The Volatility of Capital Flows in Emerging Markets: Measures and Determinants

Maria Sole Pagliari; Swarnali A Hannan

doi:10.5089/9781475585254.001.A001

Jump to Content

The Volatility of Capital Flows in Emerging Markets

Measures and Determinants

Author:

Maria Sole Pagliari

Maria Sole Pagliari https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Maria Sole Pagliari in
Current site
Google Scholar

and

Mrs. Swarnali A Hannan

Mrs. Swarnali A Hannan
Search for other papers by Mrs. Swarnali A Hannan in
Current site
Google Scholar

Contributor Notes

Authors’ E-Mail Addresses: mpagliari@economics.rutgers.edu, sahmed@imf.org

Publication Date:: 07 Mar 2017

eISBN:: 9781475585254

Language:: English

Keywords:: WP; capital flow; flow volatility; volatility estimate; standard deviation; GARCH estimate

Download PDF (5.4 MB)

Capital flow volatility is a concern for macroeconomic and financial stability. Nonetheless, literature is scarce in this topic. Our paper sheds light on this issue in two dimensions. First, using quarterly data for 65 countries over the period 1970Q1-2016Q1, we construct three measures of volatility, for total capital flows and key instruments. Second, we perform panel regressions to understand the determinants of volatility. The measures show that the volatility of all instruments is prone to bouts, rising sharply during global shocks like the taper tantrum episode. Capital flow volatility thus remains a challenge for policy makers. The regression results suggest that push factors can be more important than pull factors in explaining volatility, illustrating that the characteristics of volatility can be different from those of the flows levels.

Abstract

I. Introduction¹

Capital flow volatility is a major source of concern for macroeconomic and financial stability in emerging market and developing economies (EMDEs). Over the last three decades, the emerging economies have liberalized their capital accounts and have become more integrated in international financial markets. However, as the Global Financial Crisis (GFC) has shown, reaping the benefits of capital account liberalization while containing the associated risks remains a key challenge for many countries for several reasons.

First, EMDEs tend to receive capital flows that, even in net terms, are large relative to their domestic economies and overall absorptive capacity in terms of the size and depth of their financial systems.

Second, EMDEs are more vulnerable to shocks, partly because their economies are smaller and less diversified, and because they have less domestic economic and political stability. In addition, shocks of any kind - positive or negative, domestic or external - are exacerbated and propagated more easily in EMDEs because of structural and institutional characteristics. In particular, large capital inflows, mostly intermediated through the banking system, tend to amplify the domestic financial and real business cycles to a greater extent than in advanced economies (AEs).

Finally, the GFC has prompted new concerns that flows to EMDEs are overly sensitive to some global (push) factors that are beyond the influence of domestic policies.

Given all this, financial integration poses serious challenges to economic and financial sector stability in EMDEs.² To the extent that there exists a positive link between the stability of capital flows and economic growth (see Easterly, Islam, and Stiglitz 2001 and Ramey and Ramey 1995), maintaining a steady stream of capital flows is a policy priority in most EMDEs. Despite being a focus of policymakers, relatively few studies focus primarily on the issue of volatility. Moreover, capital flow volatility is usually quantified in policy and academic discussions using a crude measure of standard deviation of flows over a specified period. Given the importance of this issue, a better understanding of volatility in capital flows is warranted. This paper aims at providing a deeper understanding of volatility of capital flows by contributing to the existing literature along three important dimensions:

1) Measurements: The first step of this study consists of deriving measures that can usefully describe volatility. Our paper builds upon Broto et al. (2011) to provide updated capital flow volatility measures using three methodologies: rolling window standard deviation, GARCH(1,1) conditional variance and ARIMA(1,1,0) estimates.

2) Determinants: In order to pursue policies to increase resilience and help manage capital flow volatility, it is important to understand the drivers of capital flow volatility. We perform panel regressions to understand what are the key variables driving volatility, by disentangling between, on the one hand, global or push factors and, on the other hand, domestic or pull factors.

3) Components: We analyze the evolution of volatility across a wide array of instruments. In particular, the high level of granularity of our dataset allows us to achieve a greater precision in the study of particular categories of flows.

Using quarterly data over the period 1970Q1-2016Q1, we use three measures of volatility to document its evolution over time. Our measures show that net capital flows to EMDEs tend to be more volatile than those of AEs, as gross outflows tend to dampen the impact of gross inflows in the latter group but not in the former. Our computed measures for gross inflows in EMDEs suggest that portfolio and other investments are respectively around two and four times more volatile than FDI. In addition, amongst portfolio flows, portfolio debt is more volatile than portfolio equity. As to other investment flows, private flows are much more volatile than the official sector ones. Within private flows, bank flows usually tend to be more volatile than non-bank flows, but there are periods where this does not hold.

For policymakers, it is not only the level of volatility but also the changes in volatility that matter. Therefore, we use our estimates to document the changes of volatility since 2000. Our computed measures suggest that, after a spike during the GFC, capital flow volatility in EMDEs is now back to pre-crisis levels. However, volatility has sharply risen in response to some after-crisis global shocks, particularly during the taper tantrum episode. Moreover, after accounting for the global slowdown in capital flows witnessed since 2012, EMDEs capital flow volatility has either increased or remained the same as before the crisis. There exists, however, a degree of heterogeneity in the results, depending on the type of flows targeted. Indeed, while foreign direct investments (FDIs) display a more clear-cut and stable volatility pattern over time, the same does not hold for other instruments, like portfolio or other flows. In particular, debt-generating instruments, notably portfolio debt and bank flows, were the main drivers of the large swings in volatility observed soon before and in the aftermath of the GFC.

