Projecting a country’s economic growth into the medium term and beyond is notoriously difficult. At the same time, getting the growth projections wrong has major adverse consequences. For fiscal policymakers, overestimating future economic growth implies underestimating the government debt-to-GDP ratio that will be reached at the end of the projection period (in the absence of corrective policy measures). As a result, either the country will end up with a higher-than-expected debt ratio, which could result in a debt crisis, or future policymakers will have to tighten fiscal policy abruptly – with disruptive consequences – at a later stage.
Yet, economic growth forecasts often exhibit optimism bias, particularly for longer horizons. While the ex-post empirical association between a country’s growth rate in a given decade and in the following one is weak (Easterly et al. 1993), forecasters often predict continued rapid growth into the medium and long term for countries that have recently experienced strong growth. Drawing on these observations, Pritchett and Summers (2013) have recently argued that longer term growth forecasts for China and India—where growth has been exceptionally high for more than a decade—fail to take into account the “reversion to the mean” effect, whereby exceptional performance tends to dissipate. In a similar vein, there is evidence that forecasters have an especially difficult time predicting turning points in the economic cycle (Juhn and Loungani, 2002; Ahir and Loungani, 2014). Our recent work (Ho and Mauro, 2014) investigates the degree of optimism bias – and the extent to which the persistence of strong growth may be overestimated – in economic growth forecasts at longer-term horizons (up to twenty years).
How good is past growth in predicting future growth?
We begin by repeating the exercise done by Easterly et al. (1993), computing the (simple and rank) correlation coefficients between individual countries’ per capita real GDP growth rate in one decade with the previous decade (Table 1). Even with two more decades of data, the conclusion remains remarkably similar. These correlations in growth rates between decades are low, ranging between 0 – 0.5 depending on the sample and period under consideration. Thus, preliminary evidence suggests that by historical experience, past growth is often a poor predictor of future growth.
Table 1. Simple and rank correlations of growth rates across decades
Sources: World Economic Outlook, IMF, and Penn World Tables. 1/ Spearman rank correlation; 2/ Using WEO definition of fuel exporters
To more precisely gauge the persistence in countries’ growth rates, we estimate the autocorrelation coefficient for per-capita income growth in panel regressions using a large sample of countries over 1950-2010. We do this for different horizons, from one year up to twenty years. For example, for the ten-year horizon, we regress a country’s average per capita growth rate in one decade against its growth rate in the previous decade, allowing for possible convergence effect whereby poorer countries tend to grow faster. The autocorrelation coefficient is an estimate of how persistent growth is from one period to the next. Results, which are robust to different samples and estimation techniques, show that this coefficient is generally low, ranging from 0.2 to 0.35 depending on the horizon (Figure 1). In other words, empirical evidence points to low persistence of growth rates over time; a country’s growth rates have a high tendency to revert to the sample average.
Figure 1. Persistence of per capita income growth rates, 1950–2010
Sources: Penn Worlt Table 8.0 and authors' calculations.
Notes: Estimation samples use overlapping observations.
HAC estimator (Newey-West) to correct for autocorrelation in errors.
All point estimates are significant at 1% level.
Comparing long-term growth forecasts with recent growth outcomes
Do economists’ growth forecasts reflect the large degree of “mean reversion” observed in historical growth data? We assess this question by comparing long-term growth forecasts (10-year horizon and beyond) against those that one could obtain by drawing on the empirical findings above. The forecasts are prepared jointly by IMF and World Bank teams for the purpose of the debt sustainability analyses conducted for the group of about 70 low-income countries.
In particular, we construct a “reversion to the mean” forecast by applying the empirically estimated persistence coefficient at ten-year horizon (red line in Figure 2). The dots represent all available 10-year-ahead forecasts, plotted against average growth rates in the previous ten years, and summarised by the fitted line (green line). With the green line visibly above the red line, IMF/World Bank country teams predict better growth performance in the future than would be implied by our “reversion to the mean” framework. To illustrate the difference, consider a country whose per-capita income grew over the past decade at an average rate equal to the sample mean (2.4%). The typical forecast predicts this country to grow at an average rate of 3.1% in the next decade, compared to 2% predicted by our model, so that the optimism bias in this example is 1.1 percentage points. The bias tends to get larger for countries that have experienced rapid growth in the past. In the paper, we also show that the bias is larger at longer forecast horizons.
Figure 2. Forecast vs. past per capita growth, ten-year horizon (percent)
Sources: WTO, DSA (vintages 2006-13), UN, and authors' calculations
Notes: Excluding top and bottom 1 percentiles to remove outliers.
Growth forecasts converted to per capita terms using UN population projections.
How have forecasters done in the past?
Thus far, we have made the case that today’s forecasts look optimistic by comparison with past actual growth performance. But what if there are good reasons to expect that the future will be better than the past? We cannot assess that claim directly. However, an examination of historical five-year forecasts from the IMF’s World Economic Outlook database (1990-2012) shows that economists have had a pretty consistent record of forecasts that turned out to be optimistic by comparison with actual outcomes. In particular, forecast errors (defined as forecast minus actual) tend to be positive at all horizons, with the positive bias increasing as the horizon becomes longer (Figure 3). This pattern of forecast errors is statistically significant even after we control for a variety of country characteristics that may systematically influence forecast errors.
Figure 3. Mean and median forecast error by forecast horizon (percentage points)
Sources: WEO (1990-2012 vintages) and authors' calculations.
Note: forecast error = forecast – actual, actual data as of December 2013.
Searching for an explanation of optimism bias
By way of conclusion, we offer some tentative thoughts on the possible factors underlying the over-optimism found in medium and long-term growth forecasts. It is well known, from the seminal work of Easterly et al. (1993), that what the economics profession considers as the most likely “fundamental” determinants of economic growth (such as institutional quality, education, and macroeconomic policies) are persistent, whereas economic growth itself is not. Faced with a country that has been growing rapidly and broadly similar “fundamentals,” a forecaster would be hard pressed to justify why she expects growth would be lower going forward compared to the recent past. By contrast, for countries whose growth has been weak – possibly due to an economic or political crisis, or even a civil war – the forecaster is unlikely to be able or willing to assume that similar adverse shocks would recur.
Ahir, Hites and Prakash Loungani (2014), “Fail Again? Fail Better? Forecasts by Economists During the Great Recession,” George Washington University Research Program in Forecasting Seminar.
Easterly, William, Michael Kremer, Lant Pritchett, and Lawrence H. Summers (1993), “Good Policy or Good Luck? Country Growth Performance and Temporary Shocks,” Journal of Monetary Economics, Vol. 32, No. 3, pp. 459–483.
Ho, Giang and Paolo Mauro (2014), “Growth: Now and Forever?” IMF Working paper 14/117.
Juhn, Grace and Prakash Loungani (2002), “Further Cross-Country Evidence on the Accuracy of the Private Sector’s Output Forecasts,” Vol. 49, No. 1, IMF Staff Papers, pp. 49–64.
Pritchett, Lant, and Lawrence H. Summers (2013), “Asiaphoria Meet Regression to the Mean,” Harvard University.