alesina, Alberto, Roberto Perotti and Jose Tavares, 1998, "The Political Economy of Fiscal Adjustments,"
Brookings Papers on Economic Activity, Vol. 28, no. 1: 197-248.
----- and Guido Tabellini, 1990a, “Voting on the Budget Deficit,” American Economic Review, 80, No. 1, March.
----- and Guido Tabellini, 1990b, “A Positive Theory of Fiscal Deficits and Government Debt,” The Review of
Economic Studies.
Beetsma, Roel, and Massimo Giuliodori, “Fiscal Adjustment to Cyclical Developments in the OECD: An
Empirical Analysis Based on Real-time Data,” Oxford Economic Papers, 2010, 52 (3), 419–441.
Beetsma, Roel, Benjamin Bluhm Massimo Giuliodori and Peter Wierts, “From Budgetary Forecasts to Ex Post
Fiscal Data: Exploring the Evolution of Fiscal Forecast Errors in the European Union,” Contemporary Economic
Policy, Forthcoming.
Beetsma, Roel, Massimo Giuliodori Beetsma and Peter Wierts, “Planning to Cheat: EU Fiscal Policy in Real
Time,” Economic Policy, 2009, 24, 753–804.
Beetsma, Roel, Massimo Giuliodori Mark Walschot Peter Wierts , “Fifty Years of Fiscal Planning and
Implementation in the Netherlands,” European Journal of Political Economy, Forthcoming.
Bischoff, Ivo, and Wolfgang Gohout, “The Political Economy of Tax Projections,” International Tax and Public
Finance, 2010, 17 (2), 133–150.
Boylan, Richard T., “Political Distortions in State Forecasts,” Public Choice, 2008, 136, 411–427.
Bruck, Tilman and Andreas Stephan, “Do Eurozone Countries Cheat with their Budget Deficit Forecasts,” Kyklos,
2006, 59 (1), 3–15.
Frankel, Jeffrey, “Over-optimism in Forecasts by Official Budget Agencies and its Implications,” Oxford
Review of Economic Policy, 2011, 27 (4), 536–562.
Frankel, Jeffrey, and Jesse Schreger, “Over-optimistic Official Forecasts and Fiscal Rules in the Eurozone,”
Review of World Economics, 2013, 149 (2), 247–272.
Grilli, Vittorio, Donato Masciandaro, Guido Tabellini, 1991, “Political and Monetary Institutions and Public
Financial Policies in the Industrial Countries,” Economic Policy, Vol. 6, No. 13, Oct., pp. 341-392.
Jonung, Lars, and Martin Larch, “Improving Fiscal Policy in the EU: The Case for Independent Forecasts,”
Economic Policy, 2006, 21 (47), 491–534.
Lane, Philip, 2003, “The Cyclical Behaviour of Fiscal policy: Evidence from the OECD,” Journal of Public
Economics, Volume 87, Issue 12, December, Pages 2661-2675.
Muhleisen , Martin, Stephan Danninger, David Hauner, Kornelia Krajnyak and Bennett Sutton, “How Do
23
Canadian Budget Forecasts Compare With Those of Other Industrial Countries?,” IMF Working Paper, April
2005, (05/66).
Merola, Rossana, and Javier J. Perez, 2012, “Fiscal Forecast Errors: Governments vs Independent Agencies?”.
European Journal of Political Economy - 23, pp. 285-299, September 2013.
Milesi-Ferretti, Gian Maria, “Good, Bad or Ugly? On the Effects of Fiscal Rules with Creative Accounting,”
Journal of Public Economics, 2003, 88 (1-2), 377–394.
Persson, Torsten, and Guido Tabellini, 2002, “Political Economics and Public Finance,” Chapter 24, Handbook of
Public Economics, Volume 3, Pages 1549-1659.
Pina, Alvaro M., and Nuno M. Venes, “The Political Economy of EDP Fiscal Forecasts: An Empirical
Assessment,” European Journal of Political Economy, 2011, 27, 534–546.
Poplawski-Ribeiro, Marco, and Jan-Christoph Rulke, “Fiscal Expectations Under the Stability and Growth Pact:
Evidence from Survey Data,” IMF Working Paper, March 2011, (11/48).
Roubini, Nouriel, and Jeffrey Sachs, 1989a, “Government Spending and Budget Deficits in the Industrial
Economies,” Economic Policy, No. 8, April.
--------------------------, 1989b, “Political and Economic Determinants of Budget Deficits in the Industrial
Democracies,” European Economic Review, Volume 33, Issue 5, pp. 903-933.