One of the major contributions of our paper is that we produce capital flow volatility estimates, both for total flows and different instruments, for 65 individual countries.³ The individual country estimates suggest that there can be significant variation amongst countries, both in terms of the evolution over time and the relative magnitude across instruments, underscoring the importance of monitoring volatility of individual countries. These estimates can thus be useful tools in future studies dealing with cross-country comparisons.

We, then, use the volatility estimates at the country level to run panel regressions with the aim of understanding the determinants of volatility. Our regression analysis shows that push factors like the US monetary policy stance, US economic performance, and global risk aversion do influence the volatility of capital flows in EMDEs to an extent which, in certain cases, is greater than the effect of domestic and structural variables. These findings are interesting against the backdrop of recent studies (e.g., IMF 2016) that show how the recent capital flow slowdown can be predominantly attributed to the broader emerging markets economic growth slowdown. Our results suggest that the determinants of capital flow volatility can be quite different from the determinants of capital flow levels.

The remainder of this paper is structured as follows: Section 2 provides some stylized facts concerning the evolution of capital flows to EMDEs and discusses the different types of data included in the dataset; Section 3 reviews the existing literature; Section 4 contains an overview of the methodological approaches adopted and analyzes the evolution of volatility of capital flows over time, across instruments and countries; Section 5 deals with the identification of the main drivers of volatility via panel regressions; Section 6 concludes.

II. Evolution Of Capital Flows

A. Heterogeneity across groups and flows: AEs vs EMDEs, gross vs net.

International capital gross inflows have experienced a remarkable rise from the mid-1990s to the first half of 2000s, both in EMDEs and AEs (Figure 1). Following the GFC in 2008, inflows dropped sharply in 2008 for both groups and regained their upward momentum in 2009, only to fall again in late 2011 as the peripheral Euro Area sovereign debt crisis intensified. Within this broader trend, there exist key differences across income groups. While net flows in AEs are influenced by both gross inflows and gross outflows, their patterns for EMDEs is mainly determined by gross inflows (Figure 1). This is because capital flows to EMDEs have been generally driven by foreign investments into domestic assets, while the amount of domestic investments abroad have played a smaller role.⁴ Given this background, when it comes to evaluating the volatility of capital flows, it is crucial to make a clear distinction not only across groups, but also between gross and net flows.

Figure 1.

Cross border capital flows in AEs and EMDEs.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: flows are expressed as % share of group GDP. Dashed lines are quarterly moving averages. Data are quarterly over the period 1980Q1:2016Q1.Source: IMF, International Financial Statistics.

Table 1 reports some basic volatility measures of gross and net flows for both groups, computed over the period 1990Q1:2016Q1. The measures of volatility considered are two: the standard deviation of capital flows as a share of GDP and the coefficient of variation, that is the ratio between the standard deviation and the absolute value of the mean of the flows. Standard deviation is a measure suitable for making comparisons across groups whose average levels of flows are similar. When the means are widely different, however, the higher the mean the higher could be the dispersion of observations around it. In order to address this issue, the coefficient of variation is often a preferred measure to make comparisons across groups.⁵ The coefficient of variation could be problematic when the mean is close to zero, since the coefficient of variation will approach infinity and will be highly sensitive to small changes in the mean.

Table 1.

Statistics for quarterly moving averages.

Notes: average and standard deviation are expressed as % share of group GDP; (1) Gross Inflows; (2) Gross Outflows; (3) Net Inflows. Δ_y1/y2-x1/x2: difference between the averages computed over periods x₁/x₂ and y₁/y₂.

Table 1.

Statistics for quarterly moving averages.

	Average			Standard deviation			Coefficient of variation
	(1)	(2)	(3)	(1)	(2)	(3)	(1)	(2)	(3)
AEs
1990-1999	3.75	0.38	3.37	1.86	0.30	1.84	0.50	0.79	0.55
2000-2007	14.49	1.52	13.03	5.15	0.42	5.25	0.36	0.27	0.40
2008-2015	5.77	0.77	5.12	4.32	4.33	4.34	0.75	5.62	0.85
A 2000/07-1990/99	10.75	1.14	9.67	3.29	0.11	3.41	−0.14	−0.52	−0.14
A 2008/15-2000/07	−8.72	−0.75	−7.92	−0.83	3.91	−0.91	0.39	5.35	0.44
A 2008/15-1990/99	2.02	0.39	1.75	2.46	4.02	2.50	0.25	4.82	0.30
EMEs
1990-1999	2.21	1.62	0.76	1.79	1.56	0.57	0.81	0.96	0.74
2000-2007	4.94	2.40	2.55	2.40	1.10	1.61	0.48	0.46	0.63
2008-2015	4.81	1.85	2.92	1.78	1.78	1.77	0.37	0.96	0.60
A 2000/07-1990/99	2.73	0.77	1.79	0.60	−0.46	1.04	−0.33	−0.50	−0.11
A 2008/15-2000/07	−0.13	−0.54	0.38	−0.62	0.68	0.16	−0.11	0.50	−0.03
A 2008/15-1990/99	2.60	0.23	2.16	−0.01	0.22	1.20	−0.44	0.00	−0.14

Table 1.

