This Selected Issues paper and Statistical Appendix presents a number of studies designed to probe in more depth issues of fiscal adjustment, the public sector’s creditworthiness as judged by foreign lenders, and the banking system in Greece. The paper highlights that fiscal adjustment in Greece has followed a gradual path, relying to a significant extent on revenue enhancement and eschewing primary expenditure reduction. The paper also examines the evolution and tax determinants of the most widely followed indicator of the health and competitiveness of a banking system—the lending-deposit spread.

Abstract

This Selected Issues paper and Statistical Appendix presents a number of studies designed to probe in more depth issues of fiscal adjustment, the public sector’s creditworthiness as judged by foreign lenders, and the banking system in Greece. The paper highlights that fiscal adjustment in Greece has followed a gradual path, relying to a significant extent on revenue enhancement and eschewing primary expenditure reduction. The paper also examines the evolution and tax determinants of the most widely followed indicator of the health and competitiveness of a banking system—the lending-deposit spread.

III. The Contribution of the Balassa-Samuelson Effect to Inflation: Cross-Country Evidence17

A. Introduction and Summary of Results

48. Inflation in Greece is influenced in part by the catch-up of productivity in Greece’s tradable goods sector to the productivity levels of its main trading partners. As discussed in several recent staff reports (Finland, Ireland, and Spain), this “Balassa-Samuelson effect” comes about because countries that are experiencing higher productivity growth in the tradables sector than in nontradables will tend to have higher inflation rates for nontraded goods such as services, as nominal wage growth will tend to exceed productivity growth in the nontradables sector.18 In Greece, the highly centralized system of wage setting ensures that wage growth in nontradables largely keeps pace with that in tradables despite lower productivity growth. Overall inflation is a weighted average of inflation in the two sectors, so that the Balassa-Samuelson effects lead to higher inflation than would be the case were productivity growth even across sectors.

49. To the extent that Balassa-Samuelson effects account for the comparatively higher rate of inflation that is currently seen in Greece and in some smaller euro area countries such as Ireland and Portugal, this would help ease concerns that such higher inflation results from demand-side pressures related to these countries’ advanced cyclical positions. In a similar vein, the presence of Balassa-Samuelson effects potentially mitigates concerns about higher inflation giving rise to competitiveness problems. The inflation differential leads to an appreciation of the real exchange rate measured in terms of relative consumer price inflation, but this would not be considered a loss of external competitiveness to the extent that it stemmed from developments in nontradables.19

50. This chapter develops measures of relative prices, productivity, and wages for a number of countries in Europe over the period from 1960 to 1996. These are used to quantify the contribution of the Balassa-Samuelson effect to inflation, both through a simulation of the original model and through estimation of an extended version that allows for differential wage growth across sectors. The estimation results generally match the predictions of the theory, with a statistically significant long-run relationship found in most countries between sectoral price, wage, and productivity differentials. The Balassa-Samuelson effect is estimated as having added 0.5–2.5 percentage points to the annual inflation rate of most countries examined, with the largest contributions typically found in the poorer countries (those with the lowest per-capita GDP), reflecting their greater scope for productivity convergence. A 1 percentage point increase in the level of productivity in nontradables is found to lower the inflation rate in most countries considered by 0.3–0.4 percentage points for 2–5 years. This suggests that structural reforms that increase the level of productivity can have fairly prolonged effects on inflation.

51. In Greece, the Balassa-Samuelson effect is found to have contributed 1 percentage point of annual inflation on average over 1960–1996, an amount that accounts for nearly half of the average annual real appreciation of the drachma against the U.S. dollar over this 36 year period. Data for the most recent years available similarly point to a potentially important role of the Balassa-Samuelson effect: over 1990–96, nearly half of the real effective appreciation of the drachma against the ECU in terms of relative consumer prices can be accounted for by the contribution of this effect to higher inflation than in Greece’s partner countries in Europe—this portion can be considered as not representing a loss of external competitiveness. Given the scope for further convergence of productivity in tradables in Greece to the levels in the rest of Europe, the Balassa-Samuelson effect can be expected to remain a medium-term influence on inflation.

