The IMF Staff's Misleading New Evidence on Capital Controls
On February 19, 2010, the International Monetary Fund released a "Staff Position Note" on capital controls as a policy instrument.1 The paper concluded that, subject to four conditions, controls on capital inflows can be justified. The four conditions are: the economy is operating near potential, there is an adequate level of reserves, the exchange rate is not undervalued, and the excessive flows are likely to be transitory. The study acknowledged that "widespread adoption of controls could have a chilling longer-term impact on financial integration and globalization, with significant output and welfare losses" (pg. 5). It further acknowledged that "widespread use of controls, especially by systemically important countries, could also impede necessary steps to address global imbalances… where currencies are undervalued and where appreciation is needed to support global demand rebalancing" (pg. 10).
The initial press attention, however, focused on the seeming about-face from previous IMF opposition to controls, not on the caveats. Actually the Fund has not been categorically opposed to temporary restraints on capital inflows, and the paper can be seen as a further elaboration of the staff's response to the 2005 report of the Independent Evaluation Office calling for more clarity on the IMF's approach to capital account issues.2 Although the policy conditions spelled out by the authors are useful, the new empirical evidence they produce is unfortunately seriously misleading, yet it is featured prominently in some public accounts of the study.3
The authors rightly consider the recent global financial crisis to be a "natural experiment" to test whether capital controls substantially lessen the adverse impact of external shocks on emerging-market economies. Elsewhere I have shown that for 24 major emerging-market economies, the answer is "no." Reductions in average growth rates from 2005–07 to 2008–10 show no pattern whatsoever in relationship to the most widely used measure of financial openness. Neither is there any relationship between the severity of stock market declines and financial openness.4
The IMF authors, however, reach the opposite conclusion: that more financially open economies fared worse in the global financial crisis. For 37 emerging-market economies, the authors find that severity of the growth decline from the 2003–07 average to 2008–09 was associated with larger stocks of debt liabilities relative to GDP, and larger stocks of foreign direct investment (FDI) in the financial sector relative to GDP. Moreover, there was a "negative association between capital controls that were in place prior to the global financial crisis and the output declines suffered during the crisis" (pg. 13). Unfortunately, it turns out that their results are driven almost entirely on evidence from a handful of small, unrepresentative economies: the three Baltic states and Iceland.
The growth collapse was the most severe in Estonia, Latvia, and Lithuania, where average growth fell from about 9 percent annually in 2005–07 to about –7 percent in 2008–10.5 The extent of external indebtedness was the most extreme prior to the crisis in two of these same three economies (Estonia and Latvia), as shown in figure 1 reproduced from the study. Similarly, the slope of the regression line for financial sector FDI is heavily influenced by the extreme outlier case of Estonia (figure 2). It is evident even to the naked eye that if Estonia, Lithuania, and Latvia are removed from the analysis, there is no meaningful relationship between the growth collapse and either external debt or FDI in the financial sector. This diagnosis is confirmed by regression analysis.6
* Growth decline defined as average growth rates in 2008-09 relative to 2003-07 Source: Ostry et al., 2010.
* See figure 1 note Source: Ostry et al., 2010.
Should policymakers really believe that they should move toward capital controls because of the Baltics' experience? Surely not. These are tiny economies. They have an average population of 2.3 million and GDP (in 2008) of $35 billion. Brazil would not decide whether to manufacture commuter aircraft based on Baltic experience; neither should it do so in making decisions on how or how not to treat capital flows. The three states were running average current account deficits of 15 percent of GDP in 2005–07. The proper lesson is once again that extremely large current account deficits are a recipe for disaster—not that capital inflows should be controlled. Exchange rate depreciation, tighter fiscal policies, and lower interest rates are the proper way to address vulnerability from extremely large current account deficits, not capital controls.
