On the Relationship between Domestic Saving and the Current Account: Evidence and Theory for Developing Countries

• Published date: 01 August 2020
• Author: MARKUS BRUECKNER,WOJTEK PACZOS,EVI PAPPA
• Date: 01 August 2020
• DOI: http://doi.org/10.1111/jmcb.12658

DOI: 10.1111/jmcb.12658

MARKUS BRUECKNER

WOJTEK PACZOS

EVI PAPPA

On the Relationship between Domestic Saving and

the Current Account: Evidence and Theory for

Developing Countries

We examine the relationship between domestic saving and the current ac-

count in developing countries. Our three main findings are that: (i) domestic

saving has a small effect on the current account; (ii) domestic saving has a

significant positive effect on the trade balance—this effect is much larger

than the effect that domestic saving has on the current account; and (iii) do-

mestic saving has a significant negative effect on net-current transfers. We

use countries in the SSA region during the period 1980-2009 as a laboratory

for an instrumental variables (IV) approach. The IV approach enables to

obtain estimates of causal effects. Underlying the IV approach is the signifi-

cant positive first-stage response of domestic saving to plausibly exogenous

annual rainfall: an unanticipated, transitory supply-side shock. Weconstruct

a small open-economy DSGE model with debt adjustment costs and endoge-

nous current transfers to match the empirical findings. The model enables to

examine the relationship between domestic saving and the current account

for different types of shocks. An important message of our paper is that, for

developing countries, estimates of the relationship between domestic saving

and domestic investmentare not informative for answering the question how

domestic saving affects a country’saccumulation of net foreign assets.

JEL codes:F30, F32, F35, F41, F44, F62

Keywords:domestic saving, current account, current transfers, small

open-economy model, financial frictions, Feldstein–Horioka puzzle.

THE RELATIONSHIP BETWEEN DOMESTIC SAVING and the current

account is an important topic in open-economy macroeconomics. At least since

We would like to thank twoanonymous referees and the editor for thoughtful comments that substan-

tially improved this paper.

MARKUS BRUECKNER is Professor in the Research School of Economics and CAMA, Australian Na-

tional University (E-mail: markus.brueckner@anu.edu.au).WOJTEK PACZOS is at Cardiff University

(E-mail: paczosw@cardiff.ac.uk).EVI PAPPAis at Universidad Carlos III de Madrid and CEPR (E-mail:

ppappa@eco.uc3m.es).

Received September 18, 2017; and accepted in revised form March 28, 2019.

Journal of Money, Credit and Banking, Vol.52, No. 5 (August 2020)

C

2019 The Ohio State University

1072 :MONEY,CREDIT AND BANKING

Feldstein and Horioka (1980), there has been a large amount of research done on

this topic. Already at the early stage of research, the question arose how to interpret

and compare the empirical findings to the predictions from theoretical models (see,

e.g., Obstfeld 1986). Results of least squares regressions that the empirical literature

documented are silent about what types of shocks are driving the variation in domestic

saving. For example: are these permanent or transitory shocks; demand- or supply-

side shocks; anticipated or unanticipated? This is a key issue when relating empirical

results to predictions from theoretical models. In all theoretical models, one has to

specify the type of shock that is causing the variation in domestic saving. Another,

separate issue is that estimation of a causal effect of domestic saving on the current

account is complicated by the endogeneity of domestic saving. Identifying exogenous

shocks to domestic saving in macro-economic data is difficult.

Our contribution to the literature is threefold. First, we provide least squares

estimates of the relationship between domestic saving and the current account for

a large panel of developing countries that covers approximately half of the world’s

population and spans about half a century.We report least squares estimates separately

for different regions in the world; regions are defined according to the World Bank

classification. The relationships established from the least squares regressions are

interesting, but interpreting them as causal or comparing them to a model is not

straightforward. To enable causal interpretation, we use an instrumental variables