Teresa, Javier J. Perez, Mika Tujula Leal and Jean-Pierre Vidal, “Fiscal Forecasting: Lessons From the Literature
and Challenges,” Fiscal Studies, 2008, 29 (3), 347–386.
Wyplosz, Charles, 2005, “Fiscal policy: Institutions versus Rules,” National Institute Economic Review, 191, 64–
78.
Xavier, David, Hauner Debrun and Manmohan S. Kumar, “Independent Fiscal Agencies,” Journal of Economic
Surveys, 2009, 23 (1), 44–81.
24
Tables
Table 1A Data Coverage
country
Australia
Austria
Belgium
Canada
Chile
Cyprus
Czech Republic
Denmark
Estonia
Finland
France
Germany
Greece
Hungary
Ireland
Italy
Latvia
Lithuania
Mexico
Netherlands
New Zealand
Norway
Poland
Portugal
Slovakia
Slovenia
South Africa
Spain
Sweden
UK
USA
BB 1Y
1993-2012
1995-2012
1994-2009
BB 2Y
1994-2011
2002-2012
2005-2012
2006-2012
2008-2012
2009-2012
1999-2012
1999-2011
2000-2012
2000-2012
2005-2012
2006-2012
1999-2012
2008-2012
2008-2012
1995-2009
2000-2012
2009-2012
2009-2012
2005-2009
1997-2011
2011-2011
2005-2012
1998-2011
2011-2012
2008-2012
2012-2012
2009-2012
2011-2012
2000-2012
1994-2012
2012-2012
2006-2012
1995-2012
GDP 1Y
GDP 2Y
1999-2012
1999-2012
1990-2012
1994-2008
2005-2011
2005-2012
1999-2011
2006-2011
1999-2011
1999-2011
1999-2011
2000-2011
2005-2012
1999-2011
1999-2012
2005-2012
2005-2012
2000-2012
1999-2011
2000-2012
2000-2012
1991-2010
2002-2012
2005-2011
1999-2010
2005-2012
2005-2012
1998-2012
1999-2011
1999-2012
1998-2012
1992-2012
2006-2011
2006-2012
2000-2011
2007-2011
2001-2011
2000-2011
2000-2011
2001-2012
2006-2012
2000-2011
2000-2012
2006-2012
2006-2012
2005-2012
2000-2011
2006-2011
2000-2011
2006-2012
2006-2012
1999-2012
2000-2011
2000-2012
1999-2012
1992-2012
Notes: Each cell indicates the first and last year we have a matched private and official forecast as well as the
relevant realization. GDP1Y indicates one year ahead real GDP forecasts, GDP2Y indicates two year ahead real
GDP forecasts, BB 1Y indicates one year ahead budget balance forecasts, and BB 2Y indicates two year ahead
budget balance forecasts.
25
Table 2A: Summary Statistics for Budget Balance Forecasts
One Year Ahead
Consensus
Official
Actual
Mean
-1.92
-1.83
-1.86
SD
3.01
3.03
3.47
Obs.
205
205
205
Two Years Ahead
Consensus
Official
Actual
Mean
-1.83
-1.66
-2.32
SD
2.43
2.25
3.13
Obs.
138
138
138
Table 2B: Summary Statistics for Real GDP Growth Forecasts
One Year Ahead
Consensus
Official
Actual
Mean
2.48
2.62
2.30
SD
1.97
1.96
3.51
Obs.
350
350
350
Two Years Ahead
Consensus
Official
Actual
Mean
2.87
3.03
1.88
SD
1.31
1.34
3.56
Obs.
289
289
289
Notes: Consensus indicates the mean private sector forecast from Consensus Economics. Official indicates the
government forecast in their official budget document. Actual indicates the realization of the Budget Balance and
Real GDP growth in Tables 2A and 2B, respectively. SD denotes standard deviation and Obs. denotes
observations.
26
Table 3A: t-Tests for Equality of Means for Budget Balance, Common Sample
One Year Ahead
Mean
SD
P-Value
Obs.
Official Minus Consensus
0.09
0.06
0.17
205
Official Forecast Error
0.03
0.14
0.82
205
Consensus Forecast Error
-0.06
0.13
0.66
205
Two Years Ahead
Mean
SD
P-Value
Obs.
