Download Accounting for Emerging Market Countries` International Reserves

Accounting for Emerging Market Countries’ International Reserves:
Are Pacific Rim Countries Different?
*
Atish R. Ghosh
Jonathan D. Ostry
Charalambos G. Tsangarides
December 6, 2013
Preliminary draft
Do not cite or circulate beyond conference participants
Abstract
Popular perception is that emerging market economies (EMEs), and Asian Pacific Rim
countries—China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam (RIMs)—
in particular, have been rapidly accumulating reserves, perhaps beyond what is justified by
precautionary motives. This paper compares and contrasts the determinants of the demand
for international reserves in the RIM countries with other EMEs over the last three decades,
based on current and capital account risks, mercantilist motives, and other motives. Our
findings suggest shifting motives for holding reserves from insurance against current account
shocks, insurance against capital account shocks, and as by-product of possible mercantilism.
We also find some differences between country groups: RIM countries tend to hold more
reserves against current account vulnerabilities and fewer reserves against capital account
vulnerabilities, but more reserves overall. There is also greater evidence of mercantilist
motives being at play for RIM countries, though the contribution of this motive in accounting
for the rise in their reserve holdings is relatively small, peaked in 2004, and has diminished
thereafter.
JEL classification: E58, F15, F31, and F43
Keywords: International Reserves, Precautionary Demand, Mercantilism, Quantile
Regression
_____________________
*
Paper prepared for City University of Hong Kong International Conference on Pacific Rim Countries and the
Evolution of the International Financial Architecture, Hong Kong SAR, December 19-20, 2013. The views
expressed in this paper are those of the authors and should not be attributed to the IMF, its Executive Board, or
its management. Ghosh: Research Department, IMF, Suite 9-612G, 700 19th Street, N.W., Washington D.C.
20431 (email: [email protected]); Ostry: Research Department, IMF, Suite 10-700F, 700 19th Street, N.W.,
Washington D.C. 20431 (email: [email protected]); Tsangarides: Research Department, IMF, Suite 9-612B, 700
19th Street, N.W., Washington D.C. 20431 (email: [email protected]).
2
I. INTRODUCTION
Over the past few decades, despite somewhat greater exchange rate flexibility, and some
draw down during the global financial crisis, emerging market economies (EMEs) have been
accumulating large stocks of international reserves. Reserve holdings, which averaged about
5 percent of GDP in the 1980s, have been doubling every decade since, reaching some 25
percent of GDP by 2010. While foreign exchange reserves may provide useful insurance in
the face of current or capital account shocks, there is often a perception that reserves are not
being accumulated for precautionary purposes but rather as the by-product of deliberate
mercantilism. Beyond the possibility of unfair trade practices and exchange rate manipulation,
such mercantilism may perpetuate global imbalances and ultimately undermine the stability
of the international monetary system (Ghosh et al., 2010).
Asian countries, notably those on the Pacific Rim (RIM), are often singled out for
purportedly pursing export-led growth strategies by keeping their currencies undervalued, in
turn resulting in excessive reserve accumulation. For instance, writing in 2010, Paul
Krugman claimed Today [2010], China is adding more than $30 billion a month to its $2.4
trillion hoard of reserves. This is the most distortionary exchange rate policy any major
nation has ever followed.1 And as recently as 2013, Dani Rodrik has noted that: Although
China phased out many of its explicit export subsidies… mercantilism’s support system
remains largely in place. In particular, the government has managed the exchange rate to
maintain manufacturers’ profitability, resulting in a sizable trade surplus.
But does this view have any merit? In this paper, we take up that question with specific
reference to the Asian Pacific Rim countries— China, Indonesia, Korea, Malaysia,
Philippines, Thailand, and Vietnam.
There are potentially two approaches to addressing this question. The first is to try to
establish an absolute norm for how much reserves are required for precautionary purposes,
with any further accumulation that is undertaken in the context of an undervalued currency
being deemed “mercantilism”. The main difficulty with this approach is that “how much is
enough” is an evolving standard: what was sufficient for current account shocks in the 1980s
was clearly inadequate for the EME capital account crises of the 1990s. Likewise, prior to the
global financial crisis, many commentators thought that Russia’s US$600 billion of reserves
were ample, and perhaps even excessive, yet that view was rapidly revised when the central
bank spent more than one-third of its reserve stock in the space of a couple of months.
Moreover, existing reserve adequacy metrics generally have wide margins—the recently
1
Paul Krugman (14 March 2010). "Taking on China and its currency"
http://www.nytimes.com/2010/03/15/opinion/15krugman.html. Dani Rodrik 2013, The New Mercantilist
challenge, http://www.project-syndicate.org/commentary/the-return-of-mercantilism-by-dani-rodrik
3
developed IMF methodology for assessing reserve adequacy, for instance, suggests a range
between 100 and 150 percent of its Reserve Adequacy Metric as being appropriate, but for
the typical EME, that range translates into 10 percent of GDP.
In this paper, therefore, we adopt a more pragmatic tack. Rather than try to assess whether
Asian RIM countries are stockpiling excessive reserves against some absolute standard, we
compare the behavior of RIM countries to that of other emerging market economies. This
allows for evolving notions of reserve adequacy while still assessing whether RIM countries
are exceptional in their reserves accumulation behavior. Specifically, we seek to determine
whether RIM countries hold more reserves than non-RIM countries, controlling for various
current and capital account vulnerabilities; whether they react differently in terms of their
reserve holdings to such vulnerabilities; and whether a larger proportion of their reserve
accumulation over the past twenty years can be accounted for by mercantilist motives.
Building on our earlier work (Ghosh, Ostry, and Tsangarides (2012), henceforth GOT), we
define mercantilism as reserve accumulation that is unrelated to current or capital account
vulnerabilities and that takes place in the context of an undervalued currency. Our empirical
strategy proceeds in three steps. We begin by estimating a reserve demand function for a
large sample of (RIM and non-RIM) EMEs over the period 1980-2010 relating reserve
holdings (expressed in percent of GDP) to precautionary motives (current and capital account
vulnerabilities) and mercantilism. Our estimates allow for the possibility that the motives for
reserve accumulation differ over time, or according to the level of reserve holdings. Our first
test is simply whether RIM countries hold significantly more reserves than non-RIM
countries, controlling for their current and capital account vulnerabilities. Next,
we allow for the possibility that RIM countries’ reserve holdings react systematically
differently to the various vulnerabilities (as well as to currency undervaluation and the other
determinants) by introducing RIM interaction terms. Finally, we decompose countries’
reserve accumulation since 1990 into the various motives to examine their relative
importance in the RIM and non-RIM samples.
Our findings may be summarized briefly. First, for the EME sample as whole, the analysis
suggest shifting motives for reserve accumulation—in the 1980s (and at low levels of
reserves), it was mostly insurance against current account shocks; in the 1990s (especially
post-Asian crisis), insurance against capital account shocks became more important; in the
2000s (and at higher levels of reserves), there is greater evidence of mercantilism. Second,
there is some evidence that RIM countries hold more reserves than would be expected on the
basis of their current and capital account vulnerabilities. Allowing for different responses
across country groups suggests that, while RIM countries hold more reserves in general, they
hold fewer reserves against capital account shocks—the exception being the period
immediately following the Asian crisis—but hold more reserves against current account
vulnerabilities. Third, decomposing the factors behind the build-up of EME reserves since
the 1ate 1990s suggests that (the by-product of) mercantilism has played a more important
4
role in the RIM sample than in the non-RIM sample.
Overall, we conclude that there is some merit to the view that reserve accumulation by RIM
countries has been different from other EMEs, including because mercantilist motives have
been more important. Yet it is important not to exaggerate the importance of mercantilist
motives in accounting for the higher reserve accumulation. Our estimates imply that, at its
peak (in 2004), mercantilist motives accounted for an average of 3½ percent of GDP of the
stock of reserves accumulated by RIM countries (in other words, absent such mercantilist
motives, RIM countries’ average reserve holdings in 2004 would have been 3½ percent of
GDP lower). Moreover, this contribution to the stock of reserves subsequently shrank, to
around 2.3 percent of GDP by 2007, remaining roughly constant thereafter. Hence, we find
little evidence that RIM countries on average continue to accumulate reserves due to
mercantilist motives.
The rest of the paper is organized as follows. Section II reviews the literature and lays out our
empirical strategy. Section III presents our main OLS results. Section IV turns to quantile
regressions. Section V concludes.
II. EMPIRICAL STRATEGY
A. Determinants of demand for reserves
Traditional explanations—and policy prescriptions—for holding foreign exchange reserves
centered on the “insurance” that reserves provide. The original “three months of imports”
rule, for instance, was designed to help insulate developing countries against current account
shocks: shortfalls in export earnings or domestic shocks—such as natural disasters—that
might necessitate exceptionally large imports. As developing and emerging market countries
became more financially integrated, insulation against capital account shocks gained
importance—as amply demonstrated by the emerging market crises of the 1990s. By
providing confidence to investors, reserves can reduce the likelihood of a sudden stop and
rush for the exit or—if they nevertheless materialize—cushion the economy from their
consequences (Ben-Bassat and Gottlieb (1992)). Rather than three months of imports,
insurance against capital account shocks calls for reserves to be held against short-term debt
(or other skittish foreign liabilities) and possibly against M2, in case domestic investors lose
confidence in the banking system and the currency—as happened during the Mexican and
Indonesian crises—and rush for the exit as well. While in principle reserve demand should be
a function of the exchange rate regime, in practice few EMEs can be truly indifferent to the
5
level of the exchange rate, and most studies find few differences in reserve holdings
according to the country’s exchange rate regime.2
Recent cost-benefit calibration models cast reserve accumulation as an explicit optimization
problem, with reserves chosen to provide the optimal insurance against a sudden drop in
consumption given risk aversion and the costs of holding reserves (see Caballero and
Panageas (2008), Jeanne (2007), and Jeanne and Rancière (2006)). Empirical contributions
confirm the importance of precautionary determinants in explaining reserve holdings in
EMEs (see, for example, Bastourre, Carrera and Ibarlucia (2010), Obstfeld, Shambaugh and
Taylor (2010), Cheung and Qian (2009), and de Beaufort Wijnholds and Kapteyn (2001)).