(IV) approach. This is our second contribution to the literature. The instrument for

domestic saving is rainfall: an unanticipated, transitory supply-side shock. The IV

analysis is confined to sub-Saharan African (SSA) countries, for reasons described

below. For the SSA region, we can compare least squares to IV estimates. We can

compare least squares estimates for different regions, to see whether the least squares

estimates are different between the SSA region and other developing regions in

the world. Our third contribution to the literature is to build a DSGE model with

endogenous current transfers. The model delivers predictions of the relationship

between domestic saving and the current account. For a transitory productivity shock

like rainfall, we can compare the model’s predictions to the IV estimates. Beyond

comparison purposes, the theoretical model enables us to generate predictions of

the relationship between domestic saving and the current account for other types

of shocks, such as changes in interest rates on external debt or trend Total Factor

Productivity (TFP), for which, at the current date of writing, there is no clearly

exogenous, country-specific instrument available so that a causal relationship can

be estimated.

There are three main results from the least squares regressions: (i) in developing

countries the effect of domestic saving on the current account is small and, for some

regions, statistically indistinguishable from zero; (ii) there is a significant positive

and quantitatively sizable effect of domestic saving on the trade balance; and (iii)

a significant negative effect of domestic saving on net-current transfers. We show

that these results hold in developing countries across different regions in the world.

For developing countries, there is a substantial difference in the relationship between

domestic saving and the current account, and domestic saving and the trade balance.

MARKUS BRUECKNER, WOJTEK PACZOS,AND EVI PAPPA:1073

We document that this difference is specific to developing economies. For developed

economies, that is, High-Income Countries as defined by the World Bank, least

squares regressions show that there is no substantial difference in the relationship

between domestic saving and the current account, and domestic saving and the

trade balance.

In the IV regressions, we use year-to-year variations in rainfall to study how

a transitory, exogenous, and unanticipated supply-side shock to GDP affects the

relationship between domestic saving and the current account. We construct IV

estimates for a panel of 41 SSA countries during the period 1980-2009. The IV

approach is specific to the group of SSA countries. In SSA, the agricultural sector

is relatively large: the average agricultural GDP share is about one-third, and over

two-thirds of the population is employed in agriculture (World Bank 2017). It is

well documented that year-to-year variations in rainfall have a significant effect on

SSA countries’ year-to-year GDP growth (e.g., Miguel et al. 2004, Br¨

uckner and

Ciccone 2011). The novelty in this paper is to realize that because rainfall is a

transitory shock to GDP, the permanent income hypothesis predicts that domestic

saving should respond significantly to this shock as well. Indeed, our panel model

estimates show a highly significant positive effect of rainfall on domestic saving.

A 10% above country-mean increase in the level of rainfall increases the domestic

saving rate by around 1 percentage point.

The main finding from the IV analysis is that domestic saving has a quantitatively

small and statistically insignificant effect on the current account. Controlling for

country fixed effects, country-specific linear time trends, and year fixed effects the

coefficient on domestic saving in the current account equation is around 0.0 with

a standard error of around 0.2. In papers on the Feldstein–Horioka puzzle, see the

discussion below,the dependent variable is domestic investment. When the dependent

variable in the IV estimation is domestic investment, the estimated coefficient on

domestic saving is around 0.5; significantly different from zero and significantly

smaller than unity. That is: about half of domestic saving is channeled into domestic

capital accumulation. If one would have focused, exclusively, on the relationship

between domestic saving and domestic investment, as has been common practice in

previous papers on the Feldstein–Horioka puzzle, then one might have reached the

following conclusion: about half of domestic saving is used to increase net claims on

foreign assets. That would not havebeen the right conclusion for developing countries.

A direct approach to answering the question how domestic saving of developing

countries affects their net claims on foreign assets is to have in the econometric

model as dependent variable the current account (or the change in net foreign assets).

Estimation of that model shows that domestic saving has a small, near-zero, effect

on the current account. An important message is thus: for developing countries it is

in general not true that, in any given period, when domestic saving exceeds domestic

investment there will be an increase in net claims on foreign assets.

When we look at the components of the current account, we find that the effect

of domestic saving on net exports is significantly positive and quantitatively quite

large. The IV regressions yield a coefficient in the net export equation on domestic