Statistics for quarterly moving averages.

	Average			Standard deviation			Coefficient of variation
	(1)	(2)	(3)	(1)	(2)	(3)	(1)	(2)	(3)
AEs
1990-1999	3.75	0.38	3.37	1.86	0.30	1.84	0.50	0.79	0.55
2000-2007	14.49	1.52	13.03	5.15	0.42	5.25	0.36	0.27	0.40
2008-2015	5.77	0.77	5.12	4.32	4.33	4.34	0.75	5.62	0.85
A 2000/07-1990/99	10.75	1.14	9.67	3.29	0.11	3.41	−0.14	−0.52	−0.14
A 2008/15-2000/07	−8.72	−0.75	−7.92	−0.83	3.91	−0.91	0.39	5.35	0.44
A 2008/15-1990/99	2.02	0.39	1.75	2.46	4.02	2.50	0.25	4.82	0.30
EMEs
1990-1999	2.21	1.62	0.76	1.79	1.56	0.57	0.81	0.96	0.74
2000-2007	4.94	2.40	2.55	2.40	1.10	1.61	0.48	0.46	0.63
2008-2015	4.81	1.85	2.92	1.78	1.78	1.77	0.37	0.96	0.60
A 2000/07-1990/99	2.73	0.77	1.79	0.60	−0.46	1.04	−0.33	−0.50	−0.11
A 2008/15-2000/07	−0.13	−0.54	0.38	−0.62	0.68	0.16	−0.11	0.50	−0.03
A 2008/15-1990/99	2.60	0.23	2.16	−0.01	0.22	1.20	−0.44	0.00	−0.14

In AEs, capital flow volatility, as indicated by the coefficient of variation, has increased over the period 1990-2016, by 0.25, 4.82 and 0.3 for gross inflows, gross outflows and net inflows respectively. In all cases, there has been a slight decrease in volatility up to the GFC, while it has increased during and after it (see Table 1). In EMDEs, on the other hand, the volatility of gross inflows and net inflows have fallen by 0.44 and 0.14 respectively over the same period.

B. Heterogeneity across instruments: FDI, portfolio and Other flows

Besides the heterogeneities related to income groups and to the comparison between gross and net flows, another powerful source of heterogeneity stems from the composition of the aggregate flows. The overall dynamics of capital flows depend upon the dynamics of the specific components and, by analyzing the composition of the financial account, one can infer interesting information about the behaviour of international investors in different scenarios as well as have a better understanding of a country’s financial stability.

The three instruments that are usually closely monitored are⁶:

1. Foreign Direct Investments (FDI), “a category of cross-border investments associated with a resident in one economy having control or a significant degree of influence on the management of an enterprise that is resident in another economy ”;
1. Portfolio flows, “defined as cross-border transactions and positions involving debt or equity securities, other than those included in direct investment or reserve assets”;
2. Other flows, “a residual category that includes positions and transactions other than those included in direct investment, portfolio investment, financial derivatives and employee stock options, and reserve assets”, classified in government-related flows and private flows (bank and non-bank flows).

Figure 2 shows the evolution of gross and net inflows for EMDEs, broken down by instruments. While FDI represent the majority of capital inflows for EMDEs, the surge and the subsequent collapse in inflows during the GFC was mainly driven by other flows, which include bank flows. Even the post crisis pick-up in capital flows was led by debt-creating (bank and portfolio debt) instruments, with some reversals between 2015 and 2016. Therefore, one would expect different results when estimating volatility for different categories, with portfolio and other flows being characterized by a higher volatility.

Figure 2.

Quarterly moving average of Gross and Net Capital Inflows in EMDEs by instrument.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: flows are expressed as % share of group GDP.Source: IMF, International Financial Statistics.

Another aspect of particular relevance concerns the correlation across instruments and whether it has changed over time. In this regard, Figure 3 shows how the correlation coefficients across different types have evolved over time for EMDEs. The correlations across instruments are significantly different from 0 in certain periods. The most interesting case is provided by the correlation between FDI and portfolio flows, whose sign has switched from negative to positive and vice versa several times in the period covered. By comparing this picture with Figure 2, it is evident that episodes of surges coincide with a positive correlation between FDI and portfolio flows, while the same correlation turns negative soon after a stop. By contrast, this does not seem to be the case when it comes to co movements of both FDI and portfolio with other investments. The key message, however, is that there exists a degree of heterogeneity across instruments.

Figure 3.

Evolution of the correlation coefficient across different types of Gross Inflows.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: the grey, blue and red dots indicate estimates significant at 10%, 5% and 1% respectively.Source: IMF, International Financial Statistics, authors’ computations.

III. Literature Review

There exists a vast literature aimed at, on the one hand, identifying the main drivers of the evolution of international capital flows, and, on the other hand, measuring the potential gains from financial integration.⁷

A. The Determinants of Capital Flows

Most of the literature classifies the main determinants of capital flow developments in country-specific (pull) and global (push) factors (Koepke 2015). Among the latter, the most commonly identified are indicators of global risk appetite and US monetary policy. Reinhart et al. (1993) and Reinhart et al. (1996), for instance, argue that capital inflows to Latina America in the 1990s were influenced by conditions originated outside the region, a fact that increased the macroeconomic vulnerability of the economies in that region.