For East Asia in particular, Aizenman and Lee (2007) and Aizenman and Marion (2003)
provide some evidence for precautionary motives in post-Asian crisis reserve accumulation. 3
A quite different explanation for observed reserve holdings is that they are the unintentional
by-product of modern mercantilism. In this view, emerging market economies—perhaps
following the examples of Europe and Japan during the Bretton Woods period (Dooley et al.,
2002)—often seek to maintain deliberately undervalued currencies through foreign exchange
intervention (often supplemented by controls on capital inflows) in order to promote exports
as part of an export-led growth strategy. Ghosh and Kim (2008) consider an economy in
which the government has an incentive to maintain an undervalued exchange rate because it
is equivalent to an export subsidy (the cost of sterilized intervention being the analog of the
subsidy cost) in an economy where there are positive productivity spillovers, external to the
firm, of output in the tradable sector. Aizenman and Lee (2008) show that, in a two-country
game, such mercantilism can lead to the inefficient accumulation of reserves as each country
engages in beggar-thy-neighbor competitive depreciations.
Since it is not possible to observe the intentions of the monetary authorities directly, it is very
difficult to establish whether reserves are being accumulated as a by-product of mercantilist
motives or for some other reason. It is quite possible, for instance, that the central bank is
trying to maintain an undervalued currency not to promote exports or gain some unfair trade
2
If the central bank was truly indifferent to what happened to the exchange rate, it would not need to hold any
reserves. But even EMEs with more flexible exchange rates may want to insure against extreme events and hold
reserves accordingly.
3
Aizenman and Marion (2003) use the volatility of export receipts as their measure of current account shocks,
and implicitly include external debt and broad money in their analysis, but do not account for possible
mercantilism. Aizenman and Lee (2007) and Delatte and Fouquau (2012) try to capture possible mercantilist
motives, but neither includes banking system liabilities, and both are constrained to rather crude PPP-based
measures of undervaluation, which they do not find to be robust determinants of reserves.
6
advantage, but because it needs to accumulate reserves for precautionary purposes. 4 To
address this fundamental identification problem, here we follow GOT to define the
mercantilist motive as reserve accumulation that is uncorrelated with proxies of current or
capital account vulnerabilities, and that takes place in the context of a substantially
undervalued currency. For our definition, we also require that the exchange rate
undervaluation be substantial—an estimated misalignment of at least10 percent.
B. Empirical Specification
Full sample and sub-periods
For our baseline regression, we follow GOT to include various proxies of current and capital
account vulnerabilities as well as currency undervaluation (to capture the mercantilist
motive), augmenting their specification to differentiate between RIM and non-RIM countries:
ln Rit   0  CACAit   KA KAit   M Mercantilistit   R regimeit  O Otherit
  RIM RIM it  CARIM CAit  RIM it   KARIM KAit  RIM it   MRIM Mercantilistit  RIM it (1)
  RRIM regimeit  RIM it  ORIM Otherit  RIM it   it
where
CAit  ln M it , XVolit ,VolExtDemit  ,
KAit  ln BM it , KOpenit , Debtit  ,
regimeit  NeerVolit , Pegit  ,
Mercantilistit  Undervalit  , and
Otherit  OppCostit ,ln GDPpcit ,ln Popit .
-
lnRit is the natural log of reserves to GDP for country i at time t valued expressed as a
ratio to GDP;
-
CAit includes proxies for current account precautionary motives, namely, the log of
imports to GDP, the volatility of exports (measured as three-year standard deviation), and
the volatility of external demand captured by the trading partners’ growth volatility
(scaled by exports). Increases in the volatility of external demand are expected to result in
more precautionary reserve accumulation as insulation from exogenous (partner) demand,
while reserve accumulation is expected to be positively related to imports to GDP and
export volatility;
4
Of course the current account is not the only way that the central bank can accumulate reserves: it can also do
so by borrowing through the capital account.
7
-
KAit includes the log of M2 to GDP; a measure of the de jure openness of the capital
account—the Chinn-Ito index, which is based on the IMF Annual Report on Exchange
Arrangements and Restrictions; and short-term debt to GDP. All financial variables are
expected to be positively related to reserve accumulation;
-
Regimeit includes the volatility of the nominal effective exchange rate, and a dummy
variable for de facto pegged exchange rate regime. Both coefficients are expected to be
negative;
-
Mercantilist is proxy for exchange rate undervaluation, as defined in GOT. Countries
with undervalued exchange rates are expected to hold more reserves (in the target
country’s currency) needed to maintain/reduce the value of their exchange rate; and
-
Other includes a proxy for the opportunity cost of holding reserves (measured as interest
rate differential with the US), and two scaling variables, the log of population, and the log
of real GDP per capita at purchasing power parity. Reserve accumulation is expected to
be positively related to the scale variables, and negatively related to the opportunity cost.
Finally,  is a random error term.5
To examine how the estimates obtained for (1) change when the sample is divided in several
sub periods, namely (i) the 1980s, 1990-1997, 1998-2010; (ii) 1980-1997, and 1998-2010;
and (iii) 1980-1997, 1998-2003, and 2004-2010. The cutoffs are chosen to coincide with the
Asian crisis, and the post 2000 surge in reserves.
Quantile regressions
As discussed above, motives for holding reserves may have shifted over time, particularly as
EMEs have become more financially integrated. Different motives may also apply at a given
point in time, but at different points along the sample distribution of reserve holdings. For
example, countries that hold low levels of reserves may do so because they are not very
financially integrated and are mostly concerned about current account rather than capital
account shocks. Conditional quantile functions may offer a more complete picture of the
effect of the reserve demand determinants covariates on the location, scale and shape of the
distribution of the reserve accumulation. Using quantile regressions, we allow the elasticities
to vary across the various quantiles (q) of reserve accumulation. For a quantile q, equation (2)
is modified as follows:
5
We do not include country fixed effects in equation (1) as that would imply identifying the effect of the
variables solely through their time variation (and it is not very informative economically). However, we cluster
standard errors at the country level.
8
q
q
ln Rit   0q  CA
CAit   KA
KAit   Mq Mercantilistit   Rq regimeit  Oq Otherit
q
q
q
q
  RIM
RIM it  CARIM
CAit  RIM it   KARIM
KAit  RIM it   MRIM
Mercantilistit  RIM it (2)
q
q
  RRIM
regimeit  RIM it  ORIM
Otherit  RIM it  it
For various values of q, we obtain estimated intercepts and slopes and conduct tests to
examine differences across quantiles.
Data
Following GOT, we construct several estimates for misalignments based on application of
three equilibrium exchange rate methodologies, namely, the macro balance (MB), the
equilibrium real effective rate (ERER), and external sustainability (ES). We begin by
constructing misalignment estimates based on each of the three methodologies and then
combine estimates from the three methods to construct the average and median
misalignments. Median misalignments are then translated into three-way classifications for
aligned, overvalued and undervalued exchange rates. Given the uncertainty surrounding
estimates, we consider percentage deviations of [-10, 10] as aligned, less than -10 (more than
+10) undervalued (overvalued). Appendix A in GOT describes the methodologies used and
the construction of each of the misalignment estimates in detail.
The remaining variables are constructed from the IMF’s WEO, IFS, and INS databases. The
exchange rate regime classifications are derived from the IMF’s revised classification
published in the 2009 Annual Report on Exchange Rate Arrangements and Exchange
Restrictions. Using the IMF’s de facto exchange rate regime classification at the end of each
period of the analysis, we classify exchange rate regimes into fixed and non-fixed, as well as
separating the pegs into hard and soft pegs. Tables A1 and A2 in Appendix A provide details
on the sample, the variables used, and summary statistics.
III. THE STORY IN SIMPLE AVERAGES
Popular perception is that EMEs, and RIM countries in particular, have been rapidly
stockpiling reserves, perhaps beyond what is justified by precautionary motives. A first
glance at the data indeed shows not only that EMEs have increased their reserve holdings
markedly (in relation to GDP) over the past three decades, but also that RIM countries have,
on average, held more reserves (in relation to their GDP) than their non-RIM counterparts
(Figure 1). Over the full sample period, 1980-2010, RIM countries’ reserve holdings
averaged 17 percent of GDP compared to 12 percent of GDP for non-RIM countries. By the
eve of the global financial crisis, this difference was even more marked: at end-2007 reserves
amounted to 32 percent of GDP in RIM countries and 18 percent of GDP in the non-RIM
sample. Subsequently, however, while EME reserves have more than recovered from their
use during the global crisis, the difference between RIM and non-RIM countries has
narrowed to 9.7 percent of GDP.
9
Why might RIM countries hold more reserves? As discussed above, traditional explanations
would be that they face larger current or capital account vulnerabilities, and therefore have
greater country insurance needs. Again, a first glance at the data suggests this explanation
might have merit: over the full sample and various sub-periods (1980-97; 98-2004; 2005-10),
RIM countries exhibit greater current and capital account vulnerabilities (Figures 2(a)-2(d)).
For instance, imports average 35 percent of GDP in RIM sample compared to less than 30
percent of GDP in the non-RIM sample, with the difference between them rising to some 15
percent of GDP by 2007. Although RIM countries tend to have lower short-term debt
(especially in the latter part of the sample), their banking system monetary liabilities
(M2/GDP) are almost twice as large as those in non-RIM countries. But the mercantilist story
also gets some (potential) support. RIM countries’ are, on average, 7 percent more
undervalued than non-RIM countries during the whole period of analysis, and about 22
percent more undervalued during the 2000s.6
Looking across the distribution of reserve holdings suggests that the relationships may be
different at various levels of reserves (Figure 3). Median reserve holdings have been
increasing much more rapidly in RIMs compared to non-RIMs—in fact, the bottom 25th
percentile of the RIM countries almost exceeds the top 75th percentile of the non-RIM
countries. Unlike non-RIMs, the dispersion in reserve holdings across RIMs has also risen,
with the difference between the top and bottom quartiles widening from 13 percent of GDP
in 1980-1997 to 20 percent of GDP in 2005-10.
Overall, there is considerable time series and cross sectional variation in reserve holdings
between RIM and non-RIM countries, as well as within the distribution of each sample.
These results suggest that reserve demand determinants may be different between RIM and
non-RIM countries, a point that reinforces the motivation of our paper. In addition, while the
average behavior of RIM versus non-RIM countries is certainly informative, there may be a
richer story taking account of differences between these groups of countries at various points
in the reserves holdings distribution that bears examination.