52. The next section presents the theoretical background and econometric model, after which Section C summarizes the data and provides initial evidence on the relationship between sectoral productivity and inflation. Section D discusses estimation results and results for the entire sample from 1960 to 1996, while Section E concludes by assessing the contribution of Balassa-Samuelson effects to inflation and real exchange rate changes in Greece.

B. Background and Methodology

53. This section provides a brief exposition of the Balassa-Samuelson effect to motivate the empirical work; it follows the presentation in Froot and Rogoff (1995) and Alberola-Ila and Tyrväinen (1998). The model consists of a small open economy which produces tradables (indicated by a subscript, T), and nontradables (indicated by a subscript, N) according to Cobb-Douglas production functions:

YT=ATLTθKT1θ(1)
YN=ANLNγKN1γ(2)

where Y, L, and K are the quantities of output, labor, and capital, A is level of total factor productivity (TFP), and the parameters θ and γ are the respective output elasticities of tradables and nontradables with respect to the quantity of labor. With perfect competition and profit maximization, the levels of wages and the interest rate can be obtained as the marginal products of labor and capital for each sector:

R=(1θ)AT(KT/LT)θR=P(1γ)AN(KT/LN)γ(3)

and

W=θAT(KT/LT)1θW=PγAN(KN/LN)1γ(4)

where P is the relative price of nontradables in terms of tradables, PN/PT (that is, PT is set to 1 without loss of generality). Similarly, the wage, W, and interest rate, R, are expressed in terms of tradables.

54. Capital is assumed to be perfectly mobile across countries, so that the interest rate is fixed by the world interest rate. Together with the assumption that labor is mobile across sectors, this ensures that the nominal wage is the same for both sectors, and is determined entirely by conditions in the tradables sector. This can be seen in equations (3) and (4): the exogenous interest rate, R, fixes the capital-labor ratio in tradables, KT/LT, by the first equality in (3), and this determines wages, W, in the first equality of (4). The capital-labor ratio and relative price of nontradables are then solved from the second equalities in (3) and (4).

55. The Balassa-Samuelson effect on inflation differentials comes about because with equal wage growth across sectors, higher productivity growth in tradables than in nontradables means that output prices in nontradables must increase more rapidly than in tradables to ensure that product wages in that sector remain equal to the marginal product of labor while nominal wages across sectors remain equal. The Balassa-Samuelson effect thus typically leads to faster growth of unit labor costs and prices in nontradables than in tradables.20 To see this, the expressions for the sectoral capital-labor ratios are substituted from the production functions (1) and (2) into equation (4), and this is then solved for the relative price, P. Log-differentiating gives the Balassa-Samuelson relationship in growth rates:

P.=P.NP.T=(γ/θ)a.Ta.N(5)

Lower case letters denote the log of the corresponding variable in upper case, and a “dot” indicates the change over time. Note that even if TFP growth is the same in tradables and nontradables, an inflation differential between the two sectors can result from differences in labor shares across sectors.

56. The contribution of the Balassa-Samuelson effect to overall inflation depends on both the sectoral inflation differential and the share of nontradables in the aggregate price index. This is because overall inflation, π, is an average of inflation in the two sectors, with the production shares of nontradables, σ, and tradables, 1-σ, as the weights:

π=σP.N+(1σ)P.T=P.T+σP.(6)

57. The extent to which the Balassa-Samuelson effect contributes to inflation differentials across countries depends on the relative sectoral inflation differentials. With (weak) purchasing power parity, tradables prices expressed in a common currency grow at the same rate in each country so that cross-country inflation differentials depend solely on differences in each country in the contribution of sectoral productivity differentials to domestic inflation, σ ṗ. Even without purchasing power parity—which, as discussed below, does not appear to hold in Europe—the contribution of the Balassa-Samuelson effect to inflation in a country can be compared to the average contribution for its trading partners. The difference between these (between the values of σ ṗ) can be viewed as the contribution of the Balassa-Samuelson effect to cross-country inflation differences, with variation in tradables inflation across countries and the failure of purchasing power parity attributed to factors beyond the scope of this chapter.