As for the third result, it also is fragile. The test divides the sample of about 40 countries into the worst-outcome 10 percent—classified as "growth-crisis" cases—and all the others. The dependent variable is set at 1 for the four crisis cases and 0 for all others. This variable is related in a probit test to a particular measure of capital controls (Schindler).7 The four "crisis" economies are Iceland, Latvia, Kazakhstan, and Turkey. (The Schindler measure is not available for Estonia or Lithuania.) Applying the parameter estimated, movement from the 25th percentile of the controls variable to the 99th cuts the probability of a crisis outcome from about 12 percent to nearly 0. 8
Once again, however, two of the crisis countries are highly unrepresentative. Latvia is again in the benchmark group. This time, so is tiny Iceland, which is even smaller than the Baltic states in population (about 300,000) and no larger in GDP ($12 billion). Worse, Iceland is a high-income country (about $40,000 per capita at market exchange rates in 2007). If one high-income country is to be included in the sample, why not the United States and many others with fully open capital markets? Once again, it was distorted policies in other areas (failure of financial system regulation) that drove the collapse in Iceland, not financial openness per se.
In sum, although the recent IMF report contributes useful policy and situational screens for deciding whether temporary restraints on capital inflows are warranted, its new empirical evidence conveys the false impression that in general emerging-market economies with capital controls fared better in the crisis of 2008–09 than financially open economies. The authors' results, however, are extraordinarily dependent on outcomes in a few small, unrepresentative economies. This note demonstrates that two of the three empirical results of the paper do not hold after statistically controlling for the Baltic states, which recklessly pursued large current account deficits and were moreover in a unique historic phase of transition and pre-accession to a monetary union. The third result must also be suspect because fully half of the four "crisis" cases are again tiny, unrepresentative economies (Iceland and Latvia). If the test for severity of growth reduction is instead applied to a more representative set of 24 major emerging market economies (the smallest of which is the Philippines), there is no pattern of worse outcomes for more financially open economies.
1. Ostry, Jonathan D., Atish R. Ghosh, Karl Habermeier, Marcos Chamon, Mahvash S. Qureshi, and Dennis B. S. Reinhardt. 2010. Capital Inflows: The Role of Controls. IMF Staff Position Note SPN/10/04. Washington: International Monetary Fund. (February 19).
2. IMF (International Monetary Fund) Independent Evaluation Office. 2005. Report on the Evaluation of the IMF's Approach to Capital Account Liberalization. Washington.
3. 2010. Capital Controls: Fundamental Questions. Economist. (February 20): 75.
4. Cline, William R. Financial Globalization, Economic Growth, and the Crisis of 2007-09. Forthcoming. Washington: Peterson Institute for International Economics. The financial openness measure is that of Quinn and Toyoda (Quinn, Dennis and A. Maria Toyoda. 2008. Does Capital Account Liberalization Lead to Growth? Review of Financial Studies, 21, no. 3: 1403–1449.)
5. IMF (International Monetary Fund). 2009. World Economic Outlook Database. (October).
6. Thus, for the test on debt liabilities, the IMF study obtains a highly significant coefficient of 0.116 for the growth decline. Replicating this test with the data from figure 1 obtains approximately the same coefficient (0.109) and a highly significant t-statistic of 3.7. However, adding a Baltic dummy variable to this equation removes the significance of the debt variable. Thus, with Dg = reduction in growth (percentage points), D = debt liabilities as a percent of GDP, and B as a dummy variable for Baltic states, a regression shows:
Dg = 3.98 (2.5) +0.028 D (0.85) + 11.18B (3.8); adj. R2 = 0.46, with t-statistics in parentheses. The Baltics had 11 percentage points larger decline in growth than could be explained by their debt liabilities; and the debt liability coefficient is not only much smaller than before but also statistically insignificant. Similarly, for financial sector FDI the coefficient is statistically insignificant but the Baltic dummy variable coefficient explains a 12.1 percentage point growth reduction and is highly significant (t-statistic = 4.6).
8. By communication from the authors.