
 Start

4875
 Prefix

This paper was prepared while Swarnali Ahmed Hannan was an economist and Maria
Sole Pagliari was a summer intern in the Strategy, Policy and Review Department at the IMF. All remaining errors
are ours.
4 In the last decade, however, EMDEs have started to increasingly invest abroad as well (see
 Exact

Obstfeld (2012b)).
 Suffix

5 For instance, Broner and Ventura (2016) point out that financial globalization has sometimes led to capital flows
that are volatile and procyclical.
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.
 (check this in PDF content)

 Start

4908
 Prefix

This paper was prepared while Swarnali Ahmed Hannan was an economist and Maria
Sole Pagliari was a summer intern in the Strategy, Policy and Review Department at the IMF. All remaining errors
are ours.
4 In the last decade, however, EMDEs have started to increasingly invest abroad as well (see Obstfeld (2012b)). 5 For instance,
 Exact

Broner and Ventura (2016)
 Suffix

point out that financial globalization has sometimes led to capital flows
that are volatile and procyclical.
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.
 (check this in PDF content)

 Start

5113
 Prefix

All remaining errors
are ours.
4 In the last decade, however, EMDEs have started to increasingly invest abroad as well (see Obstfeld (2012b)). 5 For instance, Broner and Ventura (2016) point out that financial globalization has sometimes led to capital flows
that are volatile and procyclical.
flows and economic growth (see Easterly, Islam, and Stiglitz 2001 and
 Exact

Ramey and Ramey 1995),
 Suffix

maintaining a steady stream of capital flows is a policy priority in most EMDEs.
In this regard, the composition of the financial account is another crucial aspect to consider6. 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 shorterterm instruments like other investment
flows, an
 (check this in PDF content)

 Start

6158
 Prefix

The existing empirical literature on the volatility of capital flows can be classified into two strands.7
The first strand focuses on analyzing the difference in volatility between the capital flows to
emerging and advanced economies. For instance,
 Exact

Rigobon and Broner (2005)
 Suffix

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.
 (check this in PDF content)

 Start

6887
 Prefix

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.
 Exact

Bekaert and Harvey (1997), LagoardeSegot (2009) and
 Suffix

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.8 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 19802006.
 (check this in PDF content)

 Start

7632
 Prefix

Their main
finding is that, since 2000, global factors have become increasingly significant relative to countryspecific 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. In
another study,
 Exact

Byrne and Fiess (2016),
 Suffix

using data until 2008Q4, show how the volatility of capital
inflows to emerging markets is mainly driven by the US longterm interest rates and the real
commodity prices. Despite being a focus of policymakers, capital flow volatility is usually
quantified in policy and academic discussions using a crude measure of standard deviation of flows
 (check this in PDF content)

 Start

9082
 Prefix

Definitions are
taken from the sixth edition of the Balance of Payments and International Investment Position Manual (BPM6).
7 Previous theoretical contributions include Bacchetta and Van Wincoop (1998), Aghion et al. (2004) and
 Exact

Martin and Rey (2006).
 Suffix

8 See Chuang et al. (2009) and Jinjarak et al. (2011) for an analysis of the relationship between stock returns and
volumes.
over a specified period. Given the importance of this issue, a better understanding of volatility in
capital flows is warranted.
 (check this in PDF content)

 Start

19378
 Prefix

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
 Exact

(IMF 2016a,b),
 Suffix

where even small swings in flows can lead to substantial net outflows or sudden stops14.
12 In this section, the aggregates represent total capital flows divided by total GDP.
 (check this in PDF content)

 Start

21772
 Prefix

Nonetheless, as
already highlighted in the previous section, there exist some fundamental divergences in the
dynamics of individual components (or instruments) of the financial account. It is therefore
necessary to extend the analysis to single instruments as well.
15 Refer to
 Exact

Geweke (2005).
 Suffix

We prefer this test as it allows us to control for autocorrelation of the volatility
estimates.
16 With the notable exception of GARCH estimates for gross outflows.
RW GARCH ARIMA
GIs GOs NIs GIs GOs NIs GIs GOs NIs
EMDEs
Statistic 0.27 0.62 0.01 0.64 0.01 1.46 0.27 5.85 0.09
Pvalue 0.60 0.43 0.92 0.42 0.94 0.23 0.60 0.02 0.77
Table 1: statistics and pvalues of the Ge
 (check this in PDF content)

 Start

24778
 Prefix

than before19.
17 The differences in the behavior of net flows between AEs and EMDEs is also due to the fact that, while in AEs
capital flows are used for both risksharing and portfolio diversification, in EMDEs, by contrast, capital flows are not
only instrumental for risksharing, but also to have access to greater external financing (see
 Exact

Canuto and Ghosh (2013),
 Suffix

Ch. 3).
18 Measures for gross outflows and net inflows, however, are available upon request. Broadly speaking, the volatility
measures for gross inflows and net inflows are similar, but those of gross outflows are different.
19 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 very similar
 (check this in PDF content)