58. Previous work has focused principally on estimating versions of equation (5) by regressing relative prices on various measures of relative productivity. De Gregorio, Giovannini, and Wolf (1994) augment (5) with demand-side variables such as government spending and per-capita GDP, and find a significant relationship between changes in total factor productivity differentials and relative prices. De Gregorio, Giovannini, and Krueger (1994) provide similar evidence for European countries, but note that there are important differences between the behavior of labor productivity and total factor productivity. Alberola-Ila and Tyrväinen (1998) find co-integrating relationships between relative prices and labor productivity in several European countries, and between prices, productivity, and wages in others. They then use these relationships and the assumption of a common inflation rate in tradables to calculate the contribution of Balassa-Samuelson effects to cross-country inflation differentials. Canzoneri, Cumby, and Diba (1999) obtain a relationship between prices and labor productivity in a panel of advanced economies, but find that purchasing power parity does not hold for traded goods. Moschos and Stournaras (1998) find that purchasing power parity does not hold for prices in Greece. This suggests that taking a common inflation rate for tradables goods is suspect.

59. The approach in this chapter follows Alberola-Ila and Tyrväinen (1998) in augmenting (5) to allow for differential wage growth across sectors and then estimating a co-integrating relationship between relative prices, productivity, and wages. Sectoral wage differences are specified as offsetting a fraction, λ, of the inflationary impact of productivity differentials:

P.=P.NP.T=(γ/θ)a.Ta.Nλ(w.Tw.T)(7)

Equation (7) suggests the existence of a long-run relationship between inflation rates, wage growth, and productivity growth, and forms the basis of the empirical work. The econometric approach is to use the Johansen technique to test for the presence of a co-integrating relationship between the three differentials, and then estimate a vector autoregression (VAR) that includes this co-integrating relationship as an error-correction mechanism.

60. Rather than imposing the factor shares, θ/γ, as modifying the effect of tradables productivity, the specification simply examines the overall relationship between relative prices, p, productivity, a, and wages, w:21

ΔPt=ηpΔPt1+ηaΔat1+ηwΔwt1+αp(Pt1βaat1βwwt1)+εt(8)

The η coefficients on the variables in first differences correspond to the short-run effects, the β coefficients to the long-run equilibrium relationship (the coefficients of the co-integrating vector), and α to the rate at which prices adjust in response to deviations from equilibrium (ε is the error term in the estimation). Equation (8) is shown with 1 lag (in first differences), but the actual number of lags is determined in the estimation. The estimated equations for particular countries include constants, dummies, and trends depending on the characteristics of the country examined; these are discussed below. Alberola-Ila and Tyrväinen (1998) estimate a similar equation, but rather than including relative productivity, they restrict the coefficient on nontradables productivity to one as implied by (7), and estimate only the coefficient on tradables productivity. In contrast, the approach used here does not impose this restriction but instead looks at the overall effect of productivity differentials. Another important difference is that the estimation in this chapter uses total factor productivity as suggested by the theory rather than labor productivity.

61. Once the long-run relationship between prices, productivity, and wages is estimated, the β coefficients are used to calculate the predicted equilibrium inflation differential across sectors:

SectoralInflationDifferential=βa(a.Ta.N)avgβw(w.Tw.N)avg(9)

where (ȧT – ȧN)avg and (T – ẇN)avg are the average values of the TFP growth and wage growth differentials over the estimation period. The contribution to overall inflation is then the predicted sectoral inflation differential from (9) multiplied by the share of nontradables in production, σ; this is the amount by which inflation in a particular country is higher solely on account of differential productivity and wage growth across sectors. No assumption is made that purchasing power parity holds—tradables inflation is not assumed to be the same across countries and the contribution to inflation is calculated separately for each country.22