IV. PERIOD ANALYSIS OF RESERVE DEMAND
We estimate our full model (1) in steps, adding groups of variables (scale, regime,
opportunity cost, current account, capital account, and mercantilist) sequentially. For the
moment, in the estimation we allow RIM and non-RIM countries to differ only in the
intercept term by the inclusion of a dummy variable for RIM countries (Table 1). The
6
These are the estimates of exchange rate misalignment in percent. Given the large margins surrounding these
point estimates, in the empirical work below, we use a dummy variable for (more than) 10 percent
undervaluation.
10
regressions are estimated on centered variables so the exponentiated value of the constant
gives the average reserve holdings for the non-RIM sample.7
Adding explanatory variables in sequence suggests that exchange rate regime, current and
capital account precautionary variables, and mercantilist motives are important in explaining
reserve accumulation in the broader sample of RIM and non-RIM countries. The full model
including all explanatory variables is given in Table 1[7]. The model explains about 56
percent of the variation in reserve holdings (without the inclusion of annual or country fixed
effects), with both current account and capital account precautionary motives as well as
mercantilist motives statistically significant and with the expected signs. Countries with
higher per capita income hold more reserves; although the fixed exchange rate regime
dummy is insignificant, countries with more flexible exchange rate regimes (as measured by
the volatility of the nominal effective exchange rate) do hold fewer reserves. All of the
current account vulnerability variables are significant: a one standard deviation increase in
the imports-to-GDP ratio, the volatility of exports, or the volatility of partner country growth
increases reserve holdings by roughly one-half percent of GDP. Turning to capital account
vulnerabilities, short-term debt is insignificant, but a one-standard deviation increase in broad
money-to-GDP or financial openness, is associated with 0.2 to 1.5 percent of GDP higher
reserves. Other things equal, reserves of a country with an undervalued currency are 2
percent of GDP greater than those of a country whose currency is not undervalued.
Looking at the sub-periods (Table 1 [8-10]), suggests shifting motives for reserve holdings
by EMEs. In the 1980s, it was insurance against current account shocks was the predominant
motivation for holding reserves with capital account motives almost non-existent. Following
the capital account crises of the 1990s, insurance against capital account shocks became more
important. These effects are large: a one standard deviation increase in the short-term debtto-GDP ratio (corresponding to an increase in short-term debt of 5 percent of GDP) is
associated with 0.7 percent of GDP higher reserves. And as of the late 1990s, there is greater
evidence of mercantilism with the undervaluation dummy turning economically and
statistically significant.8
Are RIM Countries Different?
Turning to our main question of interest, the regressions in Table 1 include a dummy variable
for RIM countries. In specifications [1]-[3], which include only the “scale” variables (per
7
For example, a specification with only the scale variables (log(per capita income) and log(population))
included implies that average reserves holdings by non-RIM countries were exp(-2.453)=8.6 percent of GDP,
compared to 20.4 (=exp(-2.453+0.864)) percent of GDP in RIM countries.
8
These results are robust to a variety of specifications, instrumentation for possible endogeneity of the
regressors, and the inclusion of additional control variables; see GOT for a full discussion.
11
capita income and population), the RIM dummy is statistically significant, and indicates that
RIM countries hold some 10 to 12 percent of GDP more reserves than corresponding nonRIM countries. Once the current and capital account vulnerabilities are included in the
regression (Table 1 [4]-[7]), the RIM dummy turns insignificant—suggesting that RIM
countries are no different controlling for their vulnerabilities.
While that may be partly true, it is also possible that multi-colinearity between the various
explanatory variables is masking statistically significant differences. It is noteworthy, for
instance, that population—which is insignificant in specifications [1]-[3]—suddenly turns
positive and statistically significant. To address this concern, Table 2 reports the coefficients
of a similar regression, but where the various proxies of scale, current account vulnerabilities,
and capital account vulnerabilities are replaced by their respective first principal components.
In the full model (Table 2 [7]), the current account, capital account, and undervaluation
proxies are all statistically significant. Moreover, the RIM dummy retains its statistical
significance and implies that controlling for current and capital account vulnerabilities, RIM
countries hold 4.5 percent of GDP higher reserves than non-RIM countries (Table 2 [5]).
Adding our proxy for mercantilism, the difference between RIM and non-RIM countries
shrinks to 3.8 percent of GDP but remains statistically significant. Across sub-samples, the
RIM dummy is significant until 2004, but turns insignificant for the period 2005-10.
The results are similar, but starker, for China. The China dummy is significant even in the
full model with multiple proxies current and capital account vulnerabilities and mercantilism.
The estimates imply that China holds 11 percent of GDP more reserves than would be
expected on the basis of its current and capital account vulnerabilities (Table 2 [11]). The
results for the RIM sample are not being driven by the inclusion of China within the group,
however: excluding China leaves the RIM dummy statistically significant and of almost the
same magnitude.
Overall, therefore, controlling for precautionary motives, there is a statistically significant
difference between the reserve holdings of RIM and non-RIM countries, part—but only
part—of which is explained by greater mercantilist behavior on the part of RIM countries.
Do RIM Countries Respond Differently to Vulnerabilities?
The analysis above imposes the same slope coefficients for RIM countries as non-RIM
countries. One reason for the statistically significant RIM dummy may be that RIM countries
respond differently to the current and capital account vulnerabilities. For example, if RIM
countries were more risk averse, they may wish to hold higher reserves against a given level
of vulnerability. If so, imposing the same (lower) coefficient as the non-RIM sample would
force the difference to be captured by the RIM dummy.
To allow for this possibility, Table 3 re-estimates the regression but now allowing both the
12
intercept and the slope coefficients to differ between RIM and non-RIM countries, and Table
4 repeats this analysis using the principal component proxies.9 In either case, the analysis
suggests that RIM countries do respond to vulnerabilities somewhat differently from nonRIM countries. Specifically, RIM countries hold more reserves against current account
vulnerabilities than non-RIM countries: a one standard deviation increase in the current
account proxy raises reserve holdings by 2.5 percent of GDP in non-RIM countries but by
3.7 percent of GDP in RIM countries (Table 4 [7]). This is partly offset, however, by lower
reserve holdings against capital account shocks: a one standard deviation increase in the
capital account proxy raises reserve holdings by 3.0 percent of GDP in non-RIM countries
but by only 2 percent of GDP in RIM countries (Table 4 [7]). The negative interactive RIM
dummy on capital account vulnerabilities is significant in all sub-periods except 1998-2004
when, presumably in response to the Asian crisis, RIM countries sought to insure themselves
against capital account shocks to the same degree as other EMEs who were learning from the
financial crises of the 1990s. There are a few other differences between RIM and non-RIM
countries, with some evidence that RIM countries are more sensitive to the costs of holding
reserves (especially in the latter part of the sample, 2005-10), and in some specifications and
sub-periods, the RIM interaction term on the undervaluation variable turns negative and
statistically significant (Table 3 [8]-[9]).10
With (largely) offsetting differences in RIM countries’ responses to current account and
capital account vulnerabilities compared to non-RIM countries, the story above remains
unchanged: even allowing for the different responses to these vulnerabilities, over the full
sample period, RIM countries hold 9.5 percent of GDP than non-RIM countries—though this
differential shrinks over time, falling to 4.7 percent of GDP and turning insignificant by
2005-10.
V. QUANTILE ANALYSIS OF RESERVE DEMAND
Thus far, our discussion has been based on the average behavior of RIM and non-RIM
countries. But as noted in Section II, there is also substantial variation in reserve holdings
both within and across the two groups of countries. We therefore turn to quantile regressions,
as specified in (2), which allow the relationship between reserve holdings and the various
9
We do not interact the “scale” variables (population, per capita income) with the RIM dummy as that simply
assigns a lot of weight on the population variable, given the high correlation between RIM countries and
population (0.41; or 0.31 excluding China).
10
The interpretation of this negative interaction term for the period 1980-2004 is that RIM countries held fewer
reserves for a given degree of exchange rate undervaluation—or, turning the statement around, they were able
to achieve a given undervaluation with less reserve accumulation than non-RIM countries. One reason may be
that RIM countries were more financially closed during this period.
13
explanatory variables to vary by the level of reserve holdings.11 As before, we first estimate
the full model (Table 5), then collapse the various proxies to their first principal component
(Table 6), and finally allow for the slope coefficients to differ between RIM and non-RIM
countries.
The quantile regressions show that, at relatively low levels (in percent of GDP) of reserve
holdings, precautionary motives against current account shocks dominate (Table 5[12-15]).12
In part, this echoes the story above: reserves were low during the early part of the sample (the
1980s), and precautionary motives against current account shocks were the main driving
force behind reserve holdings. But while the low-reserves observations are predominantly in
the early part of the sample, they are not exclusively so, thus low-reserve holders in general
(regardless of the time period) tend hold them against current account shocks. Perhaps more
surprising is that the mercantilist motive also seems to be important for the low-reserve
observations; closer examination, however, shows that in many of these cases, the country
was losing reserves—that is, the “undervaluation” reflected collapsed exchange rates (mainly
during crises) as the country exhausted its reserves, rather than deliberate undervaluation for
mercantilism.13
The picture is more mixed for capital account vulnerabilities. For high reserves holders,
capital account vulnerabilities (particularly broad money) are more important, but financial
openness and short-term debt are more important at the lower end of the distribution—while
being smaller and/or insignificant for high reserves holders (this is intuitive inasmuch as
many high reserve holders have very little short-term debt, and certainly ample reserves to
cover their debt).
Are RIM Countries Different?
As before, to ensure that results are not influenced by possible co-linearity of the proxies, we
repeat the analysis using their principal components in Table 6. As in the case of the period
11
Quantile regression techniques developed by Koenker and Bassett (1978) and Koenker and Hallock (2001).
Quantile regressions allow the estimation of conditional quantile functions—models in which various quantiles
(or percentiles) of the conditional distribution of reserve accumulation are expressed as functions of the
determinants. While classical linear regression methods are based on minimizing sums of squared residuals and
estimate models for conditional mean functions, quantile regression methods are based on minimizing
asymmetrically weighted absolute residuals and can estimate conditional functions at any part of the reserves’
distribution.
12
The 25th, 50th, 75th, 90th, and 99th percentiles correspond to reserves of 5.6, 11.0, 18.4, 27.3, and 54.0 percent
of GDP, respectively. Of the 175 RIM country observations, 119 (i.e., 68 percent) are above the sample median,
and 56 (i.e., 32 percent) are below the median.
