VoxEU Column International trade Monetary Policy

Trade openness and inflation

Changes in openness to trade can disrupt the inflation forecasting on which many nations' monetary policies depend. New research suggests an innovative time-series openness measure that addresses some of the shortcomings of existing measures.

As the popularity of inflation targeting has spread, inflation forecasting has come to play a central element in many nations’ economic policy-making. Numerous emerging market and developing economies have undergone trade liberalisation. Since lowering import barriers typically exerts a downward pressure on prices, evolving trade liberalisation represents a structural break from the inflation forecasting perspective. Omitting this factor can confuse modellers studying the determinants of inflation and output. For instance, a greater degree of openness due to trade liberalisation is likely to lower the rate of inflation and may alter the influence of the real exchange rate on growth1, via the impact on the demand for exports and leakage of demand into imports. Long time series of better measures of openness should improve the modelling and forecasting of output and inflation.

Unfortunately, attempts to measure trade policy are fraught with measurement problems for observable components (such as tariffs), and by the presence of difficult-to-quantify components of policy (such as quotas and other non-tariff barriers). Several authors suggest that the sheer complexity of factors influencing trade - including tariffs and surcharges, drawbacks, quotas and licences, other non-tariff barriers such as differing international standards, exchange controls and occasionally, trade and foreign exchange sanctions – argues against any single measure adequately proxying for trade policy. From several critical surveys of empirical measures of trade policy2, the consensus appears to be that no measure of trade policy is free of methodological problems. Moreover, the measures often correlate poorly with trade volumes.

We draw on this literature to evaluate existing trade policy measures, focusing on viable indicators for use in time series macro-modelling, particularly of inflation. The most commonly used measures of trade policy outcomes, such as trade flows to GDP (in real or nominal terms), are amenable to time series application. However, they are also influenced by a range of factors including country size and location, the tradable natural resource base and the extent and ease of capital inflows. In a cross-country context, there have been attempts to remove these other factors to give ‘purged’ measures of trade-openness3. However, this is not a useful methodology for time series work as the essentially static ingredients in these regressions allow insufficient time variation within countries.

A second set of indicators, administrative or incidence-based indicators, such as nominal average tariffs and effective protection, and coverage of non-tariff barriers (NTB), can be used in time series studies. However, they suffer from measurement problems, especially severe for non-tariff barriers. Rodriguez and Rodrik4 argue these measures can effectively rank-order countries according to the restrictiveness of their trade regimes, but they and others emphasise several measurement biases. The use of unweighted average tariffs will weight too heavily those commodities that are only a small fraction of imports. On the other hand, trade-weighted average tariffs (e.g. aggregate import duties or trade taxes divided by the volume of imports) give negligible weight to prohibitive tariffs since the corresponding imports are typically low5. If statutory measures are used, these may not necessarily be binding due to poor enforcement or even corruption. Tariff averages will be poor proxies for overall restrictions where NTBs are important.

The coverage of NTBs relative to total imports is widely used, but various authors note this proxy of trade policy is also measured with error. Available data tends to report the number of tariff lines on which one of a small number of easily identifiable NTBs applies, or the percentage of product categories subject to import licenses. This is uninformative on the severity of the NTBs and tends to exclude less-easily quantifiable barriers to trade, such as local procurement requirements, or health and safety standards.

A third set of indicators combines quantitative and qualitative trade measures for an overall index capturing different aspects of trade policy, or uses institutional information to generate a subjective judgement of “openness”. Though data intensive such measures could be constructed for time series studies, but their value as trade policy measures is in doubt. An example is the widely-cited composite trade policy indicator of Sachs and Warner6, where a dichotomous variable measures zero (closed) or one (open) - depending on thresholds for incidence-based trade policy indicators, the size of the black market premium, whether the economy is socialist and whether there is a state monopoly on major exports. In a far-reaching review of growth and trade openness studies, Rodriguez and Rodrik (op. cit.) find this indicator explains growth due to two of its components being correlated with other determinants of growth, the black-market premium and state monopoly variables, with little contribution from its tariff and NTB components. The same authors are critical of subjective indices which they judge as “highly contaminated by judgement biases or lack(ing) robustness”.

Finally, a fourth set of indicators involves international price comparisons, often suitable for time series analysis. This idea lies behind Dollar’s well-known ‘‘price distortion’’ index7. Dollar uses international data on the relative price of consumption goods and tries to 'purge' it of its non-traded component by taking the residual from a regression on urbanisation, land and population. Rodriguez and Rodrik (op. cit.) demonstrate, even assuming that this method is successful, that the purged variable is theoretically inappropriate as a measure of trade restrictions when Purchasing Power Parity is violated over the sample, and/or when there are export taxes or subsidies in use. It is a well-established empirical fact that absolute and relative PPP do not hold over the medium-term, but the deviations even from relative PPP are very long-lived8. Moreover export taxes and subsidies are commonly applied, especially in developing countries.