More recent literature explores whether push or pull factors determine capital flows. Fratzscher (2012), by means of a factor model, shows how push factors were overall the main drivers of capital flows during the crisis, while pull factors have been dominant in accounting for the dynamics of global capital flows in 2009 and 2010, in particular for EMDEs. In their analysis of episodes of “surges”, “stops” (sharp increases and decreases, respectively, of gross inflows), “flight” and “retrenchment” (sharp increases and decreases, respectively, of gross outflows), Forbes and Warnock (2012) find that global risk is significantly associated with extreme capital flow episodes. Moreover, contagion, whether through trade, banking, or geography, is also associated with stop and retrenchment episodes, whereas domestic macroeconomic characteristics are generally less important. Agrippino and Rey (2014) find that one global factor explains an important part of the variance of a large cross section of returns of risky assets around the world. They interpret this global factor as reflecting the time-varying degree of market wide risk aversion and aggregate volatility, represented by the VIX, the Chicago Board Options Exchange Volatility Index. On top of that, they show that US monetary policy is a driver of this global factor in risky asset prices, the term spread and measures of the risk premium. Passari and Rey (2015) present evidence on the existence of a global financial cycle in gross cross-border flows, asset prices and leverage and identify the VIX as a suitable proxy for such cycle. This implies that there exist significant spillovers from the US monetary policy on capital flows to the EMDEs.

Other findings are provided by Ahmed and Zlate (2014), who highlight how growth and interest rate differentials between EMDEs and AEs and global risk appetite are statistically and economically important determinants of net private capital inflows to EMDEs. They also show that there is a positive effect of unconventional US monetary policy on EMDEs inflows, especially portfolio inflows, and that there have been significant changes in the behaviour of net inflows from the pre-crisis to the post-crisis period, partly explained by the greater sensitivity of such flows to interest rate differentials. Similarly, Ghosh et al. (2014b) find that global factors like US interest rates and VIX play a crucial role in determining capital surges to EMDEs. However, the occurrence and magnitude of a surge towards a particular EMDE also largely depend on domestic factors such as its external financing need, capital account openness, and exchange rate regime. Nonetheless, surges driven by exceptional behaviour of liability flows are relatively more sensitive to global factors and contagion.

Siemer et al. (2015) study the relationship between stock market volatility as a proxy for uncertainty and capital inflows using panel regressions and a portfolio choice model. They decompose each country’s market return volatility into two components: systemic volatility, measured by uncertainty betas, and country-specific volatility. While it is true that capital inflows generally respond to changes in both components, however in countries with higher uncertainty betas capital inflows tend to be more sensitive to risk.

B. The Volatility of Capital Flows

The existing empirical literature dealing with the volatility of capital flows can be classified into two strands.⁸ On the one hand, some papers focus on analyzing the difference in volatility between the capital flows to emerging and advanced economies. For instance, Rigobon and Broner (2005) show that the higher volatility in EMDEs is primarily due to these economies’ propensity to build up imbalances, which generates more persistent shocks and a higher likelihood of international contagion. Similarly, Alfaro et al. (2007) emphasize the importance of domestic factors, such as institutional quality and the soundness of macroeconomic policies, in explaining these volatility differences. Discrepancies between EMDEs and AEs can also be characterized by type of investment. For instance, the gap between the volatility of FDI and portfolio flows is found to be smaller in AEs (Goldstein and Razin 2006), whereas in EMDEs the share of FDI in total capital inflows is higher (Albuquerque 2003), as is the volatility of their portfolio flows (Tesar and Werner 1995).

Other contributions use panel data models to analyze the impact of financial integration on volatility. In this regard, Neumann et al. (2009) show that financial integration tends to increase the volatility of FDI in emerging economies, while reducing that of other debt flows in mature economies. In addition, Bekaert and Harvey (1997), Lagoarde-Segot (2009) and Umutlu et al. (2010), by focusing on prices rather than quantities, conclude that financial liberalization reduces the volatility of stock market returns in emerging economies.⁹ In a more recent study, Broto et al. (2011) analyze the determinants of the volatility of various types of net capital inflows to EMDEs, by using annual data over the period 1980-2006. Their main finding is that, since 2000, global factors have become increasingly significant relative to country-specific drivers. However, they also identify some domestic macroeconomic and financial factors that appear to reduce the volatility of certain instruments without increasing that of the others.

Alberola et al. (2016) investigate the role of international reserves as a stabilizer of international capital flows, in particular during periods of financial distress. They show that international reserves facilitate financial disinvestment overseas by residents. This partially offsets the drop in foreign capital inflows observed in such periods. Moreover, larger stocks of international reserves are linked to higher gross inflows and lower gross outflows.

In a more recent paper, Broner and Ventura (2016) investigate the effects of financial globalization in EMDEs. By emphasizing the role of imperfect enforcement of domestic debts and the interactions between domestic and foreign debts, they show that financial globalization can produce three different outcomes: i) domestic capital flight and ambiguous effects on net capital flows, investment and growth; ii) capital inflows and higher investment and growth; iii) volatile capital flows and unstable financial markets. These effects ultimately depend on the level of development, productivity, domestic savings and the quality of institutions. In this paper we empirically investigate some of these aspects by expanding the work of Broto et al. (2011) to include the GFC period.