13
Recall that our definition of mercantilism requires that the country be accumulating reserves in the context of
an undervalued currency (and that such accumulation be uncorrelated with precautionary needs).
14
analysis, collapsing the vulnerability proxies improves the precision of the estimation.
The results (Table 6 [1]-[4]) seem to suggest that RIM countries differ from non-RIM
countries only in the lower quartiles of the reserves distribution. For countries with belowmedian reserves, RIM countries hold some 3-4 percent of GDP more reserves than would be
expected given their vulnerabilities (and assuming the same response to these vulnerabilities
as other EMEs). For countries with above-median reserves, RIM countries do not appear to
hold more reserves than their peers.
The latter conclusion, however, is heavily dependent on the homogeneity assumption. In
Table 7, we allow the slope coefficients to differ across country groups. Doing so suggests
that (above the 25th percentile) RIM countries hold more reserves against current account
shocks than do non-RIM countries, but (across the reserves holding distribution), fewer
reserves against capital account shocks than their non-RIM counterparts. The net result is that
the RIM dummy becomes significant across the distribution, with the implication that,
controlling for current and capital account vulnerabilities (and the RIM countries response to
them), RIM countries hold some 3½ - 8½ percent of GDP more reserves than would be
expected.
VI. RESERVE ACCUMULATION TRENDS AND MODEL FIT
How well does the model account for RIM and non-RIM reserve accumulation? And what
has been driving reserve accumulation in recent years by RIM and non-RIM countries?
In this section we try to answer these questions by comparing the model’s fitted reserve
holdings to actual reserve holdings, and by decomposing the cumulative change in the stock
of reserves since 1990 into the various precautionary and mercantilist motives.
Figure 4 compares fitted and actual reserve holdings on the eve of the global financial crisis,
where the slope coefficients on the regressors are constrained to be the same for RIM and
non-RIM countries (i.e., Table 1 [7]). The picture reinforces the impression that RIM
countries generally hold more reserves than would be expected on the basis of their
vulnerabilities: except for Korea (and Indonesia, which is exactly on the line), all of the RIM
countries held more reserves than the model would imply—with the difference, in many
cases, exceeding 2 standard deviations of the fitted value; on average, RIM countries were
holding some 5½ percent of GDP more reserves than their corresponding non-RIM peers.
Figure 5 repeats this exercise but now using the specification where the slope coefficients are
allowed to vary between the RIM and non-RIM samples (Table 3 [7]). While this naturally
reduces the gap between actual and fitted values, it is noteworthy that most RIM countries
remain above the 45° line—with actual reserves of China, Malaysia, Thailand, and
the Philippines in 2007 exceeding the model’s fitted value by more than two standard
deviations. While the average residual for RIM countries shrinks from 5½ percent of GDP to
15
about 4½ percent of GDP, it remains substantial—consistent with the results above that, even
allowing for possible differences in responses to vulnerabilities, RIM countries hold higher
reserves than would be expected.
Accounting for Reserve Accumulation
What has been driving reserves accumulation by EMEs—and by RIM countries in
particular—over the past few years? Figure 6 (a, b, c) decompose the factors behind RIM
countries’ accumulation of reserves since 1990 using model specifications without the RIM
dummy, with the RIM dummy, and with the RIM slope coefficients respectively. The various
specifications tell a consistent story, with most of the difference between them being the split
between the contribution of current account vulnerabilities (greater when the RIM-specific
slopes are used) and the model residual.
Perhaps contrary to popular perception, and regardless of which specification is adopted, the
contribution of mercantilism is quite limited. The contribution of mercantilism, at its peak in
2004, was around 3½ percent of GDP; that is, in the counterfactual in which no mercantilist
motive was operative, RIM countries’ average reserves holding in 2004 would have been 3½
percent of GDP lower. Moreover, this 3½ percent of GDP contribution to the stock of
reserves either shrinks or stays relatively constant thereafter. In other words, there is little
evidence that, on average, RIM countries continue to accumulate reserves for mercantilist
reasons (as defined here). At the same time, it should be acknowledged that for non-RIM
countries, mercantilism never appears to have been a driving force behind their reserves
accumulation (Figure 7).
VII. CONCLUSIONS
Over the past three decades, emerging market economies have accumulated large stocks of
reserves, prompting suspicions that more than precautionary motives are at play. Asian
Pacific Rim countries, in particular, are often singled out for accumulating reserves as a byproduct of modern mercantilism—deliberate undervaluation of the currency as part of an
export-led growth strategy. In this paper, we seek to assess the validity of this view by
comparing Asian RIM countries to their non-RIM emerging market counterparts.
Our analysis points to three main conclusions. First, the reasons why EMEs hold reserves has
shifted over time, and depends on the level of reserves. In the 1980s, and at low reserve
holdings, reserves were mainly intended as insurance against current account shocks. In the
1990s, and following the emerging market financial crises in particular, insurance against
capital account shocks gained importance. Added to this, in the 2000s, is greater evidence of
mercantilist motives—that is, reserve accumulation that is uncorrelated with vulnerabilities
and that takes place in the context of a substantially undervalued currency. Second, RIM
countries do differ from their non-RIM counterparts: they hold more reserves against current
16
account vulnerabilities but fewer reserves against capital account vulnerabilities, and overall
hold more reserves than would be expected on the basis of these vulnerabilities. Adding the
mercantilist motive shrinks this difference, but only somewhat—so mercantilism provides, at
best, only a partial explanation. Third, confirming this, decomposition of the reasons behind
RIM countries’ reserve accumulation in recent decades suggests only a modest role for
mercantilist motives—with very little evidence that RIM countries, on average, continue to
accumulate reserves as a by-product of mercantilism.
We are thus left with a puzzle: why do RIM countries consistently hold more reserves than
precautionary (or even mercantilist) motives would imply? One reason may be that they want
to insure against shocks that are not captured here. Another reason may be that reserves
provide benefits beyond their insurance value. Yet a third possibility is that restrictions on
capital outflows mean that the public sector holds external assets that would otherwise be
accumulated by the private sector. Future research will need to explore these possibilities
more fully. What is clear is that, when it comes to holding reserves, RIM countries do differ
from their non-RIM emerging market counterparts—and simply blaming this on
mercantilism will not suffice.
17
REFERENCES
Aizenman, Joshua, and Jaewoo Lee, 2008, “Financial Versus Monetary Mercantilism LongRun View of Large International Reserves Hoarding,” The World Economy, Vol. 31,
pp. 593–611.
———, 2007, “International Reserves: Precautionary Versus Mercantilist Views, Theory and
Evidence,” Open Economies Review, Vol. 18, pp. 191–214.
Aizenman, Joshua, and Nancy P. Marion, 2003, “The High Demand for International
Reserves in the Far East: What Is Going On?” Journal of the Japanese and
International Economies, Vol. 17, pp. 370–400.
Bastourre, Diego, Jorge Carrera, and Javier Ibarlucia, 2009, “What is Driving Reserves
Accumulation.” Review of International Economics, Vol. 17, pp. 861–77.
Ben-Bassat, Avraham, and Daniel Gottlieb, 1992, “Optimal International Reserves and
Sovereign Risk.” Journal of International Economics, Vol. 33, pp. 345–62.
De Beaufort Wijnholds, J. Onno, and Arend Kapteyn, 2001, “Reserve Adequacy in Emerging
Market Economies.” IMF Working Paper 01/143 (Washington: International
Monetary Fund).
Caballero, Ricardo J., and Stavros Panageas, 2008, “Hedging Sudden Stops and
Precautionary Contractions.” Journal of Development Economics, Vol.85, pp. 28–57.
Cheung Yin-Wong, and Xingwang Qian, 2009, “Hoarding of International Reserves: Mrs.
Machlup’s Wardrobe and the Joneses.” Review of International Economics, Vol. 17,
pp. 824–43.
Dooley, Michael P., David Folkerts-Landau, and Peter Garber, 2005, International Financial
Stability (New York: Deutsche Bank).
Ghosh, Atish R., and Jonathan D. Ostry, 1997. “Macroeconomic Uncertainty, Precautionary
Saving, and the Current Account.” Journal of Monetary Economics, Vol. 40, pp. 121–
139.
Ghosh, Atish R., Jonathan D. Ostry, and Charalambos Tsangarides, 2010, “Exchange Rate
Regimes and the Stability of the International Monetary System.” IMF Occasional
Paper No. 270 (Washington: International Monetary Fund).
18
——— 2012, “Shifting Motives: Explaining the Buildup in Official Reserves in Emerging
Markets since the 1980s.” IMF Working Paper 12/34 (Washington: International
Monetary Fund).
Jeanne, Olivier, 2007, “International Reserves in Emerging Market Countries: Too Much of a
Good Thing?” Brookings Papers on Economic Activity: 1, Brookings Institution.
Jeanne, Olivier, and Romain Rancière, 2006, “The Optimal Level of International Reserves
for Emerging Market Countries: Formulas and Applications.” IMF Working Paper
06/229 (Washington: International Monetary Fund).
Koenker Roger, and Kevin F. Hallock, 2001, “Quantile Regression.” Journal of Economic
Perspectives, 15, pp. 143–56.
Koenker, Roger, and Gilbert Bassett Jr., 1978, “Regression Quantiles.” Econometrica, 46, pp.
33–50.
Obstfeld, Maurice, Jay Shambaugh, and Alan M. Taylor, 2010, “Financial Stability, the
Trilemma, and International Reserves.” American Economic Journal:
Macroeconomics, Vol. 2, pp. 57–94.
19
Figure 1. Reserves to GDP RIM, non-RIM, full sample
(Various periods)1
35
30
25
20
Full sample
15
RIM
non-RIM
10
5
0
1980-2010
1 RIM
1980-1997
1998-2004
2005-2010
countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
20
Figure 2a. Proxies for precautionary and
mercantilist motives
(RIM VS. non-RIM, 1980-2010)1
110
Figure 2b. Proxies for precautionary and
mercantilist motives
(RIM VS. non-RIM, 1980-1997)1
RIM
non-RIM
90
110
non-RIM
90
70
70
50
50
30
30
10
10
-10
-10
-30
RIM
-30
Imports/GDP
M2/GDP
ST Debt/GDP
Misalignment
Imports/GDP
1 RIM
countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand,
and Vietnam.
Misalignment
Figure 2d. Proxies for precautionary and
mercantilist motives
(RIM VS. non-RIM, 2005-2010)1
RIM
non-RIM
90
ST Debt/GDP
countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand,
and Vietnam.
