Relative prices and inflation: new evidence from different inflationary contexts

Abstract

This paper analyses the relationship between inflation and relative price variability, in the direction of the latter, in two countries with very different inflationary experiences: Argentina and Spain. To address this objective, using disaggregated price indexes (the Wholesale Price Index for Argentina and the Consumer Price Index for Spain), we delimitate different inflationary regimes and compute a set of regressions for each country. Our results suggest evidence in favour of the non-neutrality of inflation (mostly in hyperinflation periods) and do not support either the menu costs or the signal extraction approaches. We also detect significant structural changes in the relationship depending on the inflationary regime.

Acknowledgements

The authors acknowledge the financial support from Centro de Estudios Andaluces (Project ECO 17), as well as the useful comments from the participants in the Latin American Meeting of the Econometric Society (2004), members of the Macroeconomics Group (2004) of Centro de Estudios Andaluces, Juan F. Jimeno and anonymous referees of Centro de Estudios Andaluces (Working Paper E2004/71) and Applied Economics. The remaining errors are the authors’.

Notes

¹ Likewise, an important strand of the literature focuses on intramarket RPV. This variable can be defined as the standard deviation of relative price changes of a given product across stores around the average inflation rate of that product. A number of authors have found evidence supporting a positive correlation between intramarket RPV and inflation: Domberger (1987) for the United Kingdom, Lach and Tsiddon (1992) for Israel, Amano and Macklem (1997) for Canada, and Parsley (1996) for some cities of United States (US). However, when economies are experiencing very high inflation rates, intramarket RPV can decrease when inflation increases, as has been shown by Dazinger (1987) and Van Hoomissen (1988) for Israel, Tommasi (1993) for Argentina and Caglayan and Filiztekin (2003) for Turkey.

² Nevertheless, Fischer (1981, 1982) and Bomberger and Makinen (1993) assert that the relationship between inflation and RPV found for the US is dominated by energy and food price shocks.

³ In fact, the key factor is to determine how many months ‘around’ the current inflation make a homogeneous period of inflation. As has been said, for Argentina we select only three months, because given the great changes observed in its inflation rate, adding more lags could lead us to include months belonging to different inflationary contexts, and therefore there would be a risk of overestimating this measure. For Spain this problem is not so important, but as the results of the estimations are similar with three or more lags, we have chosen the same number of lags as for Argentina in order to compare the results.

⁴ For Argentina a slight variation of Equation 4 is used, because, as was stated in Dabús (1993), Equation 4 is not the best measure for RPV in high inflation economies. In this context, the estimation of the coefficient of variation of the price change distribution is required, instead of the simple variance, because at high inflation the latter is spuriously correlated with the mean of the distribution –the inflation rate. To avoid this problem, RPV for Argentina is defined as follows:

⁵ For a more detailed explanation of this methodology see Dabús (1993). An alternative approach to delimitate inflationary regimes could be based on changes in the variance of the inflation series, for example using the ICSS algorithm developed by Inclan and Tiao (1994) and Sansó et al . (2004). In further research we intend to compare both approaches for data from several countries.

⁶ Given the size of each sub-period for Argentina, it is not possible to develop the analysis for each one.

⁷ The degree of disaggregation can affect the estimations; therefore, in order to compare the results for both countries, homogeneity in the degree of disaggregation is required, otherwise different results can be obtained due just to the different kind of data used. Another distortion can be introduced in the results by the fact that we are using two different price indexes. On the one hand, this problem is not avoidable because the same price index, with a similar degree of disaggregation, is not available for both countries. On the other hand, Caraballo and Usabiaga (2004b) carry out a similar study for Spain using two price indexes, Producer Price Index (PPI) and Consumer Price Index (CPI), with a similar degree of disaggregation (25 categories for PPI and 33 categories for CPI), and there are not remarkable changes in the estimations. Moreover, the estimations using CPI with different degrees of disaggregation yield more relevant changes. Taking these arguments into account, it seems to be more appropriate to use different price indexes with a similar degree of disaggregation.

⁸ For example, if two prices are always equal, and every month one of those prices is sampled the first day and the other the last day, the true variability of relative prices is zero. At low inflation a low RPV should be detected, but at high inflation a higher variability will be detected, which would be just the consequence of the periodicity of price collection.

⁹ Data include 77 industrial and imported good prices, from a total of 87, for the 1960–1984 period, and 55, from a total of 64, for the 1984–1991 period.

¹⁰ In other countries different results have been obtained. For instance, for Turkey, during the 1948–1997 period, Caglayan and Filiztekin (2003) find a lower effect of inflation on relative prices during the higher inflationary period.

¹¹ In order to select the number of lags, the Akaike criterion has been applied.

¹² This result is observed especially in high inflation periods. It denotes that there are variables affecting RPV that have not been included in our equations, in particular real variables related to high economic volatility, like changes in real exchange rate and real wages – see, for example, Fischer (1981) and Dabús (1993) for further details.

¹³ As we stated in the introduction, this result has been obtained both for inflationary economies – see Logue and Willet (1976), Blejer (1979), Moura and Kadota (1982) and Dabús (2000), and stable economies – see Chang and Cheng (2000).

¹⁴ The same result was obtained for the Chinese hyperinflation (1946–1949) – see Tang and Wang (1993).

¹⁵ Recent papers show that sectoral factors have an important impact on RPV as well – see, for example, Nath (2004).

¹⁶ In general, a positive association inflation-skewness is supported by the data – see, for example, Ball and Mankiw (1995) for the US, Amano and Macklem (1997) for Canada, Hall and Yates (1998) for the United Kingdom, Aucremanne et al . (2002) for Belgium, Döpke and Pierdzioch (2003) for Germany, Assarsson (2004) for Sweden and Caraballo and Usabiaga (2004a, 2004b) for Spain.