functional income distribution and aggregate demand in the euro area
Engelbert Stockhammer
Functional Income Distribution and Aggregate Demand in the Euro Area
Institut für Geld- und Finanzpolitik University of Economics and Business Administration, Vienna (WU) Wien Study commissioned by the Chamber of Labour, Vienna
Functional Income Distribution and Aggregate Demand in the Euro Area
Engelbert Stockhammer [email protected] Institut für Geld- und Finanzpolitik University of Economics and Business Administration, Vienna (WU) Study commissioned by the Chamber of Labour, Vienna Vienna, March 2007
Acknowledgements The author is grateful to Paul Ramskogler and especially Stefan Ederer for their excellent work as project assistants and to Michael Mesch, Özlem Onaran, Werner Raza and Georg Zwiener for their ideas and suggestions. The study profited from synergies with the FWF Project No. P18419-G05.
Erhältlich bei:
Kammer für Arbeiter und Angestellte für Wien Abteilung EU und Internationales Prinz-Eugen-Straße 22-22 1040 Wien Tel: +43(01) 50165 – 2278 Fax: +43(01) 50165 - 2199 Email: [email protected] Internet: http://studien.arbeiterkammer.at
Das Werk ist urheberrechtlich geschützt. Die dadurch begründeten Rechte, insbesondere die Übersetzung, der Nachdruck, die Entnahme von Abbildungen, der Funksendungen, der Wiedergabe auf photometrischem oder ähnlichem Wege und der Speicherung in Datenverarbeitungsanlagen, bleiben, auch nur auszugsweiser Verwertung, vorbehalten. © 2007 Kammer für Arbeiter und Angestellte für Wien, 1040 Wien, Prinz-Eugen-Straße 22-22 Impressum: Medieninhaberin, Herausgeberin: Kammer für Arbeiter und Angestellte für Wien, 1040 Wien, Prinz-Eugen-Straße 22-22 Verlags- und Herstellort: Wien
Executive Summary An increase in the wage share has different effects on the various components of aggregate demand: ¾ It affects private consumption expenditure positively, because recipients of wage income have a greater propensity to consume than recipients of capital income. ¾ It has negative effects on (private) fixed capital formation, because profits (which perforce become relatively less on an increase of the wage share) have a positive effect. ¾ Net exports are also negatively impacted by an increase in the wage share, because unit labour costs rise with the wage share, which in turn reduces competitiveness. Given these positive and negative effects, it is theoretically impossible to provide a clearcut answer to the question of how an increase in the wage share affects demand a priori. Rather, two regimes are possible. If an increase in the wage share results in an increase in aggregate demand, the demand regime is referred to as “wage-led”; if it leads to a reduction in aggregate demand, it is called “profit-led”. Econometric methods are applied in this study to examine this issue for the euro area. The euro area is the prototype of a large, rather closed economy. The following picture emerges: ¾ The effects on consumption are much stronger than those on investments. The domestic sector is therefore clearly wage-led. ¾ Wage costs have a minimal impact in the euro area on net exports (relative to GDP). ¾ A 1% increase in the wage share in the euro area leads to a 0.2% rise in demand. Different methods of estimation yield similar results. However, the findings for the euro area are not directly applicable to its member states. A policy dilemma may arise. Wage moderation may well trigger short-term growth increases for certain (mostly small) countries, but if the same policy is adopted by all (EU) states, it clearly slows down demand and contributes to unemployment instead of reducing it. Wage moderation has dominated functional income distribution in Europe for about 25 years and unemployment has increased in that same period. The adjusted wage share in the euro area has fallen by 11.6 % since 1981, while the unemployment rate has increased by 1.2 %. The findings presented here help to explain this apparent paradox. At EU level, wage moderation curbs demand and therefore does not increase employment. One way out of this policy dilemma would be to coordinate wage policy throughout Europe (or within the euro area). This approach would help to stabilise income distribution. If broadly accepted, a coordinated, productivity-based wage policy could also be reconciled with the goal of price stability. It could have an inflationary effect in the transitional phase, however. 3
A coordinated wage policy, however, is not consistent with the current economic policy regime of the EU. Foreign exchange rates are now gone as an economic policy instrument, while monetary policy is uniform throughout the EU and fiscal policy is greatly limited by the Stability and Growth Pact. As a result, wage policy now bears most of the burden of adjusting the economy to asymmetrical shocks. If coordinated, wage policy would be less available as a tool for pursuing national economic policy goals. A coordination of wage policy would therefore also have to involve a change in the entire policy package in the EU, with fiscal policy taking on a more active role.