IV. Volatility Measures

The first way to expand the existing literature is to build a dataset with broader coverage, in terms of both the time period considered and the instruments included. The empirical papers previously cited use annual data that go until 2010 and do not cover all the different categories included in the balance of payments. Our paper uses quarterly data, spanning from 1970Q1 to 2016Q1, and encompasses a wider set of capital flow instruments.¹⁰ These features allow us to carry on a more in-depth analysis of how volatility has evolved both over time, across countries and across instruments. In addition, while current literature has focused on the capital flow slowdown witnessed since 2012 (IMF 2016), our updated data sheds light on how capital flow volatility has evolved since then.

A. Methodology

In order to produce reliable estimates of capital flow volatility, we consider three approaches, building on Engle et al. (2005) and Broto et al. (2011):

Standard deviations over a rolling window (RW) This consists of the standard deviation of capital flows computed over a rolling window as follows:
$σ_{i t} = (\frac{1}{n} \sum_{k = t - (n - 1)}^{t} (f l o w_{i k} - μ)^{2})^{1 / 2},$

where $μ = \frac{1}{n} \sum_{k = t - (n - 1)}^{t} f l o w_{i k}$ and flow_ik denotes capital flows in country i in period k.

In these computations, n = 4 quarters.
- As mentioned in Broto et al. (2011), there are some caveats associated with this measure:
  - a) there is a loss of data at the beginning of the sample, whose entity depends on the window length (n);
  - b) σ_it is strongly persistent and this might entail problems of endogeneity and serial correlation;
  - c) the same weights are assigned to flow_i(t-1) and flow_it-(n-1), which overly smooths volatility. As a result, volatility might be underestimated when a shock takes place and overestimated thereafter, compared to other measures.
Estimated standard deviations produced by a GARCH(1,1) model

The second alternative measure makes use of estimated conditional volatilities of a GARCH(1,1) model. The process is defined as follows:
$\begin{matrix} y_{i t} = ɛ_{i t} σ_{i t} \\ σ_{i t}^{2} = α_{0} + α_{1} y_{i (t - 1)}^{2} + α_{2} σ_{i (t - 1)}^{2}, \end{matrix}$

where y_it = Δflow_it, ε_it is a Gaussian white noise process and $σ_{i t}^{2}$ is the corresponding conditional variance.

This measure, however, has its own drawbacks. Notably:
- a) data scarcity might lead to convergence errors;
- b) Maximum-Likelihood estimates for small samples contain biases;
- c) stationarity and positivity of estimates require that $\hat{α_{1}} + \hat{α_{2}} < 1, \hat{α_{0}} > 0, \hat{α_{1}} >$ 0 and $\hat{α_{2}} > 0$ . If these conditions are violated for some country i, then the model fails to produce valid estimates for that country;
- d) in some cases, the residuals do not present ARCH effects and that makes the GARCH model not suitable.
Estimated standard deviations produced by an ARIMA(1,1,0) model

The third measure is given by the standard deviation of the residuals obtained after fitting an ARIMA(1,1,0) to the data.¹¹ Notably, we first estimate the residuals from the following AR(1) process:
$Δ f l o w_{i t} = c + β Δ f l o w_{i (t - 1)} + v_{i t}$

Then, a test is performed to detect the presence of any ARCH effects in the residuals. If the null hypothesis of heteroscedasticity is rejected:
$σ_{i t}^{2} = \frac{1}{4} Σ_{j = t - (n-3)}^{t + (n-2)} (v_{i j})^{2}$

otherwise conditional volatility is estimated by fitting a GARCH(1,1) model to the residuals.

The heteroscedasticity checks on the ARIMA residuals constitute a deviation from the reference literature, as we deem this further step to be helpful in improving the robustness of the resulting estimates vis-à-vis possible misspecifications. Moreover, this procedure is useful in addressing the shortcomings related to the GARCH(1,1) procedure explained above. For these reasons, we prefer the ARIMA estimates and use it as the benchmark measure in our regression analysis.

B. Evolution of Volatility over Time

Using the volatility estimates from the three measures, this section considers how the volatility of capital flows has evolved over time. We first look at the aggregate flows both globally and for AEs and EMDEs separately.¹² We then study the evolution of volatility for the single components of gross inflows in EMDEs only.

Total flows

When discussing volatility, policymakers are interested in both the levels as well as the changes in volatility. Using our measures, we shed light on both aspects in this section.

Figures 4a, 5a and 6a show the evolution of volatility of net flows for World, AEs and EMDEs respectively, using the three different measures. As discussed in section IV.A, the charts clearly show that RW underestimated the spike in volatility at the onset of the crisis and overestimated the fall in volatility following the end of the crisis. This underscores the need for going beyond the crude RW measure. Comparing the three measures, the GARCH estimates tend to lag slightly behind the RW and the ARIMA, and are on average higher than the estimates produced by the other two approaches. However, broadly speaking, the three estimates show similar results.

Figure 4.

Evolution of the volatility of capital flows for the World aggregate.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: Measures are expressed as % share of total GDP.Source: IMF, International Financial Statistics, authors’ computations.