Figure 2c. Proxies for precautionary and
mercantilist motives
(RIM VS. non-RIM, 1998-2004)1
110
M2/GDP
1 RIM
110
RIM
non-RIM
90
70
70
50
50
30
30
10
10
-10
-10
-30
-30
Imports/GDP
1 RIM
M2/GDP
ST Debt/GDP
Misalignment
countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand,
and Vietnam.
Imports/GDP
1 RIM
M2/GDP
ST Debt/GDP
Misalignment
countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand,
and Vietnam.
21
120
Figure 3: Reserves Distribution, various samples and periods
0
20
40
60
80
100
Reserves distribution 1980-2010 (full, RIM, and non-RIM samples)
120
Full sample
non-RIM sample
RIM sample
0
20
40
60
80
100
Reserves distribution by periods (full, RIM, and non-RIM samples)
1980-1997
1998-2004
Full sample
non-RIM sample
2005-2010
RIM sample
22
Figure 4. 2007 Actual vs. Predicted Reserves
(no differentiation RIM/non-RIM)1
60%
MYS
LBN
50%
CHN
JOR
BGR
40%
Actual 2007
RUS
MAR
30%
THA
VNM
BIH
PER
IND
20%
KOR
UKR
PHL ROM
TUN
ARM
HRV EGY
LVALTU
HUN
URY
ARG
CRI
KAZ
JAM POL
GEOIDN
BRA
GTM
TURVEN
ZAF SLV
PAK
COL
PAN CHL
MEX
DOMECU
10%
0%
0%
10%
20%
30%
40%
50%
Fitted 2007
1 RIM countries in
reserves.
red. Triangle RIM countries' actual reserves are above 2 standard deviations of the predicted
60%
23
Figure 5. 2007 Actual vs. Predicted Reserves1
60%
MYS
LBN
50%
CHN
JOR
BGR
40%
Actual 2007
RUS
THA
VNM
MAR
30%
BIH
PER
IND
20%
KOR
UKR ROM
PHL
TUN
ARM
HRV
EGY
LTU
LVA
HUN
URY
ARG
CRI
KAZ
JAM POL
GEO
BRA
IDNGTM
TURVEN
ZAF SLV
CHL
PAK
COL
PAN
MEX
DOMECU
10%
0%
0%
10%
20%
30%
40%
50%
Fitted 2007
1 RIM countries in
reserves.
red. Triangle RIM countries' actual reserves are above 2 standard deviations of the predicted
60%
24
Figure 6a. Cumulative differences decomposition 1990-2010
(RIM sample - specification without dummy or interactions)
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
1990
1992
1994
Capital account
1996
1998
Current account
2000
2002
Mercantilist
2004
Residual
2006
2008
2010
Reserves-scale
Figure 6b. Cumulative differences decomposition 1990-2010
(RIM sample - specification with RIM dummy only)
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
1990
1992
1994
Capital account
1996
1998
Current account
2000
2002
Mercantilist
2004
Residual
2006
2008
Reserves-scale
2010
25
Figure 6c. Cumulative differences decomposition 1990-2010
(RIM sample - specification with both dummy and interactions)
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
1990
1992
1994
Capital account
1996
1998
Current account
2000
2002
Mercantilist
2004
Residual
2006
2008
2010
Reserves-scale
Figure 7. Cumulative differences decomposition 1990-2010
(Non-RIM sample specification with RIM dummy only)
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
1990
1992
1994
Capital account
1996
1998
Current account
2000
2002
Mercantilist
2004
Residual
2006
2008
Reserves-scale
2010
26
Table 1. Current Account, Capital Account and Mercantilist Determinants of Reserve Demand
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
RIM
Sample
Scale
Log(per capita income)
Log(population)
scale
scale
scale
regime
scale
regime opp. cost
regime opp. cost
+CA
0.579*** 0.579***
(0.088) (0.090)
-0.073
-0.061
(0.071) (0.067)
Regime
Hard and Soft peg dummy
Volatility of neer
0.591***
(0.091)
-0.048
(0.064)
+KA
full
0.575*** 0.383***
(0.089) (0.081)
-0.055 0.142**
(0.063) (0.061)
80-97
0.438***
(0.139)
0.026
(0.088)
98-04
05-10
full
0.117
(0.093)
0.093
(0.076)
0.054
(0.092)
0.045
(0.070)
0.385***
(0.077)
0.135**
(0.055)
80-97
98-04
05-10
full
baseline
0.417***
(0.146)
0.086
(0.076)
0.113
(0.093)
0.104
(0.076)
0.047 0.378***
(0.095) (0.081)
-0.001 0.148***
(0.070) (0.052)
0.033
(0.121)
-0.007
(0.005)
0.016
(0.143)
-0.017***
(0.005)
0.056
(0.118)
-0.008*
(0.005)
0.308
0.057
(0.184)
(0.120)
-0.001 -0.019***
(0.006)
(0.005)
-0.110
(0.100)
-0.005
(0.006)
0.041
(0.119)
-0.008*
(0.005)
0.306*
0.065
(0.175)
(0.137)
-0.002 -0.019***
(0.006)
(0.005)
-0.124
(0.095)
-0.005
(0.006)
0.057
(0.119)
-0.008*
(0.005)
-0.390***
(0.061)
-0.065
(0.071)
-0.039
(0.082)
-0.393***
(0.069)
-0.025
(0.080)
-0.017
(0.095)
-0.068
(0.843)
-0.342
(0.933)
-0.013
(0.079)
0.039
(0.102)
-0.079
(0.860)
-0.110
(0.948)
-0.021
(0.078)
0.770*** 0.648***
(0.150) (0.134)
0.170** 0.239**
(0.078) (0.093)
0.311** 0.319**
(0.141) (0.128)
0.677***
(0.133)
0.212***
(0.058)
0.233*
(0.119)
0.532***
(0.153)
0.197**
(0.079)
0.395**
(0.192)
0.457**
(0.169)
1.385
(1.070)
0.206
(0.255)
0.509***
(0.149)
0.459
(1.989)
0.269
(0.579)
0.691***
(0.121)
0.215***
(0.058)
0.215*
(0.119)
0.690***
(0.149)
0.182**
(0.072)
0.292*
(0.156)
0.476***
(0.134)
1.371
(1.093)
0.231
(0.261)
0.426*** 0.690***
(0.138) (0.121)
0.386 0.215***
(1.898) (0.058)
0.163
0.231*
(0.580) (0.118)
0.100**
(0.048)
0.278***
(0.100)
0.227
(0.212)
0.113**
(0.049)
0.258***
(0.093)
0.274
(0.203)
0.060
(0.071)
0.195*
(0.104)
-1.068
(0.693)
0.123**
(0.056)
0.263**
(0.118)
0.509**
(0.242)
-0.023
(0.046)
0.426***
(0.143)
0.361**
(0.171)
0.113**
(0.046)
0.243**
(0.094)
0.285
(0.203)
0.114*
(0.059)
0.158
(0.108)
-0.275
(0.622)
0.125**
(0.057)
0.274**
(0.109)
0.497**
(0.227)
-0.030 0.115**
(0.047) (0.046)
0.391** 0.261***
(0.147) (0.092)
0.402**
0.276
(0.177) (0.199)
0.074
(0.109)
0.453*
(0.224)
0.324***
(0.081)
0.050
(0.164)
0.358***
(0.101)
-0.194
(0.174)
0.240***
(0.065)
0.054
(0.110)
0.333***
(0.081)
0.303*** 0.252***
(0.090) (0.065)
0.091
(0.142)
0.191** 0.249***
(0.089) (0.066)
0.746***
0.041
(0.157) (0.135)
Volatility of exports/GDP (3-yr sd)
Volatility of partner growth (3-yr sd)
Capital account
Financial openness
Log(broad money to GDP)
Short term debt to GDP
Mercantilist
Exchange rate undervaluation
0.864*** 0.825***
(0.180) (0.163)
0.769***
(0.157)
0.273
(0.166)
China dummy
Observations
R-squared
R-squared adjusted
Tests for groups' joined significance
p-value Scale
p-value Regime
p-value CA
p-value KA
(15)
0.074
(0.133)
-0.011**
(0.005)
Current account
Log(imports to GDP)
Constant
(14)
-0.025
0.003
(0.144)
(0.142)
-0.021*** -0.016***
(0.005)
(0.005)
Opportunity cost
Interest rate differential w/ US
RIM countries dummy
(13)
China
scale
regime
opp. cost
+mercantilist
0.515*** 0.409***
(0.087) (0.087)
0.135** 0.138**
(0.062) (0.063)
(12)
-2.453*** -2.441*** -2.430*** -2.449*** -2.554***
(0.133) (0.131)
(0.130)
(0.106) (0.100)
-2.422*** -2.553*** -2.608*** -2.361*** -2.096***
(0.128) (0.096)
(0.120)
(0.149)
(0.170)
0.357** 0.700***
-0.112
0.433**
(0.162)
(0.223)
(0.229)
(0.200)
-2.540*** -2.567*** -2.367*** -2.038*** -2.554***
(0.099)
(0.127)
(0.149)
(0.171) (0.096)
1,009
0.31
0.31
1,009
0.35
0.35
1,009
0.38
0.37
1,009
0.50
0.49
1,009
0.55
0.54
1,009
0.39
0.38
1,009
0.57
0.56
449
0.45
0.43
296
0.53
0.50
264
0.53
0.51
1,009
0.57
0.56
449
0.44
0.43
296
0.53
0.50
264
0.53
0.51
1,009
0.57
0.56
0.00
0.00
0.00
0.00
0.00
0.00
0.02
0.00
0.00
0.23
0.00
0.00
0.00
0.00
0.00
0.09
0.00
0.00
0.01
0.13
0.00
0.05
0.36
0.00
0.02
0.00
0.79
0.44
0.01
0.00
0.00
0.10
0.00
0.00
0.02
0.11
0.00
0.12
0.32
0.00
0.00
0.00
0.84
0.33
0.02
0.00
0.00
0.09
0.00
0.00
Notes:
1. Robust standard errors clustered by country in parentheses (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
27