4
1. Introduction The wage share in the euro area has seen a sharp decline of 11.6 %-points since 1981, whereas the unemployment rate has increased a further 1.2 % since that year from an already high level (see Figure 1.1). 1 Europe has been exercising wage moderation for 25 years without seeing any resulting decline in unemployment. How is that possible? Doesn’t every textbook on microeconomics say that wage decreases lead to an increase in employment? The current Non-Accelerating Inflation Rate of Unemployment (NAIRU) theory also supports this thesis. But it is not just textbooks, old and new. The EU Commission also joins in the chorus supporting “employment-friendly labour cost developments and wage setting mechanisms” (EC 2006, 40). What it actually means is wage moderation. It is argued that moderate wage increases (coupled with dynamic productivity growth) strengthen Europe’s competitiveness and lead to economic growth and employment. Figure 1.1 Wage moderation without Employment Growth in the Euro Area 70
12
68 10
Adjusted Wage Share (market prices) 66 64
8
Unemployment rate
62 60
6
58 4 56 54 2 52
adj WS unempl.rate
50
0
1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006
And who would dare dispute that wage moderation improves competitiveness? No one. But does improved competitiveness automatically mean more growth and employment? Foreign trade is just one part of aggregate demand. The other components are private consumption and private investment expenditures, public consumption and public investment expenditures. To understand the effects of wage moderation on demand in the overall economy, one must determine how all these components react to a change in income distribution.
1
Data source: AMECO, adjusted wage share at market prices
5
This empirical study focuses on precisely this issue for the euro area, namely how a change in functional income distribution affects aggregate demand. The euro area is a large, relatively closed economy. This study is structured as follows. First, section 2 discusses the theoretical background for the empirical study, namely, a neo-Kaleckian macro model that allows for wage-led and profit-led demand regimes. Then, econometric estimates for the euro area will be presented (in section 3). This section will analyse how a change in the wage share affects private consumption, private investment expenditures and foreign trade. Section 4 will discuss the economic policy implications of the findings. Wage coordination will be a pivotal point throughout. The discussion will explore implications for aggregate wage development and a design for and possible problems associated with wage coordination. Special attention will be paid to the role of wage policy in the economic policy regime of the EU. .
6
2. Theoretical Background: Wage-led and Profit-led Demand Regimes To examine the effect a change in functional income distribution has on aggregate demand, this study will make use of a general neo-Kaleckian model based on Marglin and Bhaduri (1990). In classical Kaleckian models (for closed economies), an increase in the wage share always results in increased demand. The model used here permits profit-led and wage-led growth, because it allows for a positive effect of profits on investment expenditures. The question of the overall effects that a change in functional income distribution has on demand thus becomes an empirical question. This chapter will present the theoretical background for this empirical study. Aggregate demand (Y) is the sum total of consumption (C), capital investment (I), net exports (NX) and government spending (G). All of these variables except government spending are a direct function of the profit share (π). Government spending and revenues can also react to changes in income distribution but this fact is not taken into account in the following discussion. The model (presented in Table 1) is therefore one of a private open economy and embodies several simplifications. Income distribution, namely the profit share (π), 2 is assumed to be exogenous. Feedback from growth affecting income distribution, i.e. regarding lower unemployment and a better negotiating position for employees, are not considered at this stage. 3
Table 2.1: Model Overview GDP
Y = C(π) + I(π) + NX(π) + G
Profit share
π == R/Y
Consumption
C = f (W, R)
Investments
I = f (R, Y, i)
Exports
X = f (ULC, PIMP,YW; E)
Imports
M = f (ULC, PIMP, Y)
Y: GDP; C: Private Consumption; I: Private Investments; NX: Net Exports; G: Government Spending; π: Profit share; W: Total Wages; R: Profits; i: Long-term Interest Rate in Real Terms; X: Exports; M: Imports; ULC: Unit Labour Costs; PIMP: Import Price Deflator; YW: GDP of other countries. 4
The effect an increase in the profit share has on aggregate demand and production is uncertain a priori and depends on the sum of reactions from the components of GDP, namely, consumption, investments and net exports. To determine how a change in functional income 2
Functional income distribution and its measure, the profit share, are used synonymously in this report. The wage share is by definition one minus the profit share. For theoretical modelling, it is therefore immaterial which ratio is used. 3 In terms of econometrics, the problem of simultaneity is ignored. 4 See Appendix A for more detailed definitions of the variables.
