Real Exchange Rates of Tradable and Non-tradable Goods in the European Monetary Union

The spatial spread of inflation and its consequences for real exchange rates (RER) between the members states of the European Monetary Union (EMU) is a topic of continuing empirical research. Lacking inflation convergence can cause serious problems in a monetary union.

On one hand, diverging inflation rates between the member states of a currency union can lead to diverging real interest rates, if monetary policy causes a convergence of nominal interest rates. According to “Walter’s Critique” a positive feedback loop might emerge, where high (low) inflation countries might experience low (high) real interest rates, what in turn favors spending (saving), what in turn causes again high (low) inflation rates. This way, a self-reinforcing divergence process might result.

On the other hand, price level divergence does not only affect the competitiveness of the prices for goods and services; it also affects the competitiveness of wages and salaries via its impact on unit labor costs. If real wages and the wage share in GDP shall stay constant, nominal wages must, ceteris paribus, follow the price level. This however implies an increase of unit labor costs and a loss of wage competitiveness compared to countries with lower price level growth. As a consequence, the competitiveness in the tradable goods sector of high-inflation (low-inflation) countries erodes (improves). This can result in a deindustrialization (reindustrialization) process in high-inflation (low-inflation) countries.

Given these potential problems caused by price level divergence, this study empirically analyzes the stationarity of the RERs of tradable and non-tradable goods and services between the 12 EMU founder states. The analysis is based on single country pairwise unit root and stationarity tests. Since the results indicate that RERs – even for tradable goods – are typically not stationary, cointegration tests are additionally applied in order to allow the coefficients of the RER components to deviate from unity.

Figure 1 shows the all-items HICPs of the founding member states of the EMU, which equals the HICP aggregate used by the European Central Bank (ECB) to calculate the official inflation rates of the EMU.

Figure 1 – PPP-corrected All-Items HICPs of EMU-founder States

The PPP-corrections are calculated in the following way: The original time series is rebased such that the month June 1999 equals 100%. The whole time series of a specific country is then multiplied with the ratio of the PPP-value of the specific country for the year 1999 to the average PPP-value of all 12 EMU countries in the year 1999. In this way, the PPP-information about the percentage difference of the price level of a country compared to the country average is transferred to every country-specific HICP time series.

The diagram shows that in the starting year of the EMU a maximum difference of roughly 40% existed between the price levels between Portugal and Finland. By the end of the sample period a maximum price level of roughly 60 percent existed between Portugal and Luxemburg. As the coefficient of variation shows, from the beginning of the year 1999 to the year 2010 the price levels of the member states had been converging. Since the year 2010 this trend has reversed and the price levels diverge. Since sigma-convergence (-divergence) implies also beta-convergence (-divergence) this means that over the first ten years of the EMU countries with a lower than average price levels, as Portugal Greece and Spain experienced higher than average inflation rates, while countries with a higher price levels like Germany, France, Austria, Belgium and Finland experienced lower than average inflation rates, as figure 2 shows. Exceptions like Ireland and Luxemburg are too few in number to change the overall trend. As a result, figure 2 displays a negative relationship between price level and inflation rate.

Figure 2 – Convergence and Divergence of All-Items HICPs of EMU-founder States

Legend Figure 2: The horizontal lines mark the average inflation rate over the 12 EMU-founder states over the corresponding time range, January 1999 – January 2010 and January 2010 – September 2019. The vertical lines mark the average HICP level rate over all EMU-founder states in the corresponding starting period, January 1999 and January 2010. Source: EUROSTAT

As revealed by the second diagram of Figure 2 from the beginning of the year 2010 until the presence countries like Portugal Greece, Spain and Italy with lower than average price levels have experienced lower than average inflation rates, while countries with higher than average price levels like Austria, Belgium, Finland and Luxemburg have experienced higher than average inflation rates. The only significant exception here is Ireland. Germany, Netherland and France are too close to the intersection point of the average inflation rate and the average price level curve to change the divergence trend. As a result, figure 2 displays a positive relationship between price level and inflation rate. These data are compatible with the hypothesis that the European debt crisis and/or the following policy response has caused a divergence of the price levels.

Theoretical intuition might tell that the divergence of the overall HICP price level has been mainly caused by the fact that a large part of HICP components are service-sector goods and as such non-tradables. A closer look to the disaggregated components of the HICP shows (Figure 3), however, that this intuition is wrong. While the coefficient of variation is roughly half as large for the total goods HICP component as for the total service component, a similar trend reversal is recognizable for both components. Consequently, the component of the HICP, which should mostly include tradeable goods is also not characterized by a long-run convergence of prices driven by arbitrage.