C. Data and Preliminary Evidence

62. Data for all countries but Greece and Portugal are from the OECD Intersectoral Database (ISDB), 1998 edition; for Greece and Portugal, data by sector are from the two countries’ national accounts.23 The tradables sector is comprised of manufactures and mining, while the nontradables sector includes all other sectors except agriculture.24 Agriculture is excluded because the web of subsidies and nonmarket arrangements in European agriculture are likely to distort the relationship between productivity and prices.25 This classification matches that in other papers such as De Gregorio, Giovannini, and Wolf (1994), with the exception that nontradables here include transport services, which are sometimes counted as traded in other work (the split here is meant to group together services). Government services are included in nontradables and account for more than 10 percent of GDP in all countries in Europe (and as much as 20 percent on average in Sweden). It must be noted, however, that this introduces a measurement issue, since government output is valued by the inputs rather than by the outputs as in other sectors (government-owned enterprises are, of course, counted as part of their respective industry and valued by their outputs).

63. Total factor productivity growth is calculated as growth of the residuals from a Cobb-Douglas production function, where the average share of wages out of each sector’s value-added is used to weight the growth of capital and labor inputs in the production function. Because the data include both the number of employees and total employment but wages only for employees, the average wage of employees is imputed to the self-employed to obtain the total share of wages in output. The use of production data means that the prices for tradables and nontradables are implicit price deflators, so that the measures of inflation are akin to GDP deflators rather than to the consumer price index. Overall (nonagricultural) inflation is calculated by using the value of production in each sector to calculate a weighted average of inflation in the two sectors.

64. Table 1 provides a summary of the data over the entire sample for each country. The data start in 1960 for some countries and no later than 1970 for others, and include at least the early 1990’s in each country, through 1996 in some countries. Inflation is higher on average in nontradables than in tradables in all countries but the Netherlands (where the two are nearly identical), while total factor productivity growth is higher in tradables in all countries. Average wage growth is higher in tradables than in nontradables in all countries but Portugal, though the gaps in wage growth are much smaller than the inflation or productivity differentials. This suggests that differential wage growth will at best attenuate but not completely offset Balassa-Samuelson effects stemming from productivity differentials. Finally, nontradables constitute by far the larger share of production in all countries, so that higher inflation in this sector than in tradables will have an important effect on overall inflation. As suggested by the theory, poorer countries such as Greece and Portugal—the countries with the lowest per-capita GDP of those examined—have among the highest rates of productivity growth in tradables reflecting their greater scope for productivity catch-up over the sample period.26 Greece is also less open than most of the other countries examined, with a smaller share of tradable goods, so that higher inflation in nontradables has a relatively large effect on overall inflation.

Table 1.

Data Summary

article image
Note: The rank of real GDP per-capita is constructed from the 1990 value of per-capita real GDP adjusted for changes in the terms of trade; the source is the Penn World Tables version 5.6.

65. Figure 1 depicts the inflation and productivity differentials for both the full sample (top) and for the average of the available years in the 1990s (bottom). The Balassa-Samuelson relationship, which suggests a positive relationship between sectoral productivity growth and inflation differentials, is apparent, with a correlation of 0.78 over 1960–96, and a correlation of 0.58 for the period in the 1990’s.27 This provides strong initial evidence for Balassa-Samuelson effects. The next section discusses estimation results for each country.

Figure 1.
Figure 1.

Cross-Sector Productivity Growth and Inflation Differentials, 1960-96

(In percent)

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A003

Source: OECD Intersectoral Database; and author’s calculations.

D. Estimation Results

66. Unit root tests were first run on the three variables; in all but one instance, the tests indicate that relative prices, productivity, and wages are integrated of order 1, warranting the next step of looking for a co-integrating relationship.28 The 3 equation model of the form of equation (8) was estimated separately for each country using the Johansen VAR methodology for testing of co-integrating relations; Table 2 contains selected estimation results. Either one or two lags are sufficient to account for serial correlation, while certain dummy variables outside the co-integrating vector (that is, only in the short-run part of (8), where the variables are in first differences) were necessary to control for large residuals in the estimated equations in particular years. In Finland, for example, the co-integrating relationship appears to be unchanged throughout the sample, but the short-run relationship shifts in 1991. In some countries, the co-integrating vector includes a constant and/or a trend; in Greece and Germany, for example, the level of relative prices trends upward, possibly reflecting increased effects of external competition that affects only prices of tradables but not those of nontradables.

