The causes of South Africa's next bout of capital outflows

Background

Capital flows have far-reaching implications for monetary, fiscal and financial policy. South Africa has a relatively large current account deficit that is financed by capital inflows. A reversal of capital inflows could have serious economic consequences. While the economic effects of capital flow reversals have been studied for South Africa, less is known in terms of what prompts capital flow reversals. This is the central question addressed in this note. Our analysis shows that the probability of a capital flow reversal increases in relation to:

•     Higher debt service costs.
•     Slower economic growth.
•     Sovereign ratings downgrades.
•     Higher government debt.

The effects and causes of capital flow reversals

Empirical work shows that capital flow reversals have a negative impact on the economy (for a good summary see Smit et al. (2013)). Capital flow reversals cause:

•     Sharp currency depreciations.
•     Declining economic activity.
•     Declining asset prices.
•     Current account reversal if unaccompanied by reserve buffers.

Smit et al. (2013) shows that a capital reversal of 50% reduces economic growth by 0.3 percentage points in the first year and by 1 percentage point the following year. The 10 year government bond yield increases by 3.2 percentage points in the first year following the capital flow reversal and  by a further 1.7 percentage points the next year.

Studying the determinants of capital flow reversals is justified given its effects on the economy. It is also important in the context of economic movements recently – South Africa needs to be aware that the potential for capital flow volatility increases as the Fed tapers down its quantitative easing programme. At the same time, SA policy makers need to be cognisant of the effects of a possible EU quantitative easing (QE) programme. By no means does another QE imply an increase in capital flows – this depends on the factors that influence flows (the core research question of this note).

The literature usually cites growth differentials, interest rate differentials, foreign exchange reserve, prices, financial policies and fiscal sustainability as factors that influence capital flows. We study the impact of some of these factors on capital flow reversals. A short description of the possible effects of these variables on capital flows are summarised in Table 1:

Table 1: The influence of macroeconomic variables on capital flows

Variable	Effect
Interest rates	Higher interest rates provide higher yields for foreign investors. These yields could lead to higher capital inflows. However, these yields need to be adjusted for risk. If the risk adjusted interest rate is still low, or when a country’s public finances are perceived to be unsustainable, then changes to the interest rate could have no effect on capital flows, or even lead to a reversal if rates lead increases the probability of debt default.
GDP growth differentials	Higher GDP growth could lead to an increase in net capital inflows. This often serves, alongside the stock market, as an indication of potential future gains for investment.
Expectations	Expectations regarding the financial stability of a country are important is assessing whether foreigners will invest or not. We assume that these expectations can be measured by a country’s risk rating (caution – this is usually only a measure of risk regarding a country’s foreign denominated debt). It is expected that there will be an outflow of capital when expectations worsen, i.e. a lower rating.
Inflation	Investors are often interested in real returns to investment. Inflation erodes those returns. In inflation targeting countries, high inflation would mean higher interest rates. These higher interest rates in return would reduce economic growth.
Exchange rates	It is not the level of the exchange rate that might cause an inflow or outflow, but the view about whether the exchange rate will depreciate or appreciate. While exchange rates are endogenous to capital flows, we model exchange rate deviations from equilibrium to proxy foreign investors’ views on currency movements. As an example, an investor would want buy goods in domestic currency cheaply and sell it when the currency depreciates. Here it is assumed that investors analyse this from an equilibrium perspective – assuming that any movement away from equilibrium will move back to equilibrium.

Methodology

We are interested in variables that increase the probability of a capital flow reversal from an empirical perspective. The explanatory variables include South Africa’s GDP growth differential with G7 GDP growth, debt service costs, the interest rate differential between South Africa and USA’s federal funds rate, foreign reserves as a per cent of GDP, Fitch sovereign ratings, sovereign debt as a per cent of GDP, high interest rates (measured by squaring interest rates) and inflation. Having so many explanatory variables in a regression framework can easily bias the results making inference about the size and sign of the explanatory variables impossible. As such, we use a model[1] that explicitly takes account a large number of variables without biasing the statistical significance of the estimates. Our model is estimated over 1997 to 2013q1. Our measure of capital flow reversals is measured as a binary variable that equals 1 whenever net capital flows as a per cent of GDP is less than zero and equals zero otherwise.[2] The model is set up in a way that multiple combinations of equations are estimated. In total 2^9 (512) models are estimated (there are nine variables). Our methodology allows us to evaluate the parameter distribution - The distribution helps us to assess the significance of the coefficients (i.e. how far the mode, mean and median deviates from zero) as well as whether certain variables are more important than others (as measured by the Posterior Inclusion Probability (PIP)). The PIP varies between 0 and 100, where 100 indicate that a variable was significant in modelling capital flow reversals in all 258 model combinations.