Table 2. Principal Components Analysis: Current Account, Capital Account and Mercantilist Determinants of Reserve Demand
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
scale
regime
opp. cost
+mercantilist
full
80-97
98-04
05-10
full
RIM
Sample
Scale
Regime
Opportunity cost
scale
scale
regime
scale
scale
regime
regime opp. cost
opp. cost
+CA
-0.560*** -0.549*** -0.534***
(0.113)
(0.111)
(0.105)
-0.157**
-0.124*
(0.070)
(0.069)
-0.359***
(0.095)
Current account
-0.319**
(0.122)
-0.088
(0.067)
-0.112
(0.090)
0.304***
(0.083)
Capital account
-0.181
(0.114)
-0.093
(0.060)
-0.085
(0.089)
0.252***
(0.086)
0.237***
(0.050)
Undervaluation
RIM dummy
0.927***
(0.175)
0.893***
(0.181)
0.842***
(0.175)
0.557***
(0.198)
0.385**
(0.173)
-0.535***
(0.104)
-0.124*
(0.068)
-0.377***
(0.109)
0.227**
(0.092)
0.812***
(0.172)
-0.147 -0.321**
-0.006
0.177*
(0.106) (0.132) (0.101) (0.090)
-0.088
0.026
-0.029
-0.019
(0.058) (0.078) (0.062) (0.057)
-0.079
-0.075
-1.169
-0.266
(0.093) (0.092) (0.975) (1.018)
0.287***
0.162 0.205** 0.322***
(0.083) (0.111) (0.099) (0.077)
0.251***
0.235 0.242*** 0.179***
(0.050) (0.156) (0.047) (0.032)
0.342***
0.119 0.312*** 0.337***
(0.084) (0.106) (0.081) (0.124)
0.293*
0.453*
0.290*
0.103
(0.159) (0.250) (0.155) (0.208)
China dummy
Constant
Observations
R-squared
-2.226*** -2.236*** -2.208*** -2.187*** -2.344***
(0.092)
(0.094)
(0.094)
(0.089)
(0.093)
1,009
0.250
1,009
0.277
1,009
0.301
1,009
0.353
1,009
0.413
-2.216*** -2.363*** -2.593*** -2.297*** -2.011***
(0.093)
(0.092) (0.129) (0.105) (0.073)
1,009
0.315
Notes:
1. Robust standard errors clustered by country in parentheses (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
1,009
0.444
449
0.320
296
0.457
264
0.394
(12)
(13)
China
80-97
98-04
(14)
05-10
-0.120 -0.267**
0.058
0.141
(0.089) (0.123) (0.090) (0.085)
-0.097*
0.030
-0.036
-0.055
(0.058) (0.078) (0.067) (0.057)
-0.036
-0.023
-1.090
0.229
(0.093) (0.094) (1.028) (1.007)
0.339***
0.230* 0.263*** 0.349***
(0.079) (0.114) (0.090) (0.067)
0.258*** 0.335** 0.255*** 0.168***
(0.054) (0.142) (0.053) (0.033)
0.340***
0.068 0.361*** 0.337***
(0.080) (0.100) (0.076) (0.111)
0.777*** 0.920***
0.240 0.708***
(0.139) (0.203) (0.206) (0.204)
-2.342*** -2.552*** -2.296*** -2.003***
(0.096) (0.134) (0.107) (0.073)
1,009
0.451
449
0.327
296
0.445
264
0.416
28
Table 3. Adding Interactions: Determinants of Reserve Demand
(1)
Sample
Scale
Log(per capita income)
Log(population)
scale
(2)
(4)
scale
scale
regime
scale
regime opp. cost
regime opp. cost
+CA
0.579*** 0.579***
(0.088) (0.090)
-0.073
-0.060
(0.071) (0.067)
Regime
Hard and Soft peg dummy
Volatility of neer
(3)
0.585***
(0.091)
-0.047
(0.063)
(5)
+KA
0.514*** 0.403***
(0.088) (0.096)
0.132** 0.137**
(0.063) (0.066)
full
80-97
98-04
05-10
0.569*** 0.383***
(0.090) (0.089)
-0.056 0.142**
(0.064) (0.066)
0.414**
(0.154)
0.041
(0.099)
0.118
(0.115)
0.097
(0.083)
0.046
(0.112)
0.033
(0.071)
0.052
(0.136)
-0.008
(0.005)
0.368*
0.073
(0.200)
(0.150)
-0.001 -0.021***
(0.007)
(0.006)
-0.180
(0.117)
-0.005
(0.006)
-0.389***
(0.061)
-0.046
(0.080)
-0.031
(0.089)
-0.390***
(0.068)
-0.018
(0.088)
0.010
(0.097)
-0.085
(0.930)
0.136
(1.031)
0.790*** 0.649***
(0.165) (0.147)
0.150** 0.210**
(0.065) (0.087)
0.262* 0.292**
(0.154) (0.130)
0.675***
(0.146)
0.190***
(0.064)
0.216*
(0.125)
0.581***
(0.161)
0.165*
(0.085)
0.355*
(0.203)
0.448**
(0.190)
0.414
(1.607)
0.322
(0.281)
0.529***
(0.154)
0.441
(2.594)
0.441
(0.681)
0.113**
(0.056)
0.256**
(0.124)
0.275
(0.241)
0.124**
(0.057)
0.240**
(0.116)
0.319
(0.226)
0.092
(0.092)
0.120
(0.131)
-0.894
(0.789)
0.122*
(0.062)
0.256*
(0.136)
0.514*
(0.287)
-0.013
(0.052)
0.412**
(0.173)
0.386*
(0.217)
0.173* 0.256***
(0.101) (0.083)
0.092
(0.130)
0.371***
(0.087)
0.385***
(0.123)
Capital account
Financial openness
Log(broad money to GDP)
Short term debt to GDP
Mercantilist
Exchange rate undervaluation
-0.067
(0.246)
0.003
(0.011)
Opportunity cost interaction
Interest rate differential w/ US x RIM
-0.116
(0.230)
0.002
(0.010)
-0.043
(0.177)
-0.003
(0.008)
-0.050
(0.168)
-0.002
(0.006)
-0.023
(0.243)
0.004
(0.010)
0.001
(0.168)
0.001
(0.006)
-0.390*
(0.225)
-0.008
(0.009)
0.093
(0.227)
0.019**
(0.008)
0.068
(0.163)
-0.016*
(0.008)
-1.315
(1.404)
-1.775
(1.138)
-1.217*
(0.675)
-1.189
(1.363)
-1.265*
(0.672)
-1.627**
(0.667)
-0.528
(1.194)
-1.754
(1.665)
-0.193
(0.168)
3.377*
(1.961)
0.403
(0.293)
-0.011
(0.147)
2.782*
(1.426)
0.225
(0.268)
0.028
(0.138)
2.162
(1.309)
0.247
(0.239)
-0.216
(0.203)
0.986
(1.403)
0.569
(0.571)
0.076
(0.191)
-0.663
(1.878)
-0.460
(0.319)
-0.263
(0.163)
0.310
(3.080)
-1.354
(1.118)
-0.095*
(0.053)
-0.032
(0.187)
-1.119
(0.861)
-0.094*
(0.053)
-0.033
(0.170)
-1.104
(0.812)
-0.054
(0.096)
0.262
(0.184)
0.076
(1.332)
0.067
(0.076)
0.065
(0.299)
0.777
(0.610)
-0.113
(0.070)
0.111
(0.248)
-1.388
(0.904)
Current account interaction
Log(imports to GDP) x RIM
Volatility of exports/GDP (3-yr sd) x RIM
Volatility of partner growth (3-yr sd) x RIM
Capital account interaction
Financial openness x RIM
Log(broad money to GDP) x RIM
Short term debt to GDP x RIM
Mercantilist interaction
Exchange rate undervaluation x RIM
0.864*** 0.819*** 0.844***
0.050
0.226
(0.180) (0.242)
(0.238)
(0.253) (0.271)
-2.453*** -2.440*** -2.430*** -2.440*** -2.552***
(0.133) (0.132)
(0.130)
(0.107) (0.104)
Observations
1,009
R-squared
0.31
R-squared adjusted
0.31
Tests for groups' joined significance
Scale variables p-value
0.00
Regime variables p-value
Regime variables interaction p-value
Current account variables p-value
Current account variables interaction p-value
Capital account p-value
Capital account interaction p-value
All RIM terms p-value
0.00
(10)
0.021
(0.164)
-0.017***
(0.005)
Volatility of partner growth (3-yr sd)
Constant
(9)
0.039
(0.138)
-0.006
(0.005)
Volatility of exports/GDP (3-yr sd)
RIM countries dummy
(8)
0.070
(0.151)
-0.011*
(0.006)
Current account
Log(imports to GDP)
Volatility of neer x RIM
(7)
-0.012
0.022
(0.164)
(0.163)
-0.022*** -0.016***
(0.006)
(0.005)
Opportunity cost
Interest rate differential w/ US
Regime interaction
Hard and Soft peg dummy x RIM
(6)
scale
regime
opp. cost
+mercantilist
0.089
-0.087
-0.293* -0.263**
0.018
(0.192) (0.118)
(0.169)
(0.110)
(0.201)
0.706**
0.191
0.738*
0.042
0.045
(0.288) (0.249)
(0.395)
(0.341)
(0.232)
-2.417*** -2.557*** -2.588*** -2.368*** -2.081***
(0.131) (0.102)
(0.132)
(0.158)
(0.190)
1,009
0.35
0.35
1,009
0.379
0.373
1,009
0.502
0.494
1,009
0.556
0.546
1,009
0.388
0.382
1,009
0.570
0.560
449
0.461
0.431
296
0.535
0.495
264
0.554
0.512
0.00
0.00
0.91
0.00
0.01
0.84
0.00
0.05
0.94
0.00
0.14
0.00
0.00
0.88
0.00
0.00
0.37
0.00
0.38
0.92
0.00
0.24
0.00
0.10
0.00
0.00
0.19
0.98
0.00
0.36
0.00
0.08
0.00
0.03
0.13
0.24
0.00
0.34
0.33
0.36
0.00
0.38
0.00
0.04
0.10
0.45
0.00
0.15
0.00
0.88
0.28
0.10
0.01
0.35
0.00
0.05
0.00
Notes:
1. Robust standard errors clustered by country in parentheses (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
0.00
29
Table 4. Principal Components Analysis and Interactions: Determinants of Reserve Demand
Sample
Scale
Regime
Regime x RIM
Opportunity cost
Opportunity cost x RIM
(1)
(2)
scale
scale
regime
(3)
(4)