Table 2.

Cointegrating Relationships between Relative Prices, Productivity, and Wages

article image
Note: Standard errors in parentheses.

67. In all countries but Norway and Sweden, the results of the cointegration tests (not shown) indicate at most one co-integrating vector. In some countries—notably Denmark but also France, Greece, and Portugal—the hypothesis of one co-integrating vector can be accepted at confidence levels of somewhat less than the standard 95 percent level. In these cases, one co-integrating vector is assumed, but then the estimated coefficients must be assessed to gauge whether sensible results are obtained.29 No co-integrating relationships could be estimated for Norway and Sweden using data for the entire sample period, so these countries are not considered in the results below. For Norway, this may result from the important role of oil prices, with the various oil shocks affecting the relationship between relative prices and productivity.30 For Sweden, cointegration can be found using the data through 1989, but the relationship is substantially affected by developments in the 1990s—presumably the effects of labor market changes in the early 1990s and the financial crisis in 1992—and no combination of dummy variables or trends can salvage the results.

68. The estimated coefficients for the co-integrating vector are shown in the first four columns of results in Table 2, normalized as in equation (8), with the coefficient on relative prices set to 1. The restriction that relative productivity is weakly exogenous (that is, the coefficient α in the productivity equation is zero) is not rejected in the estimation for any country. This means that any deviation from the long-run equilibrium of the Balassa-Samuelson model affects relative prices and wages but not productivity—the long-run levels of prices and wages adjust to TFP shocks but not vice-versa. In France and Germany, relative wages are also found to be weakly exogenous, indicating that the adjustment to shocks that move the economy away from the Balassa-Samuelson relationship comes about entirely through changes in relative prices. In Greece and Portugal, relative wages are found to have no effect on relative prices once relative productivity is taken into account, so that the co-integrating relationship is found to exist only between prices and productivity. Since relative productivity is again weakly exogenous in both countries, this means that the adjustment to productivity shocks comes about only through changes in relative prices.

69. The coefficients for the co-integrating relationship between prices, productivity, and wages vary across countries but have the expected signs (with only one exception): a larger TFP differential leads to a higher price differential in all countries, while a larger wage differential offsets the Balassa-Samuelson effect and tends to lead to a smaller price differential. The only exception is France, where wage differentials appear to exacerbate price differences. As seen in Table 1, however, wage growth has been remarkably similar on average between tradables and nontradables in France, so the effect of the anomalous coefficient for relative wages turns out to be quantitatively small.

70. The assumptions of weak exogeneity for TFP (and in some cases, wages) and the normalization of the coefficient on the relative inflation differential to 1 provide overidentifying restrictions that can be tested to assess whether the model is accepted by the data. The results are shown in Table 2: the model is accepted at fairly strong statistical levels in nearly every country. Finland and the United Kingdom are the only exceptions, meaning that in these countries the estimated coefficients are substantially affected by the restrictions imposed on the model. In both cases, however, the unrestricted results have the unacceptable property of explosive deviations from equilibrium (positive signs for the α coefficients), so the restrictions are imposed on the model.

71. Figure 2 shows the effects implied by the estimation results of a one percentage point shock to relative productivity on the level of relative prices (solid line) and the effect on the overall inflation rate (dashed line). The change in the level of relative prices depends on both the direct effect of productivity on prices and the response of wages to TFP and the subsequent effect of wages on prices. The contribution to overall inflation reflects the slope of the relative price response (since inflation is the change in the price level), and the share of nontradables in the economy (because the Balassa-Samuelson effect changes only prices of nontradables). The results indicate that an increase in the level of relative productivity generally leads to a higher inflation rate for several years before equilibrium is restored, though the largest effects are typically felt within the first three years. In Greece, the main effect of productivity shocks on prices diminishes rapidly, reflecting the rapid speed of adjustment (the large value for a), but the trend and constant give some recurring effects until equilibrium is reached. The productivity shock has the largest effect on prices and inflation in France, Italy, Portugal, and the United Kingdom, reflecting the large magnitudes of the long-run coefficient on TFP (first column of Table 2), and the weak (at best) response of wages in offsetting price changes in these countries. The response of prices to productivity is erratic in Denmark, possibly reflecting the lack of a long-run relationship between the variables in the model as suggested by the weak estimation results.