Our own time series measure of trade openness overcomes many of the shortcomings of existing measures, and encompasses both observable and unobservable trade policy. It is based on modelling the ratio of manufacturing imports to home demand for manufactures, a variable likely to be strongly influenced by trade openness. This is purged of other determinants by including in the model: GDP growth, the terms of trade and the real exchange rate or relative import prices. We measure observable trade policy in the model with trade-weighted tariffs. The unobservable trade policy component (NTBs and mis-measurement of tariffs) is captured in the model by a smooth, non-linear stochastic trend9. Our openness measure is constructed as a weighted combination of known trade policy and the stochastic trend, with weights from the regression coefficients in the model. In application to South Africa, the shape of the trend is likely also to reflect such factors as the lifting of capital controls and unification of dual exchange rates in the 1990s (as often used in the composite measures above), and the lifting of externally imposed trade sanctions. Our measure thus captures a broader sense of “openness” than is only due to trade policy. It is shown in Figure 1 and corresponds well with the known phases of liberalisation.

Figure 1: Openness measure and stochastic trend, plus the tariff ratios for South Africa

Our weighted openness measure and also its separate components are included in quarterly models of wholesale price inflation in South Africa. The evidence in our recent research suggests that increased openness has significantly reduced the mean inflation rate and has also reduced the exchange rate pass-through into wholesale prices. The fit of the model is best when the stochastic trend and tariff measures are included separately. The evidence from price setting suggests a larger weight on the more permanent part of tariffs than the temporary import surcharges employed mainly in the 1980s, and different weights for observed and unobserved components than in the import demand equation. The rise in trade openness in the 1990s coincides with rising labour productivity in manufacturing in South Africa, influencing unit labour costs.

Many analysts believe that the use of inflation targeting will spread. This makes it even more important for economists to develop a better understanding of the reform-linked determinants of inflation in the short and medium run. Failing to do so may lead central bank modelers to forecast from misspecified models that omit the structural breaks of past trade liberalisation, and may lead to the choice of inappropriate monetary policy.

 

 


 

Footnotes

1 Aron, Janine and John Muellbauer, “Interest rate effects on output: evidence from a GDP forecasting model for South Africa.” IMF Staff Papers 49 (November, IMF Annual Research Conference): 185-213, 2002.
2 Edwards, Sebastian, “Openness, productivity and growth: what do we really know?” Economic Journal, 108:447 (1998), 383– 398; Harrison, Ann, “Openness and Growth: a Time Series, Cross-country Analysis for Developing Countries,” Journal of Development Economics, 48:2 (1996), 419– 47; Pritchett, Lant, “Measuring Outward Orientations in Developing Countries: Can It Be Done?” Journal of Development Economics 49:2 (1996), 307-35; Rodrik, Dani and Francisco Rodríguez, (2001), “Trade Policy and Economic Growth: A Skeptics Guide to the Cross-National Evidence” (pp. 261–324), in Ben Bernanke and Kenneth Rogoff (Eds.), Macroeconomics Annual 2000. (Cambridge, MA: MIT Press); and Rose, Andrew, “Do WTO members have a more liberal trade policy?” Journal of International Economics, 63:2 (2004), 209-235.
3 Frankel, Jeffrey A. and Eduardo A. Cavallo, "Does Openness to Trade Make Countries More Vulnerable to Sudden Stops, or Less? Using Gravity to Establish Causality," NBER Working Paper No. 10957, (December, 2004).
4 Rodrik and Rodriguez, 2001 (op. cit.).
5 Anderson, Jim and Peter Neary (Measuring the Restrictiveness of International Trade Policy, Cambridge, Mass.: MIT Press, 2005) introduce an index of trade policy restrictiveness defined as the uniform tariff which maintains the same trade volume as a given tariff/quota structure. Their index overcomes the problems of the trade-weighted average tariff. However, the data requirements for its construction are rather onerous.
6 Sachs, Jeffrey, and Andrew Warner, “Economic Reform and the Process of Global Integration,” Brookings Papers on Economic Activity 1:96 (1995), 1-118.
7 Dollar, David, "Outward-Oriented Developing Economies Really Do Grow More Rapidly: Evidence from 95 LDCs, 1976-85," Economic Development and Cultural Change, 40:3 (1992), 523-544.
8 Sarno, Lucio, “Towards a solution to the puzzles of exchange rate economics: where do we stand?” Canadian Journal of Economics 38:3 (2005), 673-708.
9 In a model that captures known influences on the import ratio, any unexplained variance (apart from white noise error) is then represented by the stochastic trend (estimated using Koopman Siem J., Andrew C. Harvey, Jurgen A. Doornik, and Neil Shephard, STAMP: Structural Time Series Analyser, Modeller and Predictor (London: Timberlake Consultants Press, 2000)).

4,829 Reads