7
distribution affects consumption, income is broken down into wages (W) and profits (R) and the marginal propensity to consume is estimated for both. Since the propensity to save investment income is higher than for wage income, consumption can be expected to decline as the profit share rises. The investment function depends on profits (R), output (Y) and the long-term interest rate in real terms (i). Investments will increase as the profit share increases, because the expected profits become greater. In keeping with the standard theory, one expects a positive effect on output and a negative effect on interest rates. In an open economy, exports and imports must be included in the analysis. Exports are a function of unit labour costs (ULC), which determine the competitiveness of the domestic economy on the world market. By definition, unit labour costs are closely tied to the wage share. If unit labour costs rise, demand for domestic products will decline domestically and abroad, because domestic prices rise relative to foreign prices. Exports also depend on the GDP of the trade partner. This factor is therefore incorporated in the export function. By definition, unit labour costs have an inverse relationship to the profit share. Imports also depend on unit labour costs relative to the foreign price level and on domestic GDP. Exports should react positively to an increase in the profit share, because, by definition, this increase causes a decline in unit labour costs. If unit labour costs decline relative to foreign prices, domestic goods become more competitive on the world market. Exports are therefore stimulated. For this same reason, imports will react negatively to an increase in the wage share. As net exports are obtained by subtracting imports from exports, the total effect on net exports will presumably be positive. The expected sub-effects of a change in income distribution on demand 5 in an open economy are summarised below: •
∂C / ∂π < 0
•
∂I / ∂π > 0
•
∂NX / ∂π > 0
•
∂Y / ∂π = ?
The total effect of a change in functional income distribution on aggregate demand depends on the magnitude of the sub-effects and can only be determined empirically. If the effect is positive (∂Y/∂π > 0), the demand regime is said to be profit-led. If it is negative (∂Y/∂π < 0), demand is said to be wage-led. 6
5
To be precise, excess demand (refer also to Bowles and Boyer 1995). The above discussion pertains to the issue of whether total economic demand is wage-led or profit -led. In the introduction, the question was raised whether employment (or demand for labour) is wage-led or profit-led. The two issues are naturally closely connected. If wage changes affect labour productivity, e.g. as postulated by the efficiency wage theory, then a wage-led demand regime is a necessary but not sufficient condition for a wage-led employment regime. 6
8
Given the strength of international trade, one would empirically expect net exports to play a dominant role in determining the overall result. However, although individual countries can boost demand by expanding exports, the world as a whole cannot. It is therefore important to distinguish between the domestic sector of an economy and an open economy. In this case, the domestic sector is defined as the sum of consumption and investments, assuming no change in net exports (as would happen if wages changed in all countries simultaneously). If consumption reacts more sensitively than investments to an increase in the profit share, the domestic sector is said to be wage-led. The strategy applied for econometric estimation is similar to the one in Bowles and Boyer (1995). The model is estimated using the individual equations for saving, investments and net exports. The key divergences from Bowles and Boyer are as follows: First, the econometric specifications differ. Bowles and Boyer do not discuss any time series characteristics of the economic variables and neglect the subject of unit roots. Consequently, they use no difference or error correction models of the kind that form the core of modern time-series econometrics. Second, Bowles and Boyer use the employment rate as a proxy variable for capacity utilisation. For several European countries with long-standing unemployment, that is a misleading indicator. Therefore, output growth is used in this study instead. Brief mention should be made of the empirical literature on neo-Kaleckian models at this juncture. Gordon (1995a) estimates consumption and investments as a function of income distribution for the United States. In a vector autoregressive (VAR) model, numerous exogenous shocks are simulated. Gordon (1995b) expands