Figure 5.

Evolution of the volatility of capital flows in AEs.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: Measures are expressed as % share of total GDP.Source: IMF, International Financial Statistics, authors’ computations.

Figure 6.

Evolution of the volatility of capital flows in EMDEs.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: Measures are expressed as % share of total GDP.Source: IMF, International Financial Statistics, authors’ computations.

Comparing the levels of volatility across income groups, the charts suggest that net flows to EMDEs are more volatile than net flows to AEs. As discussed in the previous section, gross outflows tend to dampen the impact of gross inflows on net flows in AEs but not in EMDEs, resulting in higher volatility for net flows in EMDEs compared to AEs. Looking at the volatility of gross inflows and gross outflows in EMDEs separately, Figures 6b and 6c show that the volatility of inflows is on average higher than outflows, thereby driving up the volatility of net flows.

An interesting question is how the volatility has fared since the GFC, especially in EMDEs. The charts show that, barring several spikes in AEs and EMDEs, the biggest one corresponding to the GFC, the current level of volatility is comparable to the pre-crisis average. However, EMDEs volatility has been prone to bouts, as it has risen during global risk off episodes. In particular, the volatility spiked up during the taper tantrum episode. These bouts of increase in volatility thus remain a policy challenge for EMDEs, particularly in the current context of capital flow slowdown (IMF 2016), where even small swings in flows can lead to substantial net outflows or sudden stops. In fact, when adjusted for the level of flows, volatility has either increased or stayed at the same pre-crisis levels (Appendix D) and, hence, remains a major concern.

Given its size, China could have an impact in the overall capital flow volatility of EMDEs. Charts in Appendix E display the volatility of capital flows over time for World and EMDEs, excluding China. Broadly speaking, the evolution is not very different when China is excluded. However, the increases in volatility observed more recently are not as pronounced. The subsequent charts in Appendix E show that this discrepancy is associated with the volatility of other investments.

Statistical tests

We perform some tests to detect whether the mean value of the estimates has significantly changed after the GFC compared to the pre-crisis period. Table 2 reports the results for the Geweke’s separated partial means test performed over three groups (global, AEs and EMDEs) and the three different measures.¹³

Table 2 shows that the null hypothesis of equal mean values before and after the crisis is almost always rejected at 5% or 10% significance levels, indicating that there has been some shift in volatility, albeit very small in magnitude, since the GFC. There are, however, some results that show no statistical difference in mean values, notably for net flows in World (RW measure) and net flows in AEs (GARCH estimator). The same holds for the GARCH volatility of both gross and net inflows in EMDEs.

Table 2:

results of the Geweke’s separated partial means test for World, AEs and EMDEs aggregates, RW, GARCH and ARIMA volatility estimates.

Notes: GIs: Gross Inflows; GOs: Gross Outflows; NIs: Net Inflows. Statistics are based on two separate means (p=2) and covariance functions tapered at 2%; the limiting distribution is a chi-squared with 1 dof; H0: means are equal, H1: means are different.

Table 2:

results of the Geweke’s separated partial means test for World, AEs and EMDEs aggregates, RW, GARCH and ARIMA volatility estimates.

RW				GARCH			ARIMA
	GIs	GOs	NIs	GIs	GOs	NIs	GIs	GOs	NIs
Global
Statistic	9.18	6.83	1.19	62.44	53.39	3.66	20.36	15.27	3.05
P-value	0.00	0.01	0.28	0.00	0.00	0.06	0.00	0.00	0.08
AEs
Statistic	20.99	21.04	15.17	79.31	86.76	0.68	17.69	17.61	6.88
P-value	0.00	0.00	0.00	0.00	0.00	0.41	0.00	0.00	0.01
EMEs
Statistic	7.89	15.84	24.60	0.99	12.38	0.06	6.51	15.10	27.76
P-value	0.00	0.00	0.00	0.32	0.00	0.80	0.01	0.00	0.00

Table 2:

results of the Geweke’s separated partial means test for World, AEs and EMDEs aggregates, RW, GARCH and ARIMA volatility estimates.

RW				GARCH			ARIMA
	GIs	GOs	NIs	GIs	GOs	NIs	GIs	GOs	NIs
Global
Statistic	9.18	6.83	1.19	62.44	53.39	3.66	20.36	15.27	3.05
P-value	0.00	0.01	0.28	0.00	0.00	0.06	0.00	0.00	0.08
AEs
Statistic	20.99	21.04	15.17	79.31	86.76	0.68	17.69	17.61	6.88
P-value	0.00	0.00	0.00	0.00	0.00	0.41	0.00	0.00	0.01
EMEs
Statistic	7.89	15.84	24.60	0.99	12.38	0.06	6.51	15.10	27.76
P-value	0.00	0.00	0.00	0.32	0.00	0.80	0.01	0.00	0.00

Therefore, conclusions about evidence of a decrease or increase in volatility owing to the GFC depend on the type of flows targeted by the analysis and the methodology adopted. Broadly speaking, we can say that, in the aftermath of the financial crisis, EMDEs have experienced a slight increase in the volatility of gross inflows and a slight decrease in the volatility of gross outflows, on average by 0.2% and 0.16% of GDP respectively. These two effects have concurred to slightly decrease the volatility of net inflows by an average of 0.02% of GDP.

