RDP 2007-10: Trade Costs and Some Puzzles in International Macroeconomics 1. Introduction

The international macroeconomics literature has identified a number of key puzzles which are associated with empirical correlations. One, called the Feldstein-Horioka puzzle, is why domestic investment is correlated with domestic saving when, in a world with open capital markets, savings should flow to countries with the greatest investment opportunities. For example, in the sample of developed economies used in this paper, there is a correlation between annual investment and saving of 0.47 when theory would predict a low correlation.^[1] Another, called the purchasing power parity (PPP) puzzle, is why the real exchange rate is very persistent despite the relative flexibility implied by high nominal exchange rate volatility.^[2] For example, Froot and Rogoff (1995) and Rogoff (1996) find that it takes over three years for a deviation from PPP to be reduced by one-half in developed economies. A third puzzle, called the international consumption correlation puzzle, is why, in a world of international trade and capital flows, financial instruments (or other mechanisms) have not developed so as to better help consumption smoothing in the face of country-specific shocks. Some theories suggest that countries should smooth consumption such that every country's consumption is perfectly correlated with world consumption.^[3]

In a seminal paper in international macroeconomics, Obstfeld and Rogoff (2001) argue that trade costs could explain these puzzles (as well as some other puzzles, such as the extent of the bias towards consuming domestically produced goods and the bias towards domestic shares in equity portfolios). While trade costs provide a natural explanation for why individuals consume a relatively high share of domestically produced goods, this paper seeks to assess the extent to which the data suggest that trade costs can explain the Feldstein-Horioka, PPP and international consumption correlation puzzles. In essence, Obstfeld and Rogoff suggest that trade costs for goods can cause phenomena that are similar to financial market imperfections, which are natural explanations for both the Feldstein-Horioka and consumption correlation puzzles.^[4] With financial market imperfections, global savings do not necessarily flow to the most profitable investments, which would explain the high correlation of domestic investment and saving. Similarly, financial market imperfections due to trade costs may also lessen the extent of consumption smoothing across countries. Trade costs also explain why the same good may not cost the same amount in different countries and so provides a natural explanation of deviations from PPP (see the discussion in Dumas 1992).^[5]

The contribution of this paper is to assess the plausibility of the trade-cost explanation for these three puzzles within a simple empirical analysis which is consistently applied across each of the puzzles. Some existing literature looks at whether trade costs can explain the potentially related puzzle of the home bias in portfolio holdings (Coeurdacier 2006 and van Wincoop and Warnock 2006). Other work has found evidence between the home bias in portfolio holdings and the international consumption correlation puzzle (Sørensen et al 2005). Other related literature looks at the channels of consumption smoothing such as the role of government spending (Asdrubali, Sørensen and Yosha 1996, for example).

The extent to which trade costs might explain these three puzzles in macroeconomics is interesting in a number of respects. It would aid in the prediction and interpretation of movements in the real exchange rate, investment and consumption. For example, if trade costs play a role in these puzzles and continue to decline, it would be expected that real exchange rates would tend to adjust to shocks more quickly, domestic investment and saving would become less closely related and consumption would become smoother (at least to the extent of not responding to domestic shocks). It is also likely to change the way in which shocks that influence the real exchange rate influence the rest of the economy.^[6]

Perhaps as importantly, if there is evidence that trade costs explain the consumption correlation puzzle, it would suggest that lower trade costs may not only raise welfare by enabling greater opportunities to benefit from comparative advantage, but also by facilitating consumption smoothing. Similarly, if trade costs explain the Feldstein-Horioka puzzle, it suggests that lowering trading costs will help saving flow to where investment returns are higher.

To preview the results, I find some evidence that trade costs appear to play a role in each of the three puzzles for the developed economies in the sample, though sometimes the relationships are imprecisely estimated and trade costs appear to only explain certain aspects of the correlations (which are indicative of the puzzles). Figure 1 summarises these results. The horizontal axes show a measure of trade costs for each country's imports into the US. The trade costs are expressed as a percentage of the free-alongside-ship (FAS) cost. FAS cost is closely related to the more commonly cited free-on-board (FOB) cost, which includes loading costs. On the vertical axis, the top panel of Figure 1 plots a measure of the correlation between investment and saving for each country (obtained by regressing investment on saving, both as a percentage of GDP). The positive relationship suggests that lower trade costs are associated with a smaller Feldstein-Horioka puzzle (a lower correlation between investment and saving). The middle panel of Figure 1 plots trade costs against a measure of real exchange rate persistence within each country – based on the coefficient of regressing the real exchange rate on the first period lag of the exchange rate. The bottom panel of Figure 1 plots trade costs against a measure of the correlation between domestic output and consumption.^[7] These positive relationships support Obstfeld and Rogoff's (1996) theory. Later it will be argued that trade costs play an economically significant role in explaining the Feldstein-Horioka and PPP puzzles.

The rest of the paper is structured as follows. The main empirical methods used are described in Section 2. Section 3 presents the data and the main results. Some results examining whether capital restrictions play a role are presented in Section 4. Section 5 presents results using an alternative measure of trade costs and provides some additional checks of robustness. Section 6 discusses in more detail those aspects of the correlations which trade costs appear to be able to explain. Brief conclusions are drawn in Section 7.

Footnotes

This puzzle was originally described by Feldstein and Horioka (1980). More recent work by Obstfeld and Rogoff (1996) and Feldstein (2005) suggests that the extent of the puzzle has lessened but the puzzle is still important. The correlations in Table 1 suggest that the extent of the correlation has fallen from about 0.6 to 0.3. [1]

Taylor (2000) argues that it is generally difficult to test for deviations from PPP. Even if one is uncomfortable characterising the correlations discussed in this paper as puzzles, it is useful to know the extent to which they can be accounted for by trade costs for the purposes of better understanding and predicting economic developments. [2]

For example, see the discussion in Chapter 6 of Obstfeld and Rogoff (1996). In this paper, I put aside concerns that have been raised about how the low consumption correlation is only a puzzle under a set of strong assumptions (such as that the amount of leisure does not affect the marginal utility of consumption), as a number of these assumptions are fairly standard in the macroeconomic literature. [3]

Trade costs imply that a country's consumption patterns can affect prices, and, therefore, expected changes in consumption patterns can affect interest rates. Trade costs can also affect incentives for portfolio diversification and the ability to share risks internationally since payments to foreigners can only be made in the form of traded goods. Fazio, MacDonald and Melitz (2005) also suggest that trade costs can explain the Feldstein-Horioka puzzle because of the differential between consumption and output prices. [4]

Some of the controls used later in the panel data regressions could be viewed as ways of trying to control for other possible explanations of the puzzles. This is particularly relevant for the Feldstein-Horioka puzzle for which a number of explanations already exist (though Obstfeld and Rogoff 2001 describe them as not thoroughly convincing). [5]

Also, if the real exchange rate becomes less persistent it may be easier to determine what drives real exchange rate movements (see Obstfeld and Rogoff 1996). [6]

The regressions estimated for Figure 1 are the equivalent of Equations (1), (3) and (4) below without the trade-cost terms and with a constant instead of the fixed effects. [7]