this model to include the open economy and empirically examines how aggregate demand reacts to changes in functional income distribution in the United States. His conclusion is that the United States has a profitled growth regime. In their studies, both Hein and Krämer (1997) and Hein and Ochsen (2003) use a model for a closed economy based on Marglin and Bhaduri (1990). Hein and Ochsen (2003) expand the model to include the interest rate as an exogenous variable and work out various accumulation regimes depending on the sensitivity of the savings function and the investment function to the interest rate. In the empirical section of their studies, they estimate the savings function and the investment function econometrically and attempt to characterise the accumulation regimes of France, Germany, the United States and Great Britain. Stockhammer and Onaran (2004) estimate a structural VAR model consisting of variables for capital accumulation, capacity utilisation, profit share, unemployment rate and growth in labour productivity for the United States, the United Kingdom and France. They use a VAR model to take into account the reciprocal interaction of the variables. The goods market is simulated with a model based on the work of Marglin and Bhaduri (1990). It is supplemented with an equation for income distribution, a productivity function and a function for unemployment. The conclusion from the findings of the empirical study is that unemployment is determined by the goods market but that the impact of the income distribution on demand
9
and employment is very weak. Technical progress shifts income distribution in favour of profits. Onaran and Stockhammer (2005) apply a similar model to Turkey and Korea. Naastepad (2006) presents and estimates an expanded model for the Netherlands, in which the growth of labour productivity is explicitly modelled. Productivity, savings, investments and exports are estimated as individual equations. One finding is the Dutch demand regime from 1960 to 2000 was wage-led, albeit marginally so. Her overall conclusion is that Dutch demand has been relatively insensitive to changes in real wage growth since the end of World War II.
10
3. Empirical Study for the EU This chapter will empirically investigate the effects unleashed by changes in functional income distribution. To this end, functions for consumption, investment, and foreign trade will be econometrically estimated for the euro area utilising the model presented in Chapter 2. An attempt was made to keep modelling pragmatic and where possible, to assume broadly accepted behavioural functions and then expand them to include distribution variables. Unfortunately, there are hardly any reference models for the estimated behavioural equations. There is currently still a lack of macroeconomic studies on the central behavioural equations for the euro area (covering a prolonged period with annual data). One obvious reference point would be the ECB Area-Wide Model (AWM; Fagan, Henry and Mestre 2001). It is heavily theoretical, however, and contains a number of (theoretically motivated) restrictions that render the coefficients incomparable with the estimates presented here. For this reason, the WIFO Macro Model (Baumgartner, Breuss and Kaniovski 2005) was selected as the reference framework for estimations on the euro area. A pragmatic approach was also taken with regard to time-series specification. Where possible, use was made of error correction models (ECM), which currently constitute the most common time-series specification. Other specifications were used in cases where diagnostic statistics suggested their suitability; in practice, most of them were difference specifications. Income distribution was assumed to be exogenous in all estimations. Of course, this assumption is not completely accurate. In other words, the estimations do not take into account pro-cyclical mark-up prices (e.g. resulting from constant mark-ups to the normal unit costs) or the negative effects that unemployment has on wages. This approach is justified not by theory but merely by the limited scope of this study. Endogenizing income distribution would be the subject of a more extensive research project. The AMECO database (on the euro area) served as the database for estimations and contains the largest number of relevant time series of (aggregate) values for the Euro 12. The definitions of the variables can be found in the Data Base Appendix. The euro area is viewed as an entity for the entire period (approximately 1960 to 2004), even before the monetary union. Close attention was therefore paid to test for structural breaks.