Figure 3 – PPP-corrected HICP components of EMU-founder States

Figure 3 also reveals, that reducing the aggregation level further to the 3-digit level does not necessarily lead to more convergence. The HICP component “clothing” includes such items as clothing materials, garments, accessories and repair of clothing. Clothing is typically produced outside of Europe in emerging market countries. So the purchase prices for clothing should be similar for all European countries. Nevertheless, the retail prices as measured by the HICP differ widely: Prices in Ireland are roughly 80 percent below the price levels in Luxemburg or Greece at the end of the sample period. Omitting “outliers”, such as Ireland, Portugal and Greece, significantly reduces the level and the slope of the coefficient of variation, but does not eliminate its overall positive trend.

The HICP component “non-alcoholic beverages” includes such items as fruit and vegetable juices, coffee, tea, cocoa drinks, water, and soft drinks. They should be typically more tradeable as more perishable food items and less suffer from different tax regimes as “alcoholic beverages”. Nevertheless, large differences in price levels continue to persist as Figure 3 shows. Even omitting the “outlier” Finland leaves a price difference between Spain and Austria of roughly 40 percent. As Appendix Figure 2 shows, the coefficient of variation without Finland has been roughly constant. Until the year 2005 the coefficient increased, followed by a decline until the year 2014. Since then it is slightly growing again.

Items like cloth and non-alcoholic beverages are typically distributed by local retail chains, which in many European countries form national oligopolies, which exert monopolistic market power. This could be an explanation for the observed persisting price divergence.

Motor cars are a HICP component, which are typically distributed by dealership networks franchised by car producers. However, a so called “gray market” exists also, where non-franchised dealers try to profit from international price differences. On the 3-digit level the HICP component “motor cars” includes “new motor cars” as well as “second-hand motor cars”. Since the second-hand market is less controlled by car producers, more competition and arbitrage activities should be possible. However, as the corresponding diagram of Figure 2 shows, price levels for motor cars still widely differ across the founding EMU-member states and the coefficient of variation has returned to its initial level after a decrease that took place until the year 2013.

The HICP database of the statistical office of the European Union (Eurostat (2020a)) offers also data for “non-energy industrial durable goods”, which are strictly speaking not a HICP component. This aggregate includes machinery, manufacturing plants and raw materials with exception of energy used by industries or firms. These goods are typically also not sold by local retail chains but by producers themselves or international intermediaries. Hence the market structure should be quite different from that of “clothing” or “non-alcoholic beverages”. However, as the corresponding diagram in Figure 3 shows, the divergence of price levels is strong and has been significantly growing after a period of slight reduction that ended in the year 2013.

As the last two diagrams of Figure 3 display, the service sector components of the HICP are quite heterogenous. Despite the fact that services of restaurants and hotels are locally produced, labor-intensive and not internationally tradeable, the dispersion of prices of restaurant and hotel services is significantly smaller than for “total services”. One explanation might be that restaurant and hotel services offered at different tourist locations are, from the perspective of consumers, relatively close substitutes. New price comparison tools, like online booking apps, may have helped to create a relative competitive international market for such services.

The opposite situation seems to be prevalent on the market for “communication” services. Even though modern communication services rely on the same capital-intensive technology, the prices for these services differ largely between the EMU-founding member states and the divergence has been steadily growing over the sample period. This indicates that local markets in some countries might be dominated by quite narrow national oligopolies, which exert a strong market power.

Real Exchange Rate Unit Root Tests

To test the unit root hypothesis  against the stationarity hypothesis  for the 12 founding member states of the EMU, 66 country-pairs of RERs are calculated (66 = (11*12)/2). This offers also the possibility to check for potential “price level convergence-clubs”. The detailed results are displayed in the appendix tables 28 – 37. The tests are based on PPP-adjusted and unadjusted HICP data. The results show that the regression constant  fully absorbs the difference between the level of PPP-adjusted and unadjusted data. The test statistics do not differ up to the 5th decimal place. In no case the test results were affected. Table 2 summarizes the test results.

Table 2 – Summary of Unit Root Tests of the Levels of the RERs

Legend: This table summarizes the results with a rejection of the Random Walk Hypothesis at the 5% significance level of the Augmented Dickey-Fuller test (ADF) and the Phillips-Perron test (PP), the Kwiatkowski–Phillips–Schmidt–Shin tests (KPSS) and the Zivot-Andrews test (ZA). The detailed results are presented by appendix tables 2 to 28.

If a common rejection of the random walk hypothesis by the DFGLS (H0) and KPSS (H1) is applied as criterium, the random walk hypothesis can be rejected only in 4 cases out of 594 total cases: For the all-items RER of Germany/Belgium and Luxemburg/Italy and the communication services RER of Spain / Germany and Luxembourg / Greece. It is interesting that the random walk hypothesis for the RER of tradeable goods is not systematically more often rejected that for non-tradable services. It is also not possible to detect any kind of stationarity clusters between structurally similar countries.