Individual components

Even if EMDEs are moving towards a greater financial integration, they still attract inflows from foreign investors that more than outweigh investments of domestic agents abroad.¹⁴ Therefore, the paper from here focuses primarily on the analysis of the volatility and its determinants for gross (or net) inflows, as evidence suggests these are of greater importance for EMDEs.¹⁵ Figure 7 shows the evolution of volatility for individual components of EMDEs gross inflows. Portfolio and other investment flows are around two and four times, respectively, more volatile than FDI. Among portfolio flows, portfolio debt is more volatile than portfolio equity. These results are in line with literature showing that equity flows, both FDI and portfolio equity, are generally less volatile.

Figure 7.

Evolution of the volatility of gross capital inflows for EMDEs by components.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Note: measures are expressed as % of group GDP.Source: IMF, International Financial Statistics, authors’ computations.

Looking at the evolution of each instrument, the volatility estimates for FDI present a clear-cut upward trend before the GFC and a downward trend afterwards, particularly for GARCH and ARIMA estimates. For portfolio, GARCH and ARIMA estimates are higher than RW. Broadly speaking, there is not a defined trend and estimates tend to hover around the long-run averages. However, the volatility for this instrument is lower in the subsample 2009-2016. As to portfolio equity, the ARIMA and RW estimates highlight a slight upward trend for volatility until the GFC and a downward trend thereafter. Finally, other investments display the highest volatility estimates among all instruments. Even in this case, GARCH produces higher values than the other approaches.

The volatility of individual components can help track down the dynamics underlying the evolution of volatility of total capital flows. Figure 8 plots the ARIMA estimates of total inflows and the individual components. Positive changes in volatility are, broadly speaking, driven primarily by increases in the volatility of other investments, followed by the volatility of portfolio instruments (debt primarily). In the next section, we analyze in detail the volatility of the subcomponents of other investments.

Figure 8.

Evolution of the ARIMA volatility estimates of aggregate gross capital inflows and their components in EMDEs.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: measures are expressed as % of group GDP.Source: IMF, International Financial Statistics, authors’ computations.

Even if, in certain cases, volatility seems to have declined over time, the results can be different when estimates are adjusted by the levels (the idea is almost similar to computing the coefficient of variation). In particular, though net FDI volatility has gone down significantly for EMDEs, it is rising when adjusted for the levels (Appendix D, Figures 1 and 2). Similarly, net portfolio adjusted volatility has also increased in more recent periods.

Statistical tests

Table 3 reports the partial means test results for gross inflows in EMDEs, broken down by instruments. While the null hypothesis of equal means is rejected for FDI, portfolio debt and other investments across all the estimators, the volatility of portfolio inflows (GARCH) and the volatility of portfolio equity flows (RW) have means that are not statistically different before and after the GFC. Using ARIMA estimates, we can conclude that, in the aftermath of the financial crisis, EMDEs have experienced a change in the volatility of FDI, portfolio and other investments, on average, by 0.086%, -0.014% and 0.249% of GDP respectively.

Table 3.

results of the Geweke’s separated partial means test for EMDEs gross inflows, RW, GARCH and ARIMA volatility estimates.

Notes: Statistics are based on two separate means (p=2) and covariance functions tapered at 2%; the limiting distribution is a chi-squared with 1 dof; H0: means are equal, H1: means are different.

Table 3.

results of the Geweke’s separated partial means test for EMDEs gross inflows, RW, GARCH and ARIMA volatility estimates.

	FDI	Portfolio	Port. debt	Port. equity	Other
RW
Statistic	153.78	4.18	17.06	1.90	10.76
P-value	0.00	0.04	0.00	0.17	0.00
GARCH
Statistic	136.40	2.60	5.99	2.99	5.06
P-value	0.00	0.11	0.01	0.08	0.02
ARIMA
Statistic	233.30	2.86	4.52	3.21	13.81
P-value	0.00	0.09	0.03	0.07	0.00

Notes: Statistics are based on two separate means (p=2) and covariance functions tapered at 2%; the limiting distribution is a chi-squared with 1 dof; H0: means are equal, H1: means are different.

Table 3.

results of the Geweke’s separated partial means test for EMDEs gross inflows, RW, GARCH and ARIMA volatility estimates.

	FDI	Portfolio	Port. debt	Port. equity	Other
RW
Statistic	153.78	4.18	17.06	1.90	10.76
P-value	0.00	0.04	0.00	0.17	0.00
GARCH
Statistic	136.40	2.60	5.99	2.99	5.06
P-value	0.00	0.11	0.01	0.08	0.02
ARIMA
Statistic	233.30	2.86	4.52	3.21	13.81
P-value	0.00	0.09	0.03	0.07	0.00

Notes: Statistics are based on two separate means (p=2) and covariance functions tapered at 2%; the limiting distribution is a chi-squared with 1 dof; H0: means are equal, H1: means are different.

Delving Deeper into Other Investment Flows

Even though other investments are, on average, smaller in levels than equity instruments like FDI, the previous section shows that they are very relevant in determining the volatility developments for capital inflows to EMDEs. Hence, this section looks in detail at the individual components of other investment flows. Notably, other investment flows can be classified in: i) bank flows; ii) private non-bank flows; iii) official sector flows.

