Journal Issue

Germany: Selected Issues

International Monetary Fund
Published Date:
November 2004
  • ShareShare
Show Summary Details

V. Does Purchasing Power Parity Hold in the Long Run? Evidence from Germany and Switzerland63

A. Introduction

147. This chapter examines whether purchasing power parity (PPP) holds over the long run in Germany and Switzerland. If this were the case, the equilibrium real exchange rate would be relatively constant, and fluctuations of the actual real exchange rate would reverse over time. If PPP does not hold, the equilibrium real exchange rate could display drift, and assessing misalignments would depend on estimating the equilibrium drift.

148. Consensus opinion on PPP has shifted over time, as illustrated in Lothian and Taylor (1996). Studies during the Bretton Woods system of fixed parities supported a fixed real exchange rate over the long run. The monetary approach to the exchange rate even claimed PPP in the short run. However, the transition to floating currencies led to large and persistent fluctuations in real exchange rates, and new studies postulated that real exchange rates followed random walks and might not have an anchor. However, theoretical “overshooting” models (Dornbusch, 1976) were developed to help explain the longer-term deviations, and recently, the pendulum has begun to swing back once more. A number of recent studies claim to find PPP in very long data series, arguing that the time horizon is crucial (Lothian and Taylor, 1996, Glen, 1992, Abuaf and Jorion, 1990). They point out that statistical tests for stationarity are biased against PPP, since they have low power over short horizons.

149. This chapter finds that in the long run, PPP holds in Germany but not in Switzerland. It presents 115 years of data from Germany and Switzerland, two countries with very similar economic structures. While the German real exchange rate seems to exhibit PPP, Switzerland witnessed almost a century of real appreciation. After a closer look at the time series and their properties, we test some possible explanations based on different saving rates in different countries, inflation differentials, and relative market power.

B. German and Swiss Real Exchange Rates—Some Stylized Facts

150. Visual inspection of long-term time series suggests that PPP holds in Germany, but not in Switzerland. Figures V-1 and V-2 show the real effective exchange rate (REER) and an estimate of the equilibrium REER (discussed below) for Germany and Switzerland, respectively. The German REER has moved in a fairly narrow band (except for the war periods), but the Swiss franc has been on an upward trend since around 1905. On average, the Swiss franc has appreciated by around 1½ percent per year in real effective terms during the last century.

Figure V-1.Germany: Real Effective Exchange Rate

Figure V-2.Switzerland: Real Effective Exchange Rate

151. To explore these findings more formally, this chapter will use multilateral (or trade-weighted, real effective) exchange rates, in comparison with the literature which often only focuses on bilateral real exchange rates. Multilateral exchange rates allow a better assessment of PPP, as they consider actual trade patterns. The time horizon extends back to 1890, the earliest year for which there are sufficient data. Five countries are included as trading partners, accounting for about 80 percent of total foreign trade over the entire period. In the case of Germany, the trading partners are the U.S., France, the UK, the Netherlands and Italy. For Switzerland, the main trading partners are Germany, France, Italy, UK and the U.S. As partner country weights were fairly stable, Laspeyres indices of the real effective exchange rates were constructed (the Appendix presents a note on the data).

C. Time-Series Properties

152. The persistence of REER fluctuations is moderate in Germany, but high in Switzerland. Figures V-3 and V-4 show the autocorrelation functions for Germany and Switzerland, measuring the persistence of real exchange rate fluctuations. In Germany, the lagged autocorrelation coefficients quickly become statistically insignificant; in Switzerland, they become insignificant only after 20 years. This is a strong indication for a unit root, i.e., that the REER displays drift.

Figure V-3.Germany: Autocorrelation

Figure V-4.Switzerland: Autocorrelation

153. More formal tests confirm that Germany’s REER reverts to a mean, while Switzerland’s does not. Figure V-5 plots the results of augmented Dickey-Fuller tests on unit root behavior for different sample lengths. From right to left, the length of the time sample increases. On the right border, data begin in 1990, on the left border, they begin in 1890. German data are consistent with PPP (reject the unit root hypothesis at the 5 percent level) in 14 out of 21 samples, but Swiss data only in 2. The figure also questions the claim that unit root tests yield different results over long time horizons. German data are consistent with PPP over short samples as well as very long ones, but less so in intermediate ones. The Swiss data reject PPP over the whole sample. This finding is in line with Engel (2000), who cautions against making the sample period the main criterion in testing for a unit root.

