Comments on Fer.Broner, A.lberto Martin and Jaime Ventura
Comments on Fer.Broner, A.lberto Martin and Jaime Ventura
The Cost of Holding Reserves:
Evidence from India
by Abhijit Sen Gupta
India Policy Forum
New Delhi, July 15, 2008
As of December 2007, India’s international reserves were
$270 billion. This is about 18% of GDP compared to an
advanced country average of 4%.
The paper considers two questions:
Are emerging market central banks hoarding excess
What is the opportunity cost of India’s foreign reserve
For the first question, Sen Gupta begins by summarizing a
simple theoretical model of reserve demand from Jeanne
and Ranciere (2008). He then estimates reserve
holdings for a panel of 167 countries over 1980-2005.
He finds that several major emerging economies are
holding reserves well in excess of those predicted by his
model. For India, his estimate of the excess is $22
For the second question, he calculates the cost of holding
these additional reserves to be between 0.3% and 2% of
Why should central banks hold foreign reserves?
countries face uncertain export revenues and reserves
allow a country to sustain imports in the presence of
limited capacity to run a trade deficit
ii) foreign credit may be withdrawn suddenly resulting in
a net demand for foreign exchange.
Both of these entail a decision to avoid currency
Both are raised by Keynes (1930), for example, notably the
relationship of reserves to short-term foreign debt.
These have led, respectively, to rules of thumb expressed
in terms of months of import cover and short-term
amortization to GDP ratios (Greenspan-Guidotti (1999)
following an IMF proposal (1997)).
They are reflected in regressions by the inclusion of the
import share and short-term indebtedness, controlling for
per capita GDP.
Sen Gupta adds some variables to measure the role of the
exchange rate regime, domestic financial depth and
international financial integration.
Why add financial variables?
The typical answer is that these reflect the exposure of the
economy to external crises. Henry Thornton (1802)
raised this issue but articulated it in terms of external and
internal drains (as emphasized by Obstfeld, Shambaugh
and Taylor (2008)).
With an open capital account, domestic currency deposits
can be withdrawn to purchase foreign deposits or
currency. A run on the banking system (an internal
drain) coupled with an external drain leads to a
depreciation unless the lender of last resort already
holds sufficient foreign assets to exchange for domestic
public liabilities in defense of the currency.
The ratio of M2 to GDP measures the deposit base at risk
in a domestic financial crisis. Domestic demand deposits
are not a proxy – they are what matters.
The exposure of the lender of last resort to an external
drain in the event of a domestic financial crisis depends
on the openness of the capital account and, perhaps,
variables that capture domestic confidence.
The exchange rate regime may matter because it informs
us of the central bank’s exchange rate objective.
However, it can be a poor measure of the weight the
central bank places on countering exchange rate
Why go on about this?
The panel regression is a hybrid based on some loose
arguments about external crisis exposure, political
instability and exchange rate policies:
If the objective of reserve holding is insurance against
export revenue fluctuations with limited balance of
payments financing, then import share should be in the
If the objective is maintaining domestic financial stability,
then financial openness, M2/GDP and trade openness
should be in the regression.
2. Sen Gupta’s regression results tell us that reserve
holdings for India and China, among a few others,
exceed those predicted by a model that includes
variables to measure the domestic financial stability
objective of the central bank.
The central banks of these countries may anticipate further
liberalization of the capital account or perceive greater
risks (costs) of domestic financial crises than other
emerging market economy central banks. Reserve
holding should be forward looking if the central bank
expects further international integration of domestic
1. The exchange rate regime and political instability
variables are not significant for the emerging market
sample. Perhaps, the political stability story is
unsuccessful. It may be worthwhile separating floating
regimes with very low exchange rate volatility from actual
2. Per capita GDP and M2/GDP are highly correlated. I
suggest comparing regressions with each and not the
other. Also, consider the alternative of a model with
trade share and the money to output ratio.
3. The theoretical model has little to do with the empirical
model of reserves or the estimation of the opportunity
cost of reserve hoarding.
In the theoretical model, the government offers consumption
insurance to households against export revenue shocks. The
household pays a premium xR for a contingent indemnity R.
A maximum with respect to R (but not x) is found.
The calculation of the steady state reserve level reveals that this
is repeated insurance. However, reserves are (noncontingent) interest-bearing assets subject to exhaustion.
This is an incomplete model of self-insurance by the
government. It lacks the government’s budget constraint and
does not account for its liquidity constraint (stock outs).
The approach differs from buffer stock models (eg. Frankel and
Jovanovic (1981) and Flood and Marion (2003)) that
acknowledge the downward constraint and cost of holding
Opportunity cost of reserves
Three measures of the marginal cost of reserves are given.
The opportunity cost of sterilizing capital inflows is usually
measured by the real interest differential between
domestic public debt and international reserve assets
corrected for default risk.
If the central bank sells reserves to purchase government
debt, it reverses its original sterilization and the currency
Modeling the insurance motive for reserve hoarding by the
emerging markets requires an objective for the central
bank that includes the risk and costs of financial crises, a
monetary policy objective and constraints on the central
The reserve demand estimates do not account for out of
sample policy changes that affect the openness
variables. The estimates give an upper bound on the
costs of holding reserves in a liberalizing policy
environment (conversely, lower bound).