3.1 Consumption Function The consumption function was estimated in the first differences (of the logarithms) after the autoregressive distributed lag (ADL) specification clearly suggested a difference specification. The Augmented Dickey-Fullter (ADF) test for residues did not suggest a cointegration correlation (at log levels). This finding is somewhat surprising, as the consumption function can usually be modelled effectively with error correction mechanism (ECM) models. 11
The hypothesis of different propensities to consume is clearly supported (see Table 3.1). It was estimated at 0.49 (statistically significant at the 1 % level) for wages and 0.11 (statistically significant just above the 10 % level) for profits. The coefficients represent elasticities and correspond to marginal propensities to consume of 0.52 (for wages) and 0.15 (for profits). The consumption differential between wages and profits thus amounts to 0.37. 7 A 1% redistribution from wages to profits therefore reduces consumption spending by 0.37 % of GDP. Table 3.1 Dependent variable: Variable C ln (W) ln (W(-1)) ln (R) ln (R(-1)) ln (C(-1)) Δln (W) Δln (R) Adj, R² DW stat.
ADL ln (C) Coeff. Prob. 0.16 0.68 0.38 0.00 -0.39 0.00 0.12 0.12 -0.11 0.12 0.98 0.00 1.00 1.62
Differences Δln (C) Coeff. Prob. 0.01 0.00
0.49 0.11 0.78 1.45
0.00 0.10
3.2 Investments The investment function was estimated as ECM (in keeping with the WIFO specification) by imposing the restriction that the investment ratio is constant in the long run, i.e. technically speaking that the investment and output coefficients are identical. In addition, profits were included as is customary in various recent investment equations (however, not in the WIFO macro model). Table 3.2 summarises the regression results. Besides the usual ECM specification, a specification with a more generous lag structure was estimated. This increased the profit coefficient and improved its statistical significance, which nonetheless occurs only at a level of 10 %. The elasticity of investments relative to profits is modest, with a long-term value of 0.15 (=0.03/0.21). The marginal effect is 0.07. An increase in profits by 1% of GDP leads ceteris paribus to an increase in investments of 0.07 % of GDP.
7
Taking the logarithms of the variables yields an unpleasant difference relative to the theoretical model. The latter is additive in levels, while the estimated model is additive in log levels. Implicitly, thus the consumption function is subject to the Cobb-Douglas functional form. The calculated consumption difference can therefore be viewed only as an approximation.
12
Table 3.2 Dependent variable: Variable C Δln (Y) Δln (Y(-1)) Δln (R) Δln (R(-1)) ln (I(-1)/Y(-1)) ln (R(-1)) i (-1)) Adj, R² DW stat.
ECM Δln (I) Coeff. Prob. -0.30 0.00 1.89 0.00 0.13
0.49
-0.14 0.01
0.01 0.61
0.72 1.73
ECM with lags Δln (I) Coeff. Prob. -0.59 0.00 1.59 0.00 0.84 0.03 0.38 0.06 -0.18 0.38 -0.21 0.00 0.03 0.07 -0.00 0.43 0.73 1.75
3.3 Net Exports Estimating the effect on net exports is probably the most sensitive part of determining the overall effect of a change in income distribution. It was therefore done carefully utilising a dual strategy. First, a net export function was estimated directly as a function of domestic GDP growth, of key trade partners’ GDP growth, of (nominal) exchange rates and of unit labour costs. The advantage of this method is two-fold: it is simple and it answers the relevant question. Its disadvantage is that it is not compatible with the foreign trade functions in current macro econometric models (such as the WIFO Macro Model). Wage costs are generally not integrated in the import and export functions directly but rather indirectly via prices. As a second method, exports and imports were estimated separately in a first step as a function of the price level (and other control variables). This procedure follows the WIFO model, which can be considered typical of modern modelling. In a second step, prices were regressed to the nominal unit labour costs (and import prices). The price equation implicitly contains information on how high a nominal wage increase must be to increase real unit labour costs by one percent. This latter fact is relevant here. Data definitions: There are data available on the exports and imports of goods from the euro area and non-euro countries, but unfortunately none for services. For this reason, all estimates on goods exports and imports and the subsequent results were weighted with a factor reflecting the relative magnitude of service exports (in 2005). The resulting error is presumably small. For the United States, there are data for goods and services covering a long period but no trend is evident. The value averages 1.25. GDP growth of the trade partners was calculated from the weighted GDP growth rates of the US, Switzerland, Japan, Norway and Turkey. The weights correspond to the relative shares in foreign trade involved (in recent years). In an analogous step, the exchange rate was calculated from the weighted growth rates of exchange rates.