 Start

34308
 Prefix

The cases of Latin America and Europe, however, stand out, as
the volatility of debt inflows decreased in 20072009 and bounced back to unprecedented levels
in the first quarters of 2010.
23 See for instance
 Exact

Lane (2013), Lane and MilesiFerretti (2017).
 Suffix

Figure 5: ARIMA volatility estimates for gross capital inflows in EMDEs by regions and
instruments. Notes: AS=Asia, EU=Europe, LATAM=Latin America, MEAF=Middle East and
Africa; measures are computed as average across countries and expressed as % share of GDP.
 (check this in PDF content)

 Start

37301
 Prefix

of equation (1), one gets:
!"#.=AT.UVT#+T/%+2ATAEX)N(VT#,VE#)EYT+T/% (2)
where Z is the number of elements in vector J# and VT# represents the qth element of the same
vector.
Using a firstorder Taylor expansion, equation (2) can be approximated as:
!"#.=USOA′J#≈ A′\(J#)^U(J#)A (3)
24 See, for instance,
 Exact

Ahmed and Zlate (2014).
 Suffix

25 Notably, in equation (28), Devereux and Sutherland (2009) show how homecountry holdings of foreign equities
and bonds depend on a set of domestic and foreign indicators. As we mainly focus on gross inflows (so changes in
foreign holdings of domestic assets), we can flip the interpretation of this relationship so to derive a direct link
between foreigncountry holdings of domestic equities a
 (check this in PDF content)

 Start

39574
 Prefix

Therefore, given the N=33 EMDEs in our sample, the baseline dynamic panel regression model
is:
!"#.=d%!"#0%.+eTf"T#+T/%+5"#, g=1,...,_,
where f"T# are growth rates of the regressors and 5"#=i"+j#+k"#, so that we control for the
presence of both country and time fixedeffects.
26 See, for instance, Broto et al. (2011) and
 Exact

Rey (2013,2016).
 Suffix

27In our specification, we don’t include proxies for convariances among the regressors. This does not generate
endogeneity issues, as explained below.
28 Since in equation (3) the variance is linearized around the steady state, we demean the regressors, using the trend
estimated via HP filtering as a proxy for the steady state.
 (check this in PDF content)

 Start

40716
 Prefix

However, there is a strand of the literature arguing that longrun interest
rates may be at least as important for capital flows as shortrun rates, if investors prefer to diversify assets with short
maturities and assets of different maturity are imperfect substitutes (see
 Exact

Bernanke et al. (2011)).
 Suffix

As a robustness check,
we ran regressions using longterm interest rates, but the key results were substantially unchanged.
30 In this case, we follow Ghosh et al. (2014) and Ghosh, Ostry and Qureshi (2016) and we use changes in the S&P
500 realized returns volatility as an index for global risk aversion.
 (check this in PDF content)

 Start

41564
 Prefix

For a discussion about the relationship between a
country’s Net International Investment Position (NIIP) and its current account balance, as well as the risk associated
to the buildup of persistent current account imbalances, refer to
 Exact

Obstfeld (2012b) and Lane (2013).
 Suffix

Following Broto et al. (2011), we correct the covariance matrix as suggested by Driscoll and Kraay
(1998), to account for both serial and spatial (crosssectional) correlations.32 In addition, we run
two sets of diagnostic checks to detect whether this framework is capable of addressing the issues
of endogeneity in regressors and crosssectional dependence in the residuals 33.
 (check this in PDF content)

 Start

41677
 Prefix

For a discussion about the relationship between a
country’s Net International Investment Position (NIIP) and its current account balance, as well as the risk associated
to the buildup of persistent current account imbalances, refer to Obstfeld (2012b) and Lane (2013). Following Broto et al. (2011), we correct the covariance matrix as suggested by
 Exact

Driscoll and Kraay (1998), to
 Suffix

account for both serial and spatial (crosssectional) correlations.32 In addition, we run
two sets of diagnostic checks to detect whether this framework is capable of addressing the issues
of endogeneity in regressors and crosssectional dependence in the residuals 33.
 (check this in PDF content)

 Start

44686
 Prefix

Indeed, the results from the next section suggest
that the baseline regressions may be missing some dynamics (both across countries and over time)
that need to be further explored.
32 To do so, we used the  command in Stata (see
 Exact

Hoechle (2007)).
 Suffix

33 In the former case, we use the DurbinWuHausman test, while in the latter we perform the test for weak crosssectional dependence proposed by Pesaran (2015).
Indicators of the US economic stance (US real GDP and US CPI) generally have a negative effect
over the volatility of EMDEs capital inflows, except for FDI and portfolio equity inflows.
 (check this in PDF content)

 Start

44848
 Prefix

the baseline regressions may be missing some dynamics (both across countries and over time)
that need to be further explored.
32 To do so, we used the  command in Stata (see Hoechle (2007)). 33 In the former case, we use the DurbinWuHausman test, while in the latter we perform the test for weak crosssectional dependence proposed by
 Exact

