Campbell, John , “Does Saving Anticipate Declining Labor Income? An Alternative Test of the Permanent Income Hypothesis.” Econometrica 55: 1249–1273.
Carroll, Christopher D. , “The Buffer-Stock Theory of Saving: Some Macroeconomic Evidence.” Brookings Papers on Economic Activity 2: 61–135.
Fisher, Malcolm , “Exploration in Savings Behaviour,” Oxford University Institute of Economics and Statistics Bulletin XVIII: 201–278.
Ghosh, Atish R. and Jonathan D. Ostry , “Do Capital Flows Reflect Economic Fundamentals in Developing Countries?,” IMF Working Paper 93/34.
Leland, Hayne E. , “Saving and Uncertainty: the Precautionary Demand for Saving” Quarterly Journal of Economics 82: 465–473.
Miller, Bruce, L. , “The Effect on Optimal Consumption of Increased Uncertainty in Labour Income in the Multiperiod Case,” Journal of Economic Theory 13: 154–167.
Ostry, Jonathan D. , “The Balance of Trade, Terms of Trade, and Real Exchange Rate: An Intertemporal Optimizing Framework,” IMF Staff Papers 35: 541–573.
Ostry, Jonathan D. and Carmen Reinhart , “Private Saving and Terms of Trade Shocks: Evidence from Developing Countries,” IMF Staff Papers 39: 495–517.
Reinhart, Carmen M. and Peter Wickham , “Commodity Prices: Cyclical Weakness or Secular Decline?,” IMF Working Paper, forthcoming.
Skinner, Jonathan , “Risky Income, Life-cycle Consumption, and Precautionary Savings,” Journal of Monetary Economics 22: 237–255.
Svensson, Lars and Assaf Razin , “The Terms of Trade and the Current Account: The Harberger-Laursen-Metzler Effect,” Journal of Political Economy 91: 91–125.
van Wincoop, Eric , “Terms of Trade Uncertainty, Savings, and the Production Structure,” Journal of International Economics 33: 305–325.
Zeldes, Stephen P. , “Optimal Consumption with Stochastic Income: Deviations from Certainty Equivalence,” Quarterly Journal of Economics 104: 275–298.
Princeton University and International Monetary Fund, respectively. We thank Eduardo Borensztein, Carmen Reinhart, and Peter Wickham for helpful comments on an earlier draft.
Of course, the issue of the negative trend in commodity prices and the increased volatility are not independent. It may be that countries whose real incomes fall below a subsistence level owing to the secular behavior of their commodity exports may be unable to build up the necessary balances to insure themselves against greater volatility.
The early contributions include Fisher (1956), Friedman (1957), Leland (1968), Sandmo (1970), Drèze and Modigliani (1972), and Miller (1976). The recent literature includes Skinner (1988), Zeldes (1989), Caballero (1990), and Caroll (1992). See also van Wincoop (1992) for a two-period model of precautionary savings in an open economy.
Increased volatility should result in greater external savings, and so should not be affected by the country’s access to international loans. However, if a country has no access to international capital markets, and so is unable to smooth away the cyclical component of shocks to export earnings, the consumption-smoothing model presented below is unlikely to work well (for an empirical analysis of the performance of the consumption-smoothing model for a sample of developing countries, see Ghosh and Ostry (1993)). In addition, if real incomes are insufficient to maintain even a subsistence level of consumption, the feasibility of maintaining precautionary balances is called into question.
In addition, data limitations would preclude us from undertaking the empirical analysis of this paper if a more disaggregated commodity structure were employed.
The numeraire is taken to be the import good. In equation (2), therefore, the terms of trade are defined as the price of exports relative to imports. Also, bond holdings are in terms of the foreign good.
Recall that the consumption interest rate here is equal to the world interest rate since we ignore nontraded goods. This implies that the consumption rate of interest is nonstochastic here, and equal to the rate of time preference.
Export earnings are likely to be nonstationary in practice so their variance would not be defined. However, as shown below, consumption depends on the variance of the lifetime innovation to export earnings (where the innovation is by construction stationary), and so this variance is indeed well defined.
Caballero (1990) pioneers this approach in the context of a single household facing a labor income process with stochastic higher moments.
Allowing for more general ARMA processes is straightforward and does not alter the qualitative results.
It is straightforward to verify that the innovation to the Λ process, st, is proportional to the innovation to the variance process. Also, it is clear that if the variance process is an AR(1) with parameter ρ, then the Λ process will also be an AR(1) with parameter ρ.
We assume that ξ and s are independent stochastic processes.
To empirically identify precautionary savings effects through estimation of an equation such as (13), there must be sufficient variation in the volatility measure over the sample. Table 1 (discussed below) reveals considerable cross-country variation in the variance of the lifetime-innovation in export earnings. There is also substantial variation through time; for some formal tests, see Reinhart and Wickham (1994).
As is clear from (6), ξ is stationary so its variance is well defined.
This is the analogy to Campbell’s (1987) point that household saving ought to Granger-cause subsequent movements in household labor income. Thus, the longer shocks to the variance process persist, the greater will be the demand for precautionary savings, other things equal.
From (15), this variance is increasing in the persistence of the shocks to export earnings.
This part of the current account is present in both certainty-equivalent and non-certainty equivalent models alike.
Although this coefficient is not estimated here, its value using data for the United States was found to be in the range 0.2-0.5.
Indeed, as will be seen below, most of the parameter estimates are between 0.1 and 0.6, and therefore appear to reflect plausible values for the underlying parameters (which are not themselves identified).
Although not reported in the Table, the variation in the volatility measure through time is also large. Using the one-period measure, it is not uncommon, for example, to find ten- and even fourteen-fold increases in the coefficient of variation over the sample for particular countries. It is also the case that there is relatively less variation in the volatility measure for countries with relatively low average volatilities, for example among the diversified exporters or the recipients of private transfers.
This implies that the coefficient, a2, depends on relative, rather than on absolute, risk aversion. Hence the value for the risk aversion parameter of 2.0 given previously is indeed plausible.
All standard errors reported are heteroscedastic-consistent (White) standard errors.
As an empirical matter, if the variance of the lifetime innovation in export earnings is small, it also tends to be relatively stable through time, making identification of the precautionary savings motive all the more difficult.
The only exception is the five-year variance measure.
Indeed, Table 1 showed that the coefficient of variation of the lifetime innovation of export earnings was smaller for exporters with a diversified export base than for any other country grouping except for the recipients of private transfers. It is also the case that the coefficient of variation is relatively stable through time for countries in this grouping.
In some cases, the point estimate on the variance is negative, although it is never statistically significant.
This would depend, inter alia, on the persistence of shocks to the variance process.