scale
scale
regime
regime opp. cost
opp. cost
+CA
-0.560*** -0.551*** -0.531***
(0.113)
(0.110)
(0.104)
-0.160**
-0.122*
(0.070)
(0.069)
0.020
0.002
(0.037)
(0.030)
-0.355***
(0.092)
-2.087
(1.824)
Current account
Current account x RIM
-0.307**
(0.120)
-0.081
(0.067)
0.005
(0.021)
-0.127
(0.090)
-1.929
(1.970)
0.289***
(0.086)
0.090
(0.097)
Capital account
Capital account x RIM
(5)
+KA
-0.161
(0.110)
-0.073
(0.059)
-0.001
(0.018)
-0.110
(0.090)
-0.386
(0.989)
0.226**
(0.089)
0.103*
(0.056)
0.253***
(0.050)
-0.090***
(0.021)
Undervaluation
Undervaluation x RIM
RIM
Constant
Observations
R-squared
0.927*** 0.902*** 0.907*** 0.785*** 0.906***
(0.175)
(0.181)
(0.180)
(0.186)
(0.192)
-2.226*** -2.236*** -2.209*** -2.194*** -2.370***
(0.092)
(0.094)
(0.094)
(0.090)
(0.093)
1,009
0.250
1,009
0.278
1,009
0.303
1,009
0.357
1,009
0.446
(6)
scale
regime
opp. cost
+mercantilist
-0.533***
(0.103)
-0.116
(0.069)
0.028
(0.025)
-0.371***
(0.103)
-1.403
(1.624)
0.166
(0.103)
0.354*
(0.199)
0.726***
(0.155)
-2.207***
(0.093)
1,009
0.323
Notes:
1. Robust standard errors clustered by country in parentheses (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
(7)
(8)
(9)
(10)
full
80-97
98-04
05-10
-0.135 -0.318**
-0.010
0.171*
(0.103)
(0.132)
(0.108)
(0.085)
-0.071
0.045
-0.036
-0.069
(0.059)
(0.080)
(0.064)
(0.055)
0.018
0.015
0.039*
-0.004
(0.013)
(0.010)
(0.023)
(0.024)
-0.102
-0.087
-1.153
0.829
(0.094)
(0.089)
(1.056)
(1.007)
-0.374
-0.797
1.553 -4.258**
(0.822)
(0.726)
(1.413)
(1.741)
0.257***
0.147
0.191* 0.315***
(0.087)
(0.112)
(0.107)
(0.072)
0.101*
0.116
0.060
-0.033
(0.054)
(0.087)
(0.044)
(0.063)
0.262***
0.235 0.247*** 0.190***
(0.051)
(0.162)
(0.048)
(0.026)
-0.073*** -0.081***
-0.015 -0.119***
(0.019)
(0.027)
(0.022)
(0.015)
0.284***
0.106 0.328*** 0.383***
(0.098)
(0.114)
(0.090)
(0.138)
0.107
-0.169
-0.084
0.026
(0.132)
(0.216)
(0.118)
(0.195)
0.726*** 1.207*** 0.473**
0.309
(0.183)
(0.324)
(0.203)
(0.199)
-2.380*** -2.615*** -2.307*** -2.036***
(0.092)
(0.131)
(0.108)
(0.073)
1,009
0.470
449
0.349
296
0.463
264
0.465
30
Table 5. Reserve Demand Across Quantiles
Percentile
Scale
Log(per capita income)
Log(population)
Regime
Hard and Soft peg dummy
Volatility of neer
Opportunity cost
Interest rate differential w/ US
Current account
Log(imports to GDP)
Volatility of exports/GDP (3-yr sd)
Volatility of partner growth (3-yr sd)
Capital account
Financial openness
Log(broad money to GDP)
Short term debt to GDP
Mercantilist
Exchange rate undervaluation
RIM countries dummy
Constant
Quantile regression estimated coefficients
(1)
(2)
(3)
(4)
25
50
75
90
1
Inter-quantile regression results
(5)
(6)
(7)
(8)
(9)
(10)
25 vs. 50 25 vs. 75 25 vs. 90 50 vs. 75 50 vs. 90 75 vs. 90
0.324***
(0.052)
0.161***
(0.036)
0.377***
(0.033)
0.151***
(0.023)
0.298***
(0.034)
0.128***
(0.019)
0.257***
(0.048)
0.092***
(0.032)
0.052
(0.047)
-0.009
(0.032)
-0.026
(0.053)
-0.032
(0.034)
-0.068
(0.063)
-0.069
(0.044)
-0.079**
(0.034)
-0.023
(0.020)
-0.120**
(0.047)
-0.060*
(0.033)
-0.041
(0.037)
-0.036
(0.029)
0.083
(0.069)
-0.011*
(0.006)
0.020
(0.045)
-0.004
(0.003)
0.028
(0.039)
-0.001
(0.002)
0.004
(0.061)
-0.003
(0.004)
-0.063
(0.058)
0.007
(0.005)
-0.055
(0.068)
0.010*
(0.005)
-0.079
(0.082)
0.008
(0.006)
0.008
(0.041)
0.002
(0.003)
-0.017
(0.062)
0.000
(0.004)
-0.024
(0.050)
-0.002
(0.003)
0.052
(0.061)
-0.131
(0.086)
-0.098
(0.060)
-0.152**
(0.061)
-0.183**
(0.080)
-0.151**
(0.074)
-0.204**
(0.081)
0.032
(0.078)
-0.021
(0.091)
-0.053
(0.058)
0.804***
(0.087)
0.364
(1.033)
0.272*
(0.155)
0.647***
(0.063)
0.190
(0.506)
0.195*
(0.105)
0.634***
(0.050)
0.067
(0.582)
0.134
(0.123)
0.481***
(0.081)
-0.091
(0.781)
0.120
(0.189)
-0.157**
(0.078)
-0.175
(0.864)
-0.077
(0.149)
-0.170** -0.323***
(0.084)
(0.108)
-0.298
-0.455
(1.143)
(1.256)
-0.138
-0.152
(0.175)
(0.227)
-0.013
(0.057)
-0.123
(0.488)
-0.061
(0.118)
-0.166*
(0.087)
-0.281
(0.811)
-0.076
(0.200)
-0.153**
(0.073)
-0.158
(0.711)
-0.015
(0.175)
0.139***
(0.024)
0.306***
(0.065)
0.541***
(0.136)
0.082***
(0.016)
0.303***
(0.046)
0.359***
(0.083)
0.061***
(0.016)
0.349***
(0.038)
0.115*
(0.067)
0.052**
(0.022)
0.329***
(0.058)
0.034
(0.094)
-0.057*** -0.078*** -0.087***
-0.021
-0.030
(0.021)
(0.025)
(0.030)
(0.015)
(0.023)
-0.003
0.043
0.023
0.046
0.026
(0.059)
(0.063)
(0.076)
(0.039)
(0.059)
-0.181 -0.425*** -0.507*** -0.244*** -0.326***
(0.126)
(0.140)
(0.146)
(0.076)
(0.105)
-0.009
(0.019)
-0.020
(0.048)
-0.081
(0.081)
0.337*** 0.213*** 0.240*** 0.179***
(0.070)
(0.052)
(0.037)
(0.053)
0.077
-0.022 -0.273*** -0.277***
(0.105)
(0.064)
(0.050)
(0.086)
-2.948*** -2.503*** -2.113*** -1.803***
(0.057)
(0.033)
(0.035)
(0.042)
-0.125**
-0.097
-0.158*
0.027
-0.033
(0.062)
(0.066)
(0.083)
(0.042)
(0.062)
-0.099 -0.350*** -0.354*** -0.251*** -0.255***
(0.091)
(0.103)
(0.130)
(0.058)
(0.096)
0.444*** 0.834*** 1.144*** 0.390*** 0.700***
(0.049)
(0.056)
(0.063)
(0.033)
(0.044)
-0.061
(0.049)
-0.005
(0.082)
0.310***
(0.036)
Observations
1,009
1,009
1,009
1,009
1,009
1,009
1,009
1,009
1,009
1,009
Pseudo R2
0.37
0.37
0.37
0.38
Tests for groups' joined significance
Scale variables p-value
0.00
0.00
0.00
0.00
0.47
0.61
0.21
0.05
0.02
0.30
Regime variables p-value
0.05
0.33
0.53
0.62
0.13
0.11
0.22
0.71
0.96
0.78
Current account variables p-value
0.00
0.00
0.00
0.00
0.16
0.09
0.01
0.90
0.18
0.18
Capital account p-value
0.00
0.00
0.00
0.00
0.02
0.00
0.00
0.00
0.00
0.42
Notes:
1. Quantile regression estimates for 1980-2010. Standard errors obtained using bootstrapping with 1000 replications (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
31
Table 6. Principal Component Analysis: Reserve Demand Across Quantiles 1
Percentile
Scale
Regime
Opportunity cost
Current account
Capital account
Undervaluation
RIM countries dummy
Constant
Observations
Pseudo R2
Quantile regression estimated coefficients
(1)
(2)
(3)
(4)
25
50
75
90
(5)
25 vs. 50
Inter-quantile regression results
(6)
(7)
(8)
(9)
25 vs. 75 25 vs. 90 50 vs. 75 50 vs. 90
(10)
75 vs. 90
-0.230***
(0.060)
-0.129***
(0.040)
-0.032
(0.070)
0.345***
(0.055)
0.227***
(0.029)
0.344***
(0.064)
0.506***
(0.092)
-2.725***
(0.042)
-0.152***
(0.055)
-0.095***
(0.031)
-0.157**
(0.079)
0.280***
(0.044)
0.202***
(0.024)
0.350***
(0.054)
0.283***
(0.082)
-2.300***
(0.036)
-0.057
(0.047)
-0.032
(0.027)
-0.195*
(0.105)
0.317***
(0.029)
0.193***
(0.021)
0.332***
(0.050)
0.062
(0.067)
-1.866***
(0.030)
0.028
(0.047)
-0.048
(0.033)
-0.122
(0.095)
0.332***
(0.036)
0.164***
(0.029)
0.350***
(0.053)
-0.024
(0.060)
-1.583***
(0.041)
0.078
(0.057)
0.034
(0.034)
-0.125*
(0.073)
-0.065
(0.050)
-0.024
(0.025)
0.006
(0.057)
-0.223**
(0.088)
0.426***
(0.038)
0.173***
(0.064)
0.097**
(0.041)
-0.163
(0.108)
-0.028
(0.055)
-0.034
(0.031)
-0.012
(0.069)
-0.444***
(0.093)
0.860***
(0.044)
0.258***
(0.068)
0.081*
(0.047)
-0.090
(0.105)
-0.013
(0.061)
-0.063
(0.039)
0.006
(0.073)
-0.530***
(0.102)
1.142***
(0.055)
0.095**
(0.046)
0.063**
(0.028)
-0.038
(0.090)
0.037
(0.038)
-0.009
(0.024)
-0.018
(0.050)
-0.221***
(0.071)
0.434***
(0.032)
0.180***
(0.061)
0.047
(0.038)
0.035
(0.096)
0.052
(0.051)
-0.038
(0.035)
-0.000
(0.064)
-0.307***
(0.090)