Real Exchange Rate Cointegration Tests

Appendix tables 40 – 48 display the results for Johansen Cointegration tests for the 66 possible combinations of every one of the nine HICP components as well as a couple of additional estimation statistics. Table 4 provides a summary for the estimated rank of  at a significance level of 5%. As to be expected, the hypothesis of a self-stabilizing relationship between the pairs of prices is less often rejected as for the stricter unit root tests reported in table 2. However, the results indicate again, that there is not systematic difference between HICP components of tradeable goods and non-tradeable services. Now out of 594 possible cointegration relationships 112 relationships do the reject the H0 of the existence of one cointegration vector at a significance level of 5%. However only in 40 cases the point estimates of the parameters of equation (8) display the correct signs. The number of cointegration vectors for “total goods” is the same as for “total services”, even though the corresponding country pairs are different. For “communication services” the H0 of the existence of one cointegration vector cannot be rejected in 22 cases. The corresponding number for “motor cars” is 11, for “clothing” 14, for “non-alcoholic beverages” 12 and for “industrial durables” 15.

Table 4 – Summary of Johansen Cointegration Tests of the HICP Levels

Legend: This table summarizes the results of the JC tests of appendix tables 40 – 48. The significance level for the rejection of the H0 is 5%. If the H0 that one cointegration vector exists cannot be rejected the column “All” informs about all the cases and the column “Meaningful” informs about cases where the point estimates of the parameters have the correct signs for a cointegration relationship.

As the following table 5 shows, it is not possible to detect specific country-clusters, where the existence of cointegrating relationships can less often be rejected than for the rest. The country with the most cointegrating vectors is Germany (29 of 99 possible), Austria holds the second place with (25 of 99 possible). The country with the lowest cointegrating vectors in Finland (11 of 99 possible). The country pair with the largest number of cointegrating vectors is Netherlands/Germany (6 of 9 possible), on second place are the country pairs Ireland/Austria (5 of 9 possible) and Portugal/Greece (5 of 9 possible). A specific clustering around “northern” or “southern” European countries is not recognizable. It is also interesting to note, between what country pairs no cointegration relationships exist: the Netherlands and Luxemburg and Belgium and Luxemburg. Between the Netherlands and Belgium only 1 cointegration vector with wrong parameter signs can be found for “total goods”. This is astonishing, since the three countries form a customs and currency union since 1960 (Benelux-Treaty) and should thus be economically integrated.

Table 5 – Summary of Johansen Cointegration Tests of the HICP Levels

Conclusions

The empirical results presented in this paper show that the price levels of the Eurostat HICPs and 8 of its components for the founding member states of the European Monetary Union are typically not stationary around a linear trend but random walks. The country-pairs of the resulting RERs are typically also random walks and the RER components are most of the time not cointegrated. A couple of exceptions are found, but these exceptions are always related to different country pairs and do not allow the identification of a cluster of countries with mutually stationary RERs. Somewhat surprising is that the results for tradable goods do in general not display more often stationary RERs than the results for non-tradable services.

Since most trade between the EMU founding member states is intra-industrial trade, a straightforward explanation for the observed price differences for tradable goods could be monopolistic price discrimination. If this hypothesis is true, typical price discrimination factors like per-capita income differences should be able to explain, why the observed prices for tradable goods are not cointegrated. This could be one strategy for future research.

But whatever the explanation for the observed divergence of prices is, the problems for monetary policy in the EMU will remain. Under the observed empirical conditions, it cannot be expected that market forces spread the target inflation rate of the ECB evenly across the member states. This questions the European Central’s Bank’s principle of a “single monetary policy“. As mentioned in the introduction diverging inflation rates might cause self-enforcing inflationary spirals with a corresponding worsening of the competitiveness of a country.

If price arbitrage is not able to keep the development of price level together, it might be justified to consider more differentiated country-specific monetary policies than the principle of a “single monetary policy” allows. Such policies could for example include country-specific minimum reserve requirements. By Article 19.1. of the ECB-Statute, the European Central Bank has the full legal entitlement to set the minimum reserve rates. Another possibility to implement country-specific monetary policies could be the implementation of country-specific main refinancing rates. Such a regime had already been practiced by the United States’ Federal Reserve System from 1914 to 1941, when discount rates were set district by district.

Appendix Tables

Kommentar verfassen

Trage deine Daten unten ein oder klicke ein Icon um dich einzuloggen:

WordPress.com-Logo

Du kommentierst mit Deinem WordPress.com-Konto. Abmelden /  Ändern )

Google Foto

Du kommentierst mit Deinem Google-Konto. Abmelden /  Ändern )

Twitter-Bild

Du kommentierst mit Deinem Twitter-Konto. Abmelden /  Ändern )

Facebook-Foto

Du kommentierst mit Deinem Facebook-Konto. Abmelden /  Ändern )

Verbinde mit %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.