Figure 2.
Figure 2.
Figure 2.
Figure 2.

Effect of One Percentage Point Shock to Relative Total Factor Productivity on Relative Prices and the Inflation Rate

(In percent)

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A003

Source: Author’s calculations.

72. Figure 3 shows the predictions for the contribution of productivity differentials to overall inflation, calculated for both the original Balassa-Samuelson model with only relative prices and productivity as in equation (5), and for the estimation of the extended model in equations (7) and (8). As discussed above, the contribution to overall inflation is determined by the sectoral inflation from equation (9) that results from the average productivity and wage differentials over the entire sample for each country, and by the share of nontradables. In most countries, the estimated contribution to inflation is fairly close to the prediction of the model. In Greece, for example, the model predicts a contribution of 0.8 percentage points versus the results of 1.0 from the estimation. The estimated results of negative contributions of Balassa-Samuelson effects on inflation in Germany and the Netherlands stem from large long-run coefficients on wage differentials that more than offset the contribution of productivity differentials to higher inflation. This is also an important factor in Italy, where a large response of prices to wage differentials greatly diminishes the estimated contribution of productivity differentials to inflation compared to the prediction of the unaugmented model of equation (5) in which wages are assumed to grow at the same rate in both sectors.31 The opposite is the case in the United Kingdom, where productivity differentials have a large effect on relative prices—nearly twice that predicted by the model—and are only slightly offset by relative wage differentials. In France, differential wage growth across sectors substantially magnifies rather than offsets productivity differentials, resulting in the much larger contribution from the estimation than the prediction of the theory.

Figure 3.
Figure 3.

Contribution of Balassa-Samuelson Effect to Inflation, 1960-96

(In percent)

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A003

Source: Author’s calculations.*The estimated relationship is not available for Norway and Sweden because a cointegrating relationship could not be found.

E. Conclusion and Implications for Greece

73. The Balassa-Samuelson model appears to hold reasonably well in the countries examined, with statistically significant long-run relationships found between inflation and productivity differentials across sectors. On average for the period from 1960–96, the Balassa-Samuelson effect is estimated to have contributed an additional 1.0 percentage point of annual inflation in Greece. Over this period, annual inflation averaged 11.8 percent in Greece and 4.7 percent in the United States, while the exchange value of the drachma against the dollar fell by an annual average of 5.1 percent. Together, this implies an average annual real appreciation of the drachma of about 2 percent in terms of relative consumer prices against the dollar. The one percent contribution of the Balassa-Samuelson effect to higher inflation in Greece thus potentially accounts for as much as half of the real appreciation of the drachma against the dollar.32 The Balassa-Samuelson effect thus leads to an appreciation of the real exchange rate measured in terms of relative consumer price inflation, but this would not be considered a loss of external competitiveness.

74. To examine the role of the Balassa-Samuelson effect in the most recent period, Figure 4 shows the contribution of inflation calculated by using the original Balassa-Samuelson model of equation (5) and the average differentials for inflation and productivity for the available years in the 1990’s for each country. The estimated coefficients from equation (8) are not used in this exercise in order to exclude any data from earlier years. As expected, Balassa-Samuelson effects are largest in the 1990’s in the poorer countries with the largest productivity differentials across sectors. Uneven productivity growth accounts for 1.7 percentage points of inflation in Greece over 1990–96, out of an average inflation rate of 14 percent in this period. This compares to a weighted average 0.5 percentage point contribution in the other countries (weighting by each country’s share of tradable goods production), so that the Balassa-Samuelson effect potentially accounts for 1.2 percentage points of the inflation differential between Greece and the rest of Europe during this period. Taking into account the higher inflation in Greece than in the rest of Europe and the depreciation of the drachma against the ECU, Greece experienced a roughly 2½ percent annual real appreciation in terms of consumer prices. The 1.2 percentage point Balassa-Samuelson effect thus accounts for nearly half of the real appreciation in recent years—and again, this portion can be considered as not representing a loss of external competitiveness.