Results

Table 1 summarises the first set of results. The mean coefficient should be interpreted with caution. The model is a probit model and the results do not have an elasticity interpretation. The mean’s sign, however, is important. We see that higher GDP growth differentials and higher levels of reserves reduce the probability of capital flow reversals. Higher GDP growth differentials imply that macroeconomic fundamentals are good relative to the rest of the world and serves as a signal for potential investors. Higher reserves imply a higher probability of being able to absorb adverse economic shocks better. Higher debt service costs, higher sovereign debt and another sovereign ratings downgrade increase the probability of capital flow reversals. The probability of capital flow reversals for South Africa decreases in the case of higher interest rate differentials. Higher interest rate differentials can attract capital due to higher returns.

Table 1: The determinants of capital flow reversals

	PIP	Mean	SD
Growth differential	10.8	-0.02	0.09
Debt service costs (DSC)	34.3	0.26	0.09
Reserves	65.5	-0.23	0.20
Interest rate	5.3	-0.01	0.04
Inflation (infl)	9.4	0.01	0.06
Ratings (Fitch)	37.5	0.41	0.62
Equilibrium fx	15.9	0.01	0.03
Debt	18.0	0.02	0.06
Very high interest	10.3	0.00	0.00

One way to interpret the results is to analyse the probability of capital outflows over different values for our explanatory variables (everything else is evaluated at their respective means). Figure 1 shows that the probability of a sudden stop varies over different shares of reserves to GDP, different GDP growth rates and different sovereign ratings. The probability of a sudden stop is then compared when debt service costs are moderately high versus when debt service costs as a per cent of GDP is zero.

Figure 1: Probability of capital flow reversals

The vertical axes show the probability of capital flow reversals (outflows). If it equals 1 then capital flow reversals are a certainty. The horizontal axis measures the actual levels of reserves, GDP growth and ratings respectively. Reserves as a per cent of GDP vary from 3 per cent to 10 per cent as an example. The ratings are assigned numerical values where the highest rating, AAA, is assigned a 1. As is expected, the probability of a capital flow reversal decreases alongside the accumulation of reserves, higher credit scores and higher growth differentials. Interestingly, higher debt service costs are associated with a higher probability of capital outflows. In the case of having positive growth differentials, debt service costs matter a lot in terms lower the event of capital outflows. The probability of capital outflows is larger when the exchange rate has deviated far from equilibrium. Currency deviations from equilibrium often imply possible currency speculation – this can greatly affect the movement in capital flows.

Conclusion

Capital flow reversals could come about due to a number of reasons. Poor performing macroeconomic indicators such as slow GDP growth and low-rated sovereign bonds could result in an outflow of capital. Our results show that a higher level of reserves serve as a signal to manage potential economic shocks, and hence reduces the probability of a sudden stop. Higher debt service costs, alongside higher government debt increases the probability of capital flow reversals substantially.

There are some interesting policy considerations that emerge from this analysis. If the objective is to avoid an altogether outflow of capital then there are a couple of policy options. Unfortunately policy options that worked for one country during a particular period might not be that effective for another country (see Magud et al. (2011) on the effectiveness of different capital controls). It should be useful to rank and quantify the effects of various policies that mitigate capital outflows. This reduces the risk of getting things seriously wrong – such as unattended consequences of a tax on speculative flows. A convincing proposal has been put forth by Korinek (2010) to impose a Pigovian tax on inflows to mitigate possible amplification effects, or externalities, caused by outflows. Korinek (2010) using a welfare theoretic foundation for risk-adjusted capital regulations, calculates the externalities caused by various types of flows for Indonesia. He shows that externalities are amplified during crises periods. The largest externality from flows comes from dollar debt, followed by inflation linked debt. The least distortionary flows come from non-financial FDI and portfolio investments.  Regrettably little is understood regarding the macroeconomic effects of different types of flows since most studies use only aggregate measures. Thus, the correct policy response should control for the type of flows too controlling for country specific effects.

References

Magud, N., Reinhart, C.M. and Rogoff, K.S. 2011. Capital controls: Myth and reality – A portfolio balance approach. National Bureau of Economic Research. NBER Working Papper 16805.

Korinek, A. 2010. Regulating capital flows to emerging markets: An externality view. University of Maryland working paper.  

---

[1] We use a Bayesian Model Averaging (BMA) that estimates multiple combinations of models and averages out the coefficients. We use a flat prior indicating our lack of knowledge of the importance and size of the different variables. This implies that the likelihood function has a stronger weight than any prior chosen by the researcher. 

[2] There are alternative measures such as any deviation in capital flows of more than one standard deviation.