13
Table 3.3 summarises the results of the direct estimation of the net export ratio, i.e. the first estimation strategy. All variables have the signs expected with the exception of the exchange rate, which is statistically significant at the 5 % level. Owing to the high autocorrelation in the residues from the original estimation equation, the estimation was repeated with an AR(1) correction. There are hardly any substantial changes in the results. Autocorrelation problems remain in spite of a marked improvement in the DW statistic. A one per cent increase in real unit labour costs reduces net exports by about 0.1 % of GDP. After the necessary transformation to take into account service exports and imports, this corresponds to an effect of -0.15 % of GDP with a 1 % increase in the wage share 8 . Table 3.3 Dependent variable: Variable C Δln (Y) Δln (YW) Δln (E) ULC Adj, R² DW stat.
Without AR1 corr. NX / Y Coeff. Prob. 0.06 0.00 -0.15 0.01 0.16 0.01 0.01 0.58 -0.11 0.00 0.43 0.68
With AR1 corr. NX / Y Coeff. Prob. 0.05 0.06 -0.18 0.00 0.12 0.01 0.01 0.28 -0.10 0.03 0.70 1.42
The second method for estimating the effect of a change in functional income distribution on net exports is more elaborate. In a first step, the exports are regressed to the relative prices (export prices relative to import prices) and to the output of the trade partners and exchange rate (relative to the main trade partners) (Table 3.4). As there were no indications of cointegration, a difference specification (with autocorrelation correction) was selected. The coefficients have the signs expected, but at a level of only low statistical significance (at 10 % for the GDP of the trade partners and at a level of less than 10 percent for the exchange rates). The coefficient for the relative prices is statistically significant at the 1% level and has an elasticity of -1.1. A difference specification also had to be used in the import regression. A dummy variable for the years following 1995 was also included to capture a structural break (albeit on an ad hoc basis). 9 The equation contains the GDP of the Euro 12 countries as well as the exchange rates and the relative prices. While economic growth has the sign expected and is statistically significant (at the 1% level), the exchange rate has a perverse sign (and is statistically significant at the 5% level). As neither specifications with different lag structures nor specifications without exchange rates brought about any improvement, the value from the original specification was taken although it is not statistically significant. The coefficient of the relative prices is small, at 0.3. Imports therefore react weakly to changes in the price level. 8
Unit labour costs in real terms were corrected to take into account the changes in the structure of employed persons. As the ratio of total employees to dependent employees is approximately 1.25, this figure must serve as the multiplier. Refer to Marterbauer and Walterskirchen (2003) for the larger discussion. 9 The inclusion of the dummy variable is primarily statistically motivated. A possible explanation for this break is the opening up of Eastern Europe markets, which increased imports.
14
Table 3.4 Exports Dependent variable: Δln X Variable Coeff. Prob. C 0.03 0.22 Δln (YW) 0.92 0.10 Δln (E) -0.28 0.07 Δln (PX/PM) -1.14 0.00 AR(1) 0.37 0.02 Adj. R² DW stat.