Figure 4.
Figure 4.

Contribution of Balassa-Samuelson Effect to Inflation, 1990-96

(In Percentage Points)

Citation: IMF Staff Country Reports 1999, 138; 10.5089/9781451816129.002.A003

Source: Author’s calculations. Finland and Sweden are omitted as discussed in the text.

75. While the data do not include the most recent years, the results for 1990–96 suggest that some of the present inflation differential vis-à-vis the euro area countries reflects supply-side factors of productivity growth rather than Greece’s more advanced cyclical position. Looking forward, the Balassa-Samuelson effect can be expected to remain an influence on inflation, since the level of productivity in tradables in Greece remains substantially below that of its EU partners.

References

  • Aitken, Brian, 1999, “Ireland and the Euro: Productivity Growth, Inflation, and the Real Exchange Rate,” IMF Working Paper, forthcoming.

    • Search Google Scholar
    • Export Citation
  • Alberola-Ila, Enrique, and Timo Tyrväinen, 1998, “Is there Scope for Inflation Differentials in EMU? An Empirical Evaluation of the Balassa-Samuelson Model in EMU Countries,” Bank of Spain Working Paper No. 98-23.

    • Search Google Scholar
    • Export Citation
  • Balassa, Bela, 1964, “The Purchasing Power Doctrine: A Reappraisal,” Journal of Political Economy, Vol. 72, pp. 584596.

  • Bragoudakis, Zacharias, and Demetrios Moschos, 1999, “Relative Prices and Sectoral Labor Productivity Differentials: A Long-Run Analysis for Greece,” University of Athens Discussion Paper No. 1999/5, June.

    • Search Google Scholar
    • Export Citation
  • Canzoneri, Matthew, Robert Cumby, and Behzad Diba, 1999, “Relative Labor Productivity and the Real Exchange Rate in the Long Run: Evidence for a Panel of OECD Countries,” Journal of International Economics, Vol. 47, pp. 245266.

    • Search Google Scholar
    • Export Citation
  • De Gregorio, José, Alberto Giovannini, and Thomas Krueger, 1994, “The Behavior of Nontradable-Goods Prices in Europe: Evidence and Interpretation,” Review of International Economics, Vol. 2, No. 3, pp. 284305.

    • Search Google Scholar
    • Export Citation
  • De Gregorio, José, Alberto Giovannini, and Holger Wolf, 1994, “International Evidence on Tradables and Nontradables Inflation,” European Economic Review, Vol. 38, pp. 12251244.

    • Search Google Scholar
    • Export Citation
  • Froot, Kenneth, and Kenneth Rogoff, 1995, “Perspectives on PPP and Long-Run Real Exchange Rates,” Chapter 32 in Handbook of International Economics, Volume III, ed. by Gene Grossman and Kenneth Rogoff,

    • Search Google Scholar
    • Export Citation
  • Lipschitz, Leslie, and Donogh McDonald, 1991, “Real Exchange Rates and Competitiveness: A Clarification of Concepts, and Some Measurements for Europe,” IMF Working Paper 91/25, March (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Moschos, Demetrios, and Y. Stournaras, 1998, “Domestic and Foreign Price Links in an Aggregate Supply Framework: The Case of Greece,” Journal of Development Economics, Vol. 56, pp. 141157.

    • Search Google Scholar
    • Export Citation
  • Samuelson, Paul A., 1964, “Theoretical Notes on Trade Problems,” Review of Economics and Statistics, Vol. 46, pp. 145164.

