Pricing-to-market and the volatility of UK export prices

Abstract

This empirical study examines the Pricing-To-Market (PTM) behaviour of 20 UK export sectors. Using both Exponential General Autoregressive Conditional Heteroscedasticity (EGARCH) and Threshold GARCH (TGARCH) estimation methods, we find evidence of PTM that is accompanied by strong conditional volatility and weak asymmetry effects. The PTM estimates suggest that when the currency of exporters appreciates in the current period, exporters pass-on between 31% and 94% of the Foreign Exchange (FX) rate increase to importers. However, both export price changes and producers’ prices are sluggish, perhaps being driven by coordination failure and menu driven costs, amongst others. Furthermore, export prices contain strong time varying effects which impact on PTM strategy. Exporters do not typically appear to put much more weight on negative news of (say) an FX rate appreciation compared to positive news of an FX rate depreciation. Much depends on the export sector.

Notes

¹ A response to a permanent FX rate change can take the form of changes in export prices, following an expansion or contraction of production and distribution, whilst a response to temporary FX rate changes can take the form of a temporary reduction in export prices, which in turn will reduce profit margin.

² Otani et al. (2006, p. 71) find support for time variation in the ERPT coefficients of Japanese imports. Baldwin (1988) also finds instability in the ERPT coefficients of US import prices.

³ Prior empirical work tends to concentrate on the imports/exports of other major industrial countries, such as Germany, Japan and the US (see, e.g. Marston, 1990; Knetter, 1993).

⁴ The financial accounting literature suggests that firms seek to smooth their income to bring about stability on reported results (see, e.g. Beattie et al., 1994). This literature is, however, more concerned with accounting manipulation within the structure of General Accepted Accounting Principles (GAAP).

⁵ Both Giovannini (1988) and Marston (1990) separate the FX rate effects into expected and unexpected. They conclude that firms deliberately PTM even after allowing for unexpected FX rate changes.

⁶ Menon (1995) notes that differences in the choice of data and the methodology employed give rise to substantial differences in the ERPT estimates of prior studies. He reviews 43 studies on ERPT and shows that almost two-thirds use the standard OLS method.

⁷ Both the Osborn et al. (1988) and Augmented Dickey–Fuller (ADF) statistics were used to test for unit roots, respectively, at seasonal and zero frequencies. The results indicate that the univariate series contain a unit root such that the first difference of each univariate series is stationary. These and other results not fully presented are available on request.

⁸ The finding of a strong conditional volatility also suggests that the use of the standard OLS is likely to generate inefficient parameter estimates. Even so, GARCH effects will not be strong in studies like those of Kanas (1997), Gil-Pareja (2001) and Parsons and Sato (2008) that estimate PTM using quarterly data.

⁹ These studies also show that the degree of PTM varies across countries. Hung et al. (1993), for example, show that Japanese and German exporters tend to use incomplete pass-through pricing strategy, such that export prices in their foreign markets are not substantially affected by exchange rate changes. This contrasts with the behaviour of US exporters. Also, Feenstra et al. (1996) show that French auto exporters tend to PTM much more than German, US, Norwegian and Swedish auto exporters. Differences in the pricing strategy can reflect variation in competitive strength across countries as well as the characteristics of the particular export sector (see also Gagnon and Knetter, 1995).

¹⁰ The difference accounts for the export and import pressures due to the export of goods from the UK to various destinations and the import of similar goods into the UK. Clearly, the competitive strength of UK exporters in foreign markets will depend on the ability of foreign firms to satisfy their own local markets. The reverse is also true. can be thought of as an error-correction mechanism popular in economic literature. This interpretation is useful since we find that and are cointegrated for 75% of the pairwise export/import sectors using the Johansen cointegration tests. Specifically, we did not find the pairwise export and import series for the Metal Ores, Iron and Steel, Paper and Paperboard, Electronic Machinery and Machinery sectors are cointegrated. Where cointegration is found, Equation 1 can be considered to be a restricted version of an error-correction model.

¹¹ Tange (1997) uses up to six lags of FX rate changes to capture FX rate elasticity.

¹² Doornik and Ooms (2003) show that dummy variables are best placed in the variance equation since to include them as regressors can lead to multimodality.

¹³ We use six lags of the explanatory variables, even if the AIC indicates that a VAR in differences based on Equation 1 require no more than two lags. This approach enables us to incorporate as much information as possible in the empirical model for estimation. is, however, allowed one lag as we treat this variable as a measure of the long-run adjustment between the levels of exports and imports.

¹⁴ Hall et al. (2000, p. 432) also indicate that ‘… time-dependent pricing was more common than state-dependent pricing, with 79% of the respondents reporting that they reviewed their prices at a specific frequency.’ The study does not discriminate between domestic and export prices. However, as export markets are likely to be more competitive than domestic markets, the price changes for export markets are likely to be more frequent than twice yearly.

¹⁵ Dias et al. (2007) shows that the frequency of price changes across countries for the same sectors are not too different.

¹⁶ Recall that Kasa's (1992) model relies on an exporter reacting to permanent FX rate changes while also absorbing the effects of temporary FX rate changes. We capture a permanent component in respect of conditional volatility although the sign of the coefficient depends on the estimation method.