Economists traditionally view fluctuations in economic variables as structural and long-run or cyclical and short-run. In the analysis of economic activity, we distinguish between potential output and business cycle fluctuations, while labor market economists often distinguish between natural and cyclical unemployment. Scholars analyze structural and cyclical factors separately, even within different disciplines such as growth theory and business cycle analysis. Moreover, this short-run/long-run distinction extends beyond output and employment to other macro-financial variables.
Among these variables, capital flows are particularly important for emerging market and developing countries (EMDEs). For example, in these countries, capital outflows can bring about a sudden halt and generate significant negative social costs by triggering currency depreciation, higher inflation, lower output, and sometimes even financial crises (Ostry et al., 2012; Fernández et al., 2015). In line with this, Burger et al. (2022) introduce a measure of capital flows. Natural Level of Capital Flows And they show that when observed flows are large and deviate significantly from this level, they usually subsequently stop abruptly.In studies of capital flows, as with other macro-financial variables, a distinction is usually made only between short-run and long-run fluctuations.
However, this differentiation faces a major challenge, as several variables, including credit growth, asset prices, and even productivity and GDP growth, also exhibit important fluctuations in the medium term. In the presence of financial frictions, credit supply and asset demand depend on investors' beliefs about borrowers' default probabilities (Aikman et al., 2014). Because these expectations depend on the behavior of others, a small change in fundamentals that prompts some investors to reassess their beliefs may prompt most investors to follow suit, leading to larger and longer fluctuations in credit growth and asset prices. For example, a small increase in economic growth within a traditional business cycle may prompt many investors to reassess their default probabilities and grant more credit. This increase in credit supply may then lead to a large and sustained expansion of credit and asset prices beyond the time frame of a typical business cycle.
Similarly, productivity growth affects GDP in the medium run, and not just in the long run as traditionally thought. This is because the adoption of new technologies is driven by investments in R&D, which have booms and busts. Periods of high levels of investment in R&D accelerate the adoption of technology, leading to long periods of above-average and below-average productivity growth. This phenomenon is reflected in the “medium-run business cycle” of GDP and other variables (Comin and Gertler, 2006).
Medium-term fluctuations are related to several variables, demonstrating that it is not sufficient to limit the analysis to the short and long term. It is the medium-term rather than the short-term cycle of the deviation of capital flows from their natural value (KF*gap) defined by Burger et al. (2022) that explains the ability to predict sudden halts. We calculate short- and medium-term deviations from the KF* gap for several EMDEs and conduct logit regressions to estimate their individual impacts on the likelihood of sudden outages over different time frames. Following Burger et al. (2022) and Forbes and Warnock (2021), we control for world GDP growth, changes in the world money supply and world monetary policy rates, the VIX index, oil prices, and regional real GDP growth. In a recently published but yet to be published working paper, we also conduct the same exercise to assess the impact on vulnerability to global shocks.
The results show that the predictive power of each component of the KF*Gap varies significantly (Figure 1). The business cycle component significantly predicts sudden stops only six quarters ahead. In contrast, the medium-term component predicts sudden stops as early as one quarter ahead and remains significant over longer periods. When the business cycle component is two standard deviations above the mean, the probability of a sudden stop increases slightly from 5% to 10%. However, for the medium-term component, this probability rises dramatically from 5% to 25%-53%. Thus, The KF* gap cannot be assumed to simply reflect business cycle trends: its medium-term variations are essential to more accurately predict sudden stops.