  • Summers, Robert, and Alan Heston, 1991, “Perm World Table (Mark 5): An Expanded Set of International Comparisons, 1950–1988,” Quarterly Journal of Economics, Vol. 106, pp. 32768, May.

    • Search Google Scholar
    • Export Citation
17

Prepared by Phillip Swagel.

18

See Balassa (1964) and Samuelson (1964). Froot and Rogoff (1995) provide an updated discussion including a survey of recent empirical work.

19

See Lipschitz and McDonald (1991) for a discussion of this effect and implications for measures of competitiveness.

20

This will not necessarily be the case for unit labor costs if there is a divergence between total factor productivity, which drives the Balassa-Samuelson effect on prices, and labor productivity, which determines unit labor costs.

21

Three equations are estimated in the VAR (one per variable), but the focus here is on relative prices, so only this equation is written out.

22

This calculation also implicitly assumes that each country was on average in equilibrium over the sample period, since only the equilibrium relationship between prices, productivity, and wages is used to calculate the contribution of the Balassa-Samuelson effect to inflation. As discussed below, some changes in the relationship are taken into account in the estimation by the addition of trend terms or shifts in the coefficients when these are needed to estimate a statistically significant long-run relationship.

23

The countries include Belgium, Denmark, Finland, France, Germany, Greece, Italy, the Netherlands, Norway, Portugal, Sweden, and the United Kingdom, Spain is not included because capital stock data are not available by sector, while Ireland is excluded because Aitken (1999) shows that the usual measure of TFP growth is distorted by the presence of large multinationals.

24

The industry breakdown follows the ISIC classification for production. Tradables consist of mining and quarrying, and manufacturing. Nontradables are: electricity, gas, and water; construction; wholesale and retail trade, restaurants and hotels; transport, storage, and communication; finance, insurance, real estate and business services; community, social and personal services; and government services.

25

Agriculture comprises less than 5 percent of the value of output on average over the sample period in all countries except Denmark (5.0 percent), France (5.5 percent), Italy (6.6 percent), Finland (9.3 percent), Portugal (12.0 percent), and Greece (13.8 percent).

26

The income ranking in Table 1 is based on real per-capita GDP for 1990 from the Penn World Tables International Comparison Project (popularly known as the Summers and Heston database, version 5.6). This provides internationally comparable measures of output based on price deflators with common baskets of goods and is the most widely used database with which to make cross-country comparisons of per-capita output. See Summers and Heston (1991).

27

Finland and Sweden are omitted from the calculations for 1990–96 because productivity growth and inflation differentials in these countries appear to be affected by particular events in this period that result in unusually large values for relative productivity growth. These include the financial crisis in the early 1990s in Sweden, and the development of the high technology sector in Finland.

28

The only exception where stationarity appears more likely than not is for relative prices in Denmark; the results are mixed for wages in Italy and productivity in the Netherlands, with acceptance or rejection of a unit root depending on the number of lags and inclusion of a trend in the augmented Dickey-Fuller test. The finding of a long-run co-integrating relationship for Denmark is somewhat weaker than in most other countries, which is consistent with the lack of a unit root in relative prices.

29

Bragoudakis and Moschos (1999) estimate the model of Alberola-Ila and Tyrväinen (1998) for Greece over 1962–97 (and again, use labor productivity rather than total factor productivity), but find no evidence of cointegration and implausibly large coefficients for the response of prices to productivity and wages. In contrast, the estimation here uses total factor productivity and produces reasonable coefficients though the existence of a stable long-run relationship in Greece is accepted at a lower level of statistical significance than in some of the other countries.

30

However, adding oil prices or dummy variables in years of oil shocks as exogenous variables did not result in a finding of cointegration.

31

Leaving government services out of nontradables eliminates most of the wage differential between tradables and nontradables, giving an estimated contribution to inflation close to the predicted 1.8 percentage points from the model.

32

The amount would be smaller to the extent that the Balassa-Samuelson effect contributed to higher inflation in the United States, but this contribution is likely to be quite small.

Greece: Selected Issues
Author: International Monetary Fund