Composition of Other Investment Flows

Before discussing the volatility measures, Figure 9 depicts the evolution of other gross inflows and shows their composition (in levels), by disentangling between flows to the public sector and flows to the private sector. Private flows - which make up most part of the aggregate other investment flows - reversed and experienced a temporary stop during the GFC, which resulted in the sudden stop of aggregate gross inflows to EMDEs over the same period. After the GFC, the level of flows to private sector appears markedly lower than before, with two episodes of reversal experienced in 2015 and 2016. This behaviour broadly coincides with the general slowdown in aggregate capital flows to EMDEs after the crisis.

Figure 9.

Other gross capital inflows in EMDEs by sector.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: measures are expressed as % share of group GDP; private sector flows are the sum of banking flows and flows to non-banking companies.Source: IMF, International Financial Statistics, authors’ computations.

Figure 10 shows the evolution of other gross inflows by decomposing the flows to private sector into flows to banks and non-banking companies. Flows to countries for which data are not classified between banking and non-banking sectors are not included in these aggregates. The “private – not classified” flows include flows to countries for which data are not classified between banking and non-banking flows (see Figure 10). Before the crisis, most of other flows consisted of private flows to both banking and non-banking sectors, which contributed to the surge of gross capital inflows to EMDEs over the same period. On the other hand, after the GFC, both these types experienced a decrease in volumes and, in certain periods, they almost disappeared from the capital account (e.g. in the second half of 2009).

Figure 10.

Other gross inflows in EMDEs by sector.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: measures are expressed as % share of group GDP.Source: IMF, International Financial Statistics, authors’ computations.

Volatility of Other Investment Flows

Figure 11 below shows the evolution of volatility for the other gross inflows¹⁶ broken down by their subcomponents. It can be noticed that the volatility of private flows is much higher than the volatility of flows through the official sector. The latter, indeed, after a spike during the crisis, has stabilized around a level which is close to 0.

Figure 11.

Volatility estimates for other gross capital inflows in EMDEs by components.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: measures are expressed as % share of group GDP; aggregates for flows to the banking and non-banking sector do not include non-reporting countries.Source: IMF, International Financial Statistics, authors’ computations.

In the private sector, flows to banks display a slight decrease towards the end of the time period considered, while the volatility of flows to the non-banking sector has stabilized around a lower long-run average after the crisis. All in all, we can say that the volatility of other investments is almost completely driven by the volatility of private flows and, for these latter ones, volatility is around equally distributed across the banking and non-banking sectors.

C. Volatility Measures for Individual countries

Given that the previous section shows how the three approaches produce very similar estimates across different types of flows and instruments, the analysis carried out in the remainder of the paper will be based on our preferred ARIMA estimates only.

Apart from computing volatility of aggregate series, the same three measures are used to calculate the volatility for the individual EMDE countries. The volatility measures of individual countries are useful for several reasons. First, they can be important tools in policy discussions related to understanding and mitigating undesirably high volatility in a particular country. Second, the volatility measures of single components will inform policymakers about which instruments they should monitor more closely. Third, these measures can be extremely useful in studies and research geared towards understanding and comparing volatility across countries and regions, as they can shed light on the instruments responsible for possible shifts in volatility. Fourth, as shown in our subsequent analysis, the estimates of individual countries can help understand the relationship between EMDEs and AEs capital flow volatility, as well as the determinants of EMDEs capital flow volatility.

Figure 12 shows the median values of volatility estimates for each country by instruments for gross inflows.¹⁷ The green bar represents the median value for each group. As expected, median volatility is higher for portfolio flows than other types, in line with the findings witnessed for aggregate series. The median variance of overall portfolio flows is lower than the sum of variances of portfolio debt and portfolio equity flows, suggesting that there might be a negative correlation between these two subcategories. These charts also highlight the presence of a great degree of heterogeneity across countries, that is concealed when looking at the trend of aggregate flows. For instance, while the conclusion that, on median, portfolio flows are much more volatile than FDIs, however there are some countries (e.g., Chile) whose FDIs are much more volatile than portfolio inflows, or for which, at least, the volatility measures are of comparable magnitude across the two instruments (e.g., Latvia and Kazakhstan).

Figure 12.

Medians of ARIMA volatility estimates for gross inflows by instrument and across countries.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001

Notes: measures are expressed as % of GDP.Source: IMF, International Financial Statistics, authors’ computations.

In addition, Figure 13 displays the empirical distributions of the median volatilities for each instrument. The density functions of the median volatilities are a mixture of multiple distributions and this suggests the presence of important outliers in the series, something which is also underlined by the positive skewness and the high kurtosis of the density functions.¹⁸ In particular, the distributions of median volatilities for portfolio and portfolio debt are shifted to the right, compared to the other ones, indicating how median volatilities are generally higher for these types of instruments. Moreover, the kurtosis of the same distributions is the highest among all the instruments (closely followed by other investments). This indicates that the probability of observing extreme values (outliers) is greater in these cases than for FDI or portfolio equity, making the aggregate results more prone to be influenced by extreme observations.

Figure 13.

Cross-sectional Empirical Probability Density Functions of median ARIMA volatility estimates for gross inflows by instrument.

Citation: IMF Working Papers 2017, 041; 10.5089/9781475585254.001.A001