Pesaran (2015).
 Suffix

Indicators of the US economic stance (US real GDP and US CPI) generally have a negative effect
over the volatility of EMDEs capital inflows, except for FDI and portfolio equity inflows. All in
all, a 1% increase either in the US real GDP or in the US CPI decreases the volatility of gross
capital inflows by 6.31% and 2.17% of GDP respectively.
 (check this in PDF content)

 Start

46985
 Prefix

Domestic Factors
Tables 3 and 4 also show that domestic factors are not relevant drivers of capital flows volatility
in EMDEs, thus implying a predominance of global factors influence. While this finding is in line
with evidence provided by some of the existing literature (see, for instance, Broto et al (2011) and
 Exact

Byrne and Fiess (2016)),
 Suffix

yet it has also been pointed out how these effects over capital flow levels
are highly heterogenous across countries, due to differences in the quality of domestic institutions,
country risk and the strength of domestic macroeconomic fundamentals (Fratzscher (2012), IMF
(2016a)).
 (check this in PDF content)

 Start

47241
 Prefix

finding is in line
with evidence provided by some of the existing literature (see, for instance, Broto et al (2011) and
Byrne and Fiess (2016)), yet it has also been pointed out how these effects over capital flow levels
are highly heterogenous across countries, due to differences in the quality of domestic institutions,
country risk and the strength of domestic macroeconomic fundamentals
 Exact

(Fratzscher (2012), IMF (2016a)).
 Suffix

Moreover, it has also been shown that these effects are characterized by a certain degree
of variation over time, since the GFC has actually strengthen the role played by global factors in
driving the level of capital flows.
 (check this in PDF content)

 Start

53549
 Prefix

. ***
p<0.01, ** p<0.05, * p<0.1.Some countries have been excluded from the pool due to lack of an adequate number of observations
Extended regression: the role of commodity prices
In this session, we expand the baseline regression to account for commodity prices, as they are
deemed influential drivers for capital flows towards both exporting and importing economies
 Exact

(Byrne and Fiess (2016)).
 Suffix

Notably, a positive shock to commodity prices increases bank
investments to commodity exporting countries, whose financial institutions, in turn, reinvest their
increased revenues into the international financial markets, to diversify their portfolios.35
Nonetheless, changes in commodity prices are also positively correlated with capital inflows to
commodity importers.
 (check this in PDF content)

 Start

55665
 Prefix

In
particular, and more in line with expectations, an increase in S&P 500 realized returns volatility
exerts a positive effect on the volatility of almost all the instruments, with estimates ranging from
0.48% for total private investments to 5.9% for FDI (with an overall effect of about 5% of GDP).
35 See Arezki, Mazarei, and Prasad (2015) and
 Exact

Arezki, Obstfeld and MilesiFerretti (2016).
 Suffix

36 See Ahmed, Curcuru and Warnock (2015).
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
VARIABLES Total FDI Portfolio Portfolio Debt Portfolio Equity Other Banks Official Nonbanks Total private
σ2(t1) 0.825*** 0.658*** 0.306*** 0.317*** 0.657*** 0.241*** 0.762*** 0.202*** 0.735*** 0.741***
(0.0499) (0.0675) (0.0197) (0.0178) (0.0705) (0.0258) (0.0545) (0.0216) (0.0411) (0.0625)
US shadow ra
 (check this in PDF content)

 Start

63366
 Prefix

The latter one, on the other hand, can be identified with the taper
tantrum episode, but it does not coincide with any significant change in the effects of US M2.
Finally, commodity price index seems to exert a significant effect on capital inflow volatility only
in the preGFC period. This is a new finding compared to existing literature
 Exact

(Byrne and Fiess (2016)),
 Suffix

as it shows how, in the last 10 years, the influence of commodity prices has been greatly
reduced compared to that of other global factors.
<HDU
86VKDGRZUDWH
<HDU
5HDOLVHGUHWXUQVYRODWLOLW\
<HDU
865HDO*'3
<HDU
86&3,
<HDU
860<HDU
86&XUUHQW$FFRXQW
&R
 (check this in PDF content)

 Start

74531
 Prefix

—the push factors are more important than pull factors—is yet another
reminder that the determinants of the level of capital flows can be very different from the
determinants of the volatility of capital flows. However, we also show how certain domestic
(pull) factors have started to play a relevant role after the GFC. This is even more interesting
against the backdrop of recent studies
 Exact

(IMF 2016)
 Suffix

that show how the recent capital flow
slowdown can be predominantly attributed to the broader emerging markets economic growth
slowdown. While this study is a reminder that the findings from the push vs pull factors
literature do not always hold when it comes to the study of capital flow volatility, it is still
possible to detect some commonalities in the effects of push and pull factors over
 (check this in PDF content)