0.37 1.69
Imports Dependent var.: Δln M Variable Coeff. Prob. C -0.03 0.11 Δln Y 2.32 0.00 Δln E -0.33 0.02 Δln E (-1) Δln (P(-1)/PM(-1)) 0.29 0.13 T95 0.06 0.00 Adj. R² 0.46 DW stat. 1.74
In two auxiliary estimations, the price level was explained in terms of nominal unit labour costs, import prices and GDP growth, while export prices were explained in terms of domestic price level and import prices. Both estimations were conducted in differences and included autocorrelation correction. A one per cent increase in nominal unit labour costs increases domestic inflation by 0.54; a one per cent increase in domestic inflation increases export prices by 0.47. From the price equation, one can conclude that a 2.2% increase in nominal unit labour costs is required to increase real unit labour costs by 1 %. Prices would then rise by 1.2%. Table 3.5 Dependent variable: Variable C Δln Y Δln PM Δln ULC AR(1) Adj. R² DW stat.
Δln P Dependent variable: Coeff. Prob. Variable 0.02 0.01 C 0.16 0.04 Δln PM 0.03 0.11 Δln P 0.54 0.00 AR(1) 0.85 0.00 0.95 Adj. R² 2.12 DW stat.
Δln PX Coeff. Prob. 0.00 0.36 0.51 0.00 0.47 0.00 0.44 0.00 0.94 2.15
As the calculation of the effects of income distribution on exports and imports is composed of multiple steps, Table 3.6 summarises the relevant parameters to provide an overview. The export and import ratio display a trend, so that elasticities can be converted to marginal effects in various ways. This can be done either using the mean, which corresponds to the average effect of the period, or using the value of the final year (2005), which corresponds to the effect at the end of the period. The overall effect of functional income distribution on exports comprises the effect of real unit labour costs on prices and the effects of prices on export prices. This elasticity value is then converted to a marginal value and lastly, the unit wage costs are converted to the wage share and the result is weighted with the relative share of services as a percentage of total exports. In sum, a 1% increase in the wage share results in a decrease in exports equal to 0.07% of GDP. On average the effect over the observed period was only about 0.05. The analogous values for imports amount to 0.04 and 0.03 % of GDP, respectively. 15
Table 3.6 Exports ∂X/∂Px ∂Px/∂P ∂P/∂RULC 1/RULC X/Y 2005 -1.14 0.47 1.19 0.63 0.47 1.19 0.63 X/Y Ø -1.14 Imports ∂M/∂P ∂P/∂RULC 1/RULC M/Y 2005 0.29 1.19 0.63 1.19 0.63 M/Y Ø 0.29
X/Y 0.10 0.07 M/Y 0.11 0.08
ET/ED G&S / G 1.27 1.25 1.27 1.25 ET/ED G&S / G 1.27 1.25 1.27 1.25
Effect -0.07 -0.05 Effect 0.04 0.03
The effect of a one per cent increase in the wage share on the net export ratio is thus -0.11 in the final year and -0.08 on average. The difference between the two values indicates the increase that has occurred as a result of growth in foreign trade, i.e. globalisation. Both values are lower than the one obtained in the first direct estimation method. There is no obvious explanation for this fact. As will become apparent, however, this difference elicits no qualitative change in the overall result.
3.4 Total Effect The sub-results can now be brought together. The different methods used for calculating the effect on net exports are indicated separately. The (positive) effect an increase in the wage share has on private consumption is substantially greater (0.37% of GDP) than the (negative) effect it has on investments (0.07 % of GDP). The domestic sector of the euro area is therefore clearly wage-led. The net export effect is greater than that of investments (ranging from 0.07% to 0.15% of GDP depending on the estimation method employed). Table 3.7 presents three methods for estimating or calculating the net export effect. The first is a direct method, in which the net export ratio was regressed to real unit labour costs and other control variables. This approach yields the strongest effect on net exports. The second is an indirect procedure, in which exports and imports were regressed to the price level (and other variables) and in a second step, the price level was expressed in terms of nominal unit labour costs (and other variables). This method tends to correspond more closely to the customary modelling strategy currently in use. This method yields elasticities, which are then converted to marginal effects with the current export and import ratios (column “XM ratio 2005”) or the average (column “XM ratio Ø”) export and import ratios. The former indicates the net export effect “today” (0.1); the latter, the average effect (0.07) for the period as a whole. The