Official Minus Consensus
0.17**
0.07
0.02
138
Official Forecast Error
0.66***
0.21
0.00
138
Consensus Forecast Error
0.49**
0.21
0.02
138
Table 3B: t-Tests for Equality of Means for Real GDP Growth, Common Sample
One Year Ahead
Mean
SD
P-Value
Obs.
Official Minus Consensus
0.14***
0.03
0.00
350
Official Forecast Error
0.32**
0.13
0.01
350
Consensus Forecast Error
0.18
0.12
0.16
350
Mean
SD
P-Value
Obs
Official Minus Consensus
0.16***
0.03
0.00
289
Official Forecast Error
1.15***
0.21
0.00
289
Consensus Forecast Error
0.99***
0.21
0.00
289
Two Years Ahead
27
Table 4: Government-Private Disagreement over Budget Balance and Official Budget Balance Forecast Errors
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
VARIABLES
BBEgt+1
BBEgt+2
BBEgt+1
BBEgt+2
BBEgt+1
BBEgt+2
BBEgt+1
BBEgt+2
Gov-Con BBt+i
0.934**
0.608**
0.873*
0.296
0.888***
0.678***
0.821*
0.343
(0.349)
(0.227)
(0.461)
(0.306)
(0.296)
(0.223)
(0.402)
(0.248)
-0.050
0.560***
-0.019
-0.113
(0.156)
(0.191)
(0.191)
(0.162)
(0.075)
(0.029)
(0.102)
(0.032)
205
138
205
138
205
138
205
138
0.177
0.040
0.280
0.181
0.445
0.485
0.535
0.601
Country FE
No
No
Yes
Yes
No
No
Yes
Yes
Year FE
No
No
No
No
Yes
Yes
Yes
Yes
Countries
20
17
20
17
20
17
20
17
Constant
Observations
R-squared
-1.568*** -0.782*** -1.551*** -0.825***
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: The LHS variable in every column is the Official Budget forecast Error. 1Y indicates the one year ahead
error and 2Y indicates the two year ahead error. Gov-Con BB is the official forecast budget balance minus the
Consensus forecast of the budget balance.
28
Table 5: Government-Private Disagreement over GDP Growth Rates and Real GDP Growth Rate Forecast Errors
(1)
VARIABLES
Gov-Con GDPt
(2)
GDPEgt+1 GDPEgt+2
(3)
(4)
(5)
(6)
(7)
(8)
GDPEgt+1
GDPEgt+2
GDPEgt+1
GDPEgt+2
GDPEgt+1
GDPEgt+2
1.146*
0.143
1.194*
-0.076
0.695*
0.600
0.716*
0.525
(0.572)
(0.507)
(0.657)
(0.633)
(0.350)
(0.405)
(0.396)
(0.582)
0.156
1.131***
0.082
0.714***
1.062***
4.508***
1.060***
4.581***
(0.120)
(0.170)
(0.090)
(0.074)
(0.023)
(0.394)
(0.026)
(0.567)
350
289
350
289
350
289
350
289
R-squared
0.078
0.000
0.115
0.070
0.561
0.619
0.589
0.659
Country FE
No
No
Yes
Yes
No
No
Yes
Yes
Year FE
No
No
No
No
Yes
Yes
Yes
Yes
Countries
29
27
29
27
29
27
29
27
Constant
Observations
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Notes: The LHS variable in every column is the Official Real GDP Forecast Error. 1Y indicates the one year
ahead error and 2Y indicates the two year ahead error. Gov-Con GDP is the official forecast real GDP growth rate
minus the Consensus forecast of the real GDP growth rate.
29
Figure 1: Greek Over-Optimism
Notes: Data from Greece’s Stability and Convergence Programs.
30
Figure 2: Italian Budget Balance Forecasts
One-Year Ahead
Two-Years Ahead
Notes: Year is year being forecast.
31
Figure 3: Italian Real GDP Growth Forecasts
One-Year Ahead
Two-Years Ahead
Notes: Year is year being forecast.
32
Figure 4: Budget Balance forecasts, One-Year Ahead
Figure 5: Budget Balance forecasts, Two-Years Ahead
Notes: Only countries with more than 4 observations are included.
33
Figure 6: Real GDP Growth forecasts, One-Year Ahead
Figure 7: Real GDP Growth forecasts, Two-Years Ahead
Notes: Only countries with more than 4 observations are included.
34
Figure 8: Budget Balance Forecasts and Realizations in the Eurozone, Two-Years ahead, Through 2009
Figure 9: Budget Balance Forecasts and Realizations in the Eurozone, One-Year ahead, Through 2008
35
36