0.717***
(0.045)
0.085*
(0.046)
-0.017
(0.030)
0.073
(0.090)
0.015
(0.034)
-0.029
(0.029)
0.018
(0.050)
-0.086
(0.067)
0.283***
(0.037)
1,009
0.295
1,009
0.276
1,009
0.267
1,009
0.271
1,009
1,009
1,009
1,009
1,009
1,009
Notes:
1. Quantile regression estimates for 1980-2010. Standard errors obtained using bootstrapping with 1000 replications (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
32
Table 7. Principal Components Analysis and Interactions: Reserve Demand Across Quantiles 1
Quantile regression estimated coefficients
Percentile
Scale
Regime
Regime x RIM
Opportunity cost
Opportunity cost x RIM
Current account
Current account x RIM
Capital account
Capital account x RIM
Undervaluation
Undervaluation x RIM
RIM countries dummy
Constant
Observations
Pseudo R2
Inter-quantile regression results
(1)
25
sqreg25
(2)
50
sqreg50
(3)
75
sqreg75
(4)
90
sqreg90
(5)
(6)
(7)
(8)
(9)
(10)
25 vs. 50 25 vs. 75 25 vs. 90 50 vs. 75 50 vs. 90 75 vs. 90
iqreg2550 iqreg2575 iqreg2590 iqreg5075 iqreg5090 iqreg7590
-0.177***
(0.057)
-0.103**
(0.044)
0.036**
(0.018)
-0.053
(0.073)
-1.191
(1.513)
0.281***
(0.064)
0.065
(0.057)
0.259***
(0.032)
-0.056***
(0.018)
0.277***
(0.090)
0.133
(0.133)
0.869***
(0.132)
-2.776***
(0.044)
-0.093
(0.062)
-0.090***
(0.031)
0.008
(0.011)
-0.167**
(0.078)
-0.518
(0.838)
0.275***
(0.044)
0.098***
(0.037)
0.231***
(0.025)
-0.060***
(0.010)
0.309***
(0.070)
0.026
(0.115)
0.632***
(0.109)
-2.320***
(0.036)
-0.040
(0.057)
-0.015
(0.027)
0.009
(0.011)
-0.210**
(0.101)
0.284
(0.614)
0.282***
(0.031)
0.107***
(0.030)
0.222***
(0.026)
-0.073***
(0.011)
0.273***
(0.061)
0.119
(0.100)
0.389***
(0.093)
-1.898***
(0.032)
0.069
(0.043)
-0.008
(0.032)
0.014
(0.013)
-0.123
(0.091)
-0.165
(0.999)
0.291***
(0.041)
0.107***
(0.036)
0.182***
(0.041)
-0.095***
(0.015)
0.268***
(0.075)
0.181
(0.120)
0.258***
(0.095)
-1.612***
(0.047)
0.084
(0.057)
0.013
(0.037)
-0.029*
(0.016)
-0.114
(0.077)
0.673
(1.268)
-0.006
(0.055)
0.033
(0.052)
-0.029
(0.026)
-0.004
(0.016)
0.031
(0.079)
-0.107
(0.120)
-0.237*
(0.124)
0.456***
(0.038)
0.137**
(0.068)
0.088**
(0.043)
-0.027
(0.019)
-0.157
(0.111)
1.475
(1.470)
0.001
(0.063)
0.042
(0.058)
-0.038
(0.033)
-0.017
(0.018)
-0.004
(0.091)
-0.014
(0.142)
-0.480***
(0.138)
0.878***
(0.045)
0.247***
(0.065)
0.095*
(0.050)
-0.022
(0.021)
-0.070
(0.106)
1.026
(1.671)
0.010
(0.070)
0.042
(0.065)
-0.077
(0.048)
-0.039*
(0.022)
-0.009
(0.106)
0.048
(0.161)
-0.612***
(0.150)
1.164***
(0.058)
0.053
(0.055)
0.075***
(0.028)
0.002
(0.012)
-0.043
(0.094)
0.802
(0.781)
0.007
(0.036)
0.009
(0.033)
-0.009
(0.027)
-0.013
(0.011)
-0.035
(0.061)
0.093
(0.101)
-0.242**
(0.095)
0.422***
(0.033)
0.163**
(0.066)
0.082**
(0.038)
0.006
(0.015)
0.044
(0.094)
0.353
(1.173)
0.016
(0.051)
0.009
(0.043)
-0.048
(0.044)
-0.035**
(0.015)
-0.041
(0.087)
0.155
(0.136)
-0.374***
(0.121)
0.708***
(0.051)
0.110**
(0.050)
0.007
(0.030)
0.005
(0.012)
0.087
(0.081)
-0.449
(0.879)
0.009
(0.036)
-0.000
(0.034)
-0.039
(0.036)
-0.022*
(0.013)
-0.005
(0.071)
0.062
(0.112)
-0.132
(0.091)
0.286***
(0.040)
1,009
0.310
1,009
0.300
1,009
0.299
1,009
0.314
1,009
1,009
1,009
1,009
1,009
1,009
Notes:
1. Quantile regression estimates for 1980-2010. Standard errors obtained using bootstrapping with 1000 replications (*** p<0.01, ** p<0.05, * p<0.10).
2. RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
33
APPENDIX A: DATA AND SUMMARY STATISTICS
Table A1. Countries in the sample, and variable definitions and sources
Countries in the sample
Argentina, Armenia, Bosnia and Herzegovina, Brazil, Bulgaria, Chile, China, Colombia, Costa Rica, Croatia, Dominican Republic, Ecuador, Egypt, El Salvador, Georgia, Guatemala, Hungary, India
Indonesia, Jamaica, Jordan, Kazakhstan, Korea, Latvia, Lebanon, Lithuania, Malaysia, Mexico, Morocco, Pakistan, Panama, Peru, Philippines, Poland, Romania, Russia, South Africa, Thailand,
Tunisia, Turkey, Ukraine, Uruguay, Venezuela, Vietnam.
RIM countries are China, Indonesia, Korea, Malaysia, Philippines, Thailand, and Vietnam.
Variable definitions and sources
Variable
Description
Source
Log(reserves to GDP)
Log(per capita income)
Log(population)
Log(imports to GDP)
Volatility of neer
Interest rate differential w/ US
Natural log of ratio of foreign exchange reserves to USD GDP
Natural log of per capita income (at PPP)
Natural log of population
Natural log of ratio of imports to GDP
Twelve month standard deviation of end of period nominal effective exchange rate
ln[(1 + i)/(1 + iUS)], where iUS is the US interest rate corresponding to the definition
used for the national interest rate (deposit, money market, t-bill rate, lending)
3 year standard deviation of export to GDP (goods)
Chinn-Ito index index measuring a country's degree of capital account openness
Natural log of ratio of M2 to GDP
Dummy variable equal to 1 if the currency pegged and zero otherwise
Ratio of total short-term debt outstanding to GDP
3 year standard deviation of growth of trading partners' real GDP
Dummy variable equal to 1 if the currency is undervalued and zero otherwise
IMF,
IMF,
IMF,
IMF,
IMF,
IMF,
Volatility of exports/GDP (3-year sd)
Financial openness
Log(broad money to GDP)
Peg dummy
Short term debt to GDP
Volatility of partner growth (3-year sd)
Undervaluation dummy
Variable
IMF, World Economic Outlook and authors' calculations
Chinn and Ito (2009)
IMF, World Economic Outlook
Ghosh, Qureshi, and Tsangarides (2011)
IMF, World Economic Outlook
IMF, World Economic Outlook and authors' calculations
Authors' calculations
Table A2. Summary Statistics in the RIM sample
Mean
Std. Dev.
Min
Log(reserves to GDP)
Reserves to GDP
Log(per capita income)
Log(population)
Log(imports to GDP)
Volatility of neer
Interest rate differential w/ US
Volatility of exports/GDP (3-year sd)
Financial openness
Log(broad money to GDP)
Peg dummy
Short term debt to GDP
Volatility of partner growth (3-year sd)
Undervaluation dummy
Variable
World Economic Outlook
World Economic Outlook
World Economic Outlook
World Economic Outlook
International Financial Statistics and authors' calculations
International Financial Statistics and authors' calculations
-1.96
14%
8.07
18.41
-0.97
12.41
0.03
0.03
0.17
-0.35
0.54
0.13
0.11
0.42
0.79
220%
0.90
1.28
0.56
7.14
0.05
0.03
1.24
0.67
0.50
0.08
0.10
0.50
-3.91
2%
6.08
16.44
-2.45
2.41
-0.09
0.00
-1.84
-2.30
0.00
0.02
0.01
0.00
Table A3. Summary Statistics in the non-RIM sample
Mean
Std. Dev.
Min
Log(reserves to GDP)
Reserves to GDP
Log(per capita income)
Log(population)
Log(imports to GDP)
Volatility of neer
Interest rate differential w/ US
Volatility of exports/GDP (3-year sd)
Financial openness
Log(broad money to GDP)
Peg dummy
Short term debt to GDP
Volatility of partner growth (3-year sd)
Undervaluation dummy
-2.39
9%
8.59
16.72
-1.23
13.18
0.15
0.03
0.14
-0.92
0.60
0.14
0.14
0.29
0.88
240%
0.68
1.46
0.59
8.27
0.43
0.10
1.58
0.57
0.49
0.19
0.18
0.46
-5.42
0%
6.03
14.56
-3.07
0.81
-0.12
0.00
-1.84
-2.62
0.00
0.00
0.00
0.00
Max
-0.62
54%
10.30
21.02
0.01
45.45
0.28
0.17
2.48
0.66
1.00
0.45
0.66
1.00
Max
0.07
107%
9.88
20.92
-0.06
41.22
5.08
2.83
2.48
1.05
1.00
1.73
2.37
1.00