Figure V-5.Unit Root Tests

D. Why PPP May Not Hold in the Long Run

154. Purchasing power parity holds only under some restrictive assumptions. In particular, it relies on the assumption of one representative good in the economy. Trade and arbitrage will then equalize prices in all countries. A more realistic setting, however, recognizes differentiated goods. In this case, PPP does not need to hold. With two goods, one tradable (T) and one non-tradable (N), consumer prices are a weighted average of two prices PT and PN:

An asterisk (*) denotes prices of trading partners. If lower-case letters denote logs, the real exchange rate r is then defined as the price difference between the home and foreign country, corrected by the nominal exchange rate e:

r = cpicpi* – e

Substituting the consumer price indices, this expression becomes

155. Thus the equilibrium real exchange rate can vary with changes in the relative price of non-tradables. In a two-good setting, the real exchange rate can vary because of (1) deviations from the law of one price for tradables, and (2) changes in the relative price of non-tradables across countries, known as the “Engel decomposition” (Engel, 1999). Only by coincidence will the relative price of non-tradables be equal in different countries. Burstein, Neves and Rebelo (2003) argue that including distribution costs and commercialization margins, the weight of non-tradables in the CPI is as high as 85 percent, which suggests ample scope for deviations from PPP.

156. considering this perspective, the evidence for PPP in long data series for Germany is quite striking. Parallel changes in non-tradable prices require similar technology and preference parameters among trading partners, and also similar shocks to the economy. Germany indeed shared with its main trading partners some large shocks to the economy, e.g., a history of war destructions, reconstruction booms, oil price increases and economic integration.

E. Equilibrium Real Exchange Rates: Saving, Competition, and Inflation

157. The long-run appreciation of the Swiss franc appears to confirm that equilibrium real exchange rates can drift. In the Swiss case, the upward drift even coexisted with sustained and large current account surpluses suggesting that it is not a misalignment. One possible explanation that is sometimes offered is the “Balassa-Samuelson” effect, which arises if productivity growth is faster in the tradable sector than in non-tradables. Many studies on this effect were done for transition economies. Mihaljek and Klau (2004) survey them and conclude that the results are disappointing. Cross-country productivity trends turn out to be quite similar, possibly because the diffusion of technology is fast. The same conclusion emerges from the German and Swiss data that were used for this chapter: the labor productivity effects were found to be statistically insignificant.

158. An alternative explanation points out that Swiss household saving rates have been very high compared to other countries. This approach suggests a closer look at preferences. A key parameter is the propensity to save, or the rate of time preference. Since there are no long time series on national savings, Isard and Farruquee (1998) propose an indirect measure: private savings are proxied by the share of the population at work. People at work generate the income that yields national saving, while all others dissave. Thus, their share in the population could be expected to correlate with national saving. Figure V-6 shows employment ratios for Germany and Switzerland, cast in relative terms, i.e., compared to trading partners. By this proxy, Switzerland is seen to have higher employment—and hence saving—ratios than their trading partners. Household saving ratios in Germany also were relatively high, but much less so than in Switzerland.

Figure V-6.Relative Employment Ratio

159. Swiss public sector saving was also high, but Germany recorded some large deficits. Public saving was proxied by relative fiscal balances (Figure V-7). During the world wars and hyperinflation, Germany had large deficits relative to its trading partners. Switzerland posted relative surpluses most of the time, in part because it stayed out of wars. On balance, public deficits absorbed private savings in Germany, while public surpluses augmented private saving in Switzerland.

Figure V-7.Relative Fiscal Surplus

(In percent of GDP)

160. High relative saving can lead to a trend real appreciation. High saving fosters capital accumulation and favors the production of traded goods, which are more capital intensive. Rising wages in this sector attract labor away from non-tradables, where production declines. With unchanged demand, a declining supply of non-tradables leads to an increase in their relative price. The real exchange rate appreciates.

161. High saving and the expansion of tradables sectors generate current account surpluses. Switzerland has experienced persistent and increasing current account surpluses in the second half of the 20th century, rising to 15 percent of GDP. The corresponding capital exports led to the accumulation of net foreign assets of close to 150 percent of GDP (Figure V-8). In contrast, total savings in Germany broadly matched that of its trading partners and did not affect relative prices. The current account fluctuated around balance, and net foreign asset reached at most 10 percent of GDP.

Figure V-8.Net Foreign Assets

(In percent of GDP)

162. Swiss markets were more monopolistic than Germany’s, keeping prices high. Market power differentials on domestic product markets may also affect the real exchange rate. While Germany has always been at the forefront of European economic integration, Switzerland mostly stayed out. Increased competition and strong antitrust regulation led to a decline in monopolistic pricing power in Germany and the EU, but not in Switzerland. This further restricted the supply of Swiss non-tradables and raised their relative price. Figure V-9 shows a proxy for the market power of Swiss firms, the ratio of consumer to producer prices. Compared to trading partners, this measure has increased by about 1/3 since the end of the second world war, while German markups remained flat.

Figure V-9.Switzerland: Relative Markup

(2003 = 100)

163. Finally, inflation differentials may explain the large “spikes” in the real effective exchange rates during the 1920s. During the German hyperinflation the real effective exchange rate depreciated sharply, while it appreciated in Switzerland. It is known from recent experiences in developing countries that high inflation depresses the real exchange rate. An example is Mexico (Figure V-11), and Braumann (2000) confirms this pattern in a wider sample of 23 inflation episodes. Calvo and Végh (1993) argue that high inflation leads to a temporary contraction of demand. Real money balances fall as inflation increases. This makes transactions more cumbersome and expensive, and households postpone consumption. While the resulting excess supply of tradables can be exported, the excess supply of non-tradables cannot. This leads to a fall in their relative price, and to a real depreciation. Switzerland experienced the flipside of this pattern as the German hyperinflation caused its currency to appreciate.

Figure V-10.Mexico: Inflation and Real Exchange

164. Changes in these key parameters can be used to derive empirically the Fundamental Equilibrium Exchange Rate (FEER) (e.g., Hinkle and Montiel, 1999). This method allows the long-run real exchange rate to depart from a flat PPP line if fundamental factors change. The CGER macrobalance method developed at the IMF by Isard and Farruquee (1996) is closely related. Table V-1 present the results of FEER estimations for Germany and Switzerland that use the above proxies for relative saving, competition, and inflation differentials as regressors. All coefficients have the expected sign and are statistically significant, with exception of net foreign assets in Germany. This is not surprising in light of their modest magnitude. The coefficients are then applied to 5-year moving averages of the regressors to derive the equilibrium real exchange rates in Figures V-1 and V-2.

Table V-1.FEER Estimation Results (OLS)
Dependent variable: Real effective exchange rate. T-values in brackets.
Employment ratio1.53*(2.42)0.78*(3.09)
Fiscal surplus0.86*(3.29)0.35*(2.73)
Net foreign assets-0.09(1.65)0.13*(7.26)
Inflation differential-6.73*(2.51)-14.90*(2.31)
Adj. R20.440.81

165. The FEER analysis suggests that Germany’s real effective exchange rate is close to equilibrium, while Switzerland’s has room to appreciate. These results are very similar to those obtained in the CGER macrobalance analyses, which use the current account instead of the real exchange rate as the endogenous variable. As discussed in chapter II.B of the 2002 Selected Issues Paper on Germany (IMF country report No. 02/240), Germany’s current account can be considered in equilibrium at a surplus of around 2 percent of GDP at this time. The actual surplus is somewhat higher, but taking into account cyclical slack would bring it close to the FEER. The same can be said for Switzerland, where estimates of the equilibrium current account are around 5 percent of GDP, as compared to over 10 percent actual value.

166. In sum, there is evidence that PPP held over the past century in Germany, but not in Switzerland. Germany’s fundamental production and preference parameters have stayed close to those of its main trading partners, keeping relative prices in line. In contrast, Switzerland’s high saving and its sheltered, monopolistic product markets have put constant pressure on the real exchange rate to appreciate. Calculations of the German fundamental equilibrium exchange rate suggest that the current REER is close to its equilibrium level. How to reconcile this external and quite competitive equilibrium with dormant domestic demand and low labor utilization is a challenge that requires further research.

APPENDIX Data Sources

167. Consistent data sources are crucial for long-run studies. The period after World War II is covered by standard databases from the OECD and the IMF (International Financial Statistics). Peacetime and some wartime data before this time were taken from Mitchell (2003, financial data and prices) and Maddison (1991, real variables). Gaps remained for real activity in Switzerland before 1930, and general macro information on the world wars for France and Germany, and the German hyperinflation. Specialized academic literature was used to fill those gaps: Gerlach and Gerlach (2002) for Swiss GDP, INSEE (1977) for French wartime data, Braun (1990), Hoffmann (1965), and Sommariva and Tullion (1987) for Germany. A particular problem was the reconstruction of public sector balances during the 1930-40s for Germany, as budget data were a state secret for most of these years and few records remain. Braun (1990) provides below-the-line financing estimates for this time.


Prepared by Benedikt Braumann.

Other Resources Citing This Publication