RDP 2013-09: Terms of Trade Shocks and Incomplete Information 1. Introduction

Commodity prices are typically much more volatile than those of other goods or services.

… [M]any of the price movements last just long enough to convince investors and governments that ‘this time it is different’. And there is always a chance that some day it will be different. In the intervening period, long-range investments may have been set in train, new facilities built, and workers relocated … If prices stay high (or low) for a sufficiently long time, these reallocations of capital and labour could well be warranted and yield valuable returns … The trouble is that businesses, households, and policy-makers often get caught out … The inherent difficulty associated with predicting how long a boom (or bust) might last, and how high (or low) prices might go, makes the process extremely risky. Critics worry that a commodity-based economy will constantly find itself in motion, never quite settling down. John Murray, Deputy Governor of the Bank of Canada, 6 May 2010 (Murray 2010)

The terms of trade of many commodity-producing small open economies can be characterised by a succession of slow-moving long-run trends, augmented by high-frequency transitory fluctuations. As an illustration, Figure 1 shows the terms of trade – defined as the ratio of export prices to import prices – for six small open economies between 1961 and 2010. For each country, the data line represents the level of the terms of trade in logs, while the trend line shows an HP-filtered trend.^[1] Although the exact patterns differ across countries, each has experienced periods in which the trend terms of trade persistently decreased as well as periods in which it persistently increased. Changes in the trend terms of trade are often large. For example, the trend terms of trade decreased by around 50 per cent in Mexico during the 1980s and increased by over 50 per cent in Australia during the 2000s. Deviations from trend are also substantial. During the early 1970s, New Zealand's terms of trade was at times 30 per cent above its trend level.

The presence of both persistent and transitory movements in the terms of trade matters because the optimal response to a terms of trade shock depends upon the persistence of the shock. A positive terms of trade shock is similar to a positive productivity shock in that it allows an economy to sustainably increase consumption without a corresponding increase in factor inputs. A simple permanent income model would suggest that consumption-smoothing households will respond to a temporary increase in the terms of trade by saving some of the windfall and increasing consumption by the annuity value of the shock. In contrast, a permanent increase in the terms of trade will induce a larger immediate consumption response and a smaller increase in saving.^[2]

But in order for households and firms to respond in this textbook manner, they must first be able to identify which shocks are permanent and which are transitory. There is some reason to believe that they can do so. Unlike many other drivers of macroeconomic fluctuations – such as productivity or consumption preference shocks – the terms of trade are observable. Moreover, for many countries, changes in the terms of trade reflect broad global economic developments. For example, the increases in the terms of trade during the 2000s for the economies shown in Figure 1 were largely due to rising commodity prices, driven by strong economic growth in countries such as China and India (Kearns and Lowe 2011; Plumb, Kent and Bishop 2013; Kilian and Hicks forthcoming). To the extent that agents recognise the underlying causes of changes in the terms of trade, it seems plausible to think that they are able to forecast the persistence of these changes accurately.

And yet there is also evidence which suggests that identifying the persistence of terms of trade shocks is difficult. Consider Figure 2. This shows the evolution of the terms of trade in Australia during the 2000s, as well as successive forecasts of the the terms of trade published by the Reserve Bank of Australia (RBA). It is striking how consistently the forecasts underestimated the persistence of increases in the terms of trade despite the fact that many of the underlying drivers of the terms of trade boom were at least partly observable.^[3] Of course, a number of interpretations of Figure 2 are possible. It may be that the persistence of terms of trade shocks are predictable, but that forecasters made mistakes during the recent boom. For example, most forecasters appear to have underestimated the effect of strong growth in emerging market economies in Asia on demand for commodities. Also, forecasters may have overestimated the speed with which additional supply would come on stream. Alternatively, the prices of Australia's exports may have experienced a succession of positive, but temporary, shocks. Under this interpretation, forecasts like those in Figure 2 were optimal, but Australia's terms of trade merely received an unusual sequence of shocks. The results of this paper, however, suggest an alternative interpretation for the patterns of Figure 2, namely that the terms of trade does experience persistent shocks, but that it is difficult to identify these shocks in real time.

To reach this conclusion, I augment an otherwise standard small open economy model to include incomplete information about the persistence of terms of trade shocks. I then estimate the model using Bayesian methods on Australian data. The results suggest that agents face considerable difficulties in untangling the persistence of terms of trade shocks. In fact, agents' beliefs about the future path of the terms of trade are largely independent of the type of terms of trade shock that hits the economy. Consequently, it should come as no surprise that the response of the economy to terms of trade shocks differs substantially from that implied by models in which agents are perfectly informed about the nature of these shocks.

As well as documenting the existence of incomplete information about the persistence of terms of trade shocks, I also examine its implications for macroeconomic volatility. As the quotation at the beginning of this paper illustrates, it is often argued that an inability to forecast accurately the persistence of commodity price shocks exacerbates the macroeconomic volatility of small open economies. I demonstrate that, at least in the model used in this paper, this is not the case. This is because, while incomplete information about the persistence of terms of trade shocks increases the volatility of investment, it also encourages households to respond more cautiously to changes in the terms of trade. This makes consumption, the trade balance and output less volatile than they would be if agents had full information.

This paper is related to several strands of literature. Most directly, it complements work examining the effects of incomplete information about the composition of structural shocks, as in Angeletos and La'O (2010) and Blanchard, L'Huillier and Lorenzoni (forthcoming). An application of this methodology to international macroeconomics is found in Boz, Daude and Durdu (2011), who estimate open economy real business cycle models for Canada and Mexico that include uncertainty about the persistence of productivity shocks. This paper contributes to this literature in two ways. First, it provides empirical evidence of the existence of incomplete information about the persistence of an economically meaningful shock that has not previously been examined. Beyond this modest goal, the paper may also shed light on the pervasiveness of informational frictions about other shocks. Because terms of trade shocks are observable and can be rationalised in terms of broader economic developments, it seems plausible that households and firms have more information about these shocks than they do about other, unobserved shocks. Consequently, estimates of the extent of uncertainty regarding the persistence of terms of trade shocks may well represent a lower bound of the uncertainty regarding other shocks.

The paper also contributes to the literature examining the determinants of business cycles in small open economies. Aguiar and Gopinath (2007) demonstrate that a small open economy business cycle model can better match the moments of macroeconomic variables in developing economies if augmented with persistent shocks to the growth rate of productivity, which accumulate over time, to accompany standard transitory mean-reverting productivity shocks. Boz et al (2011) demonstrate that a similar result can be obtained with smaller productivity shocks if one assumes that agents have incomplete information about whether shocks are temporary or permanent. An open question in both of these papers is why some economies should experience more persistent, or less observable, shocks than others. This paper provides a potential answer to this question by highlighting the difficulty of identifying the persistence of commodity price shocks. If developing economies are more exposed to commodity price movements than advanced economies, then commodity price shocks could provide one explanation for why the nature of shocks to developing and advanced economies appears to differ.

This paper is also related to the literature describing the response of small open economies to terms of trade shocks. Key theoretical papers in this literature include Harberger (1950) and Laursen and Metzler (1950), who use a simple Keynesian approach, and Sachs (1981), Obstfeld (1982) and Svensson and Razin (1983), who examine the response to a terms of trade shock in an intertemporal optimisation setting. A number of papers have examined these relationships empirically. Otto (2003) constructs structural VAR models for a number of small open economies to examine the effect of transitory terms of trade shocks on the trade balance. He concludes that a positive terms of trade shock generally leads to an improvement in the trade balance, consistent with a basic consumption-smoothing model of the current account in a model with only transitory shocks. Kent and Cashin (2003) separate economies into those whose terms of trade shocks are typically permanent and those whose terms of trade shocks are typically transitory. They find that a positive terms of trade shock leads to a deterioration in the current account in the former economies and an improvement in the latter. They argue that their results are also consistent with standard intertemporal approaches to the current account in which agents smooth their consumption in response to transitory shocks and adjust consumption and investment more substantially in response to persistent shocks.

Other papers have examined the importance of terms of trade shocks as a source of macroeconomic fluctuations. The empirical results here are inconclusive. Based on structural VARs estimated for a number of developing countries, Broda (2004) concludes that terms of trade shocks typically explain less than 10 per cent of output volatility in developing countries. In contrast, using a simulated real business cycle model, Mendoza (1995) finds that terms of trade disturbances explain 56 per cent of output fluctuations in developing countries and 33 per cent of output fluctuations in advanced economies.

To some extent, the results of this paper reinforce those of the previous empirical literature. For example, I find that transitory positive terms of trade shocks lead to an increase in net exports while permanent positive terms of trade shocks trigger a decrease in net exports. However, as outlined in Blanchard et al (forthcoming), if agents have incomplete information about the persistence of shocks then it is not possible for an econometrician to identify permanent and transitory shocks in the data. Consequently, the finding that agents are largely unable to differentiate between permanent and transitory terms of trade shocks raises questions about the identification of these shocks in other papers.^[4]

The paper proceeds as follows. Section 2 outlines the model and clarifies the information structure. Section 3 describes the estimation and summarises the key results. Sections 4 and 5 discuss the implications of the empirical results for the response of the economy to terms of trade shocks. Section 6 reports a series of robustness checks and Section 7 presents conclusions.

Footnotes

Although the data are quarterly, the trend was calculated using a smoothing parameter of 64,000 rather than the usual 1,600. This reflects the fact that commodity price cycles – which drive the terms of trade in these economies – are typically longer than business cycles. [1]

The responses to temporary and permanent terms of trade shocks may differ from this simple permanent income example, depending, for example, on consumers' willingness to substitute intertemporally and between tradeable and non-tradeable goods. Nonetheless, the key point that the optimal responses to transitory and permanent shocks differ is generally true. [2]

It is worth noting that the RBA forecasts were not unusual in underestimating the persistence of the increase in the terms of trade. [3]

This issue may be less of a concern for Kent and Cashin (2003) as they do not identify individual transitory or permanent shocks. Their approach can be viewed as implicitly assuming that agents have no information about the persistence of individual terms of trade shocks and merely expect the persistence of the average shock. It turns out that this assumption about the information structure is not a bad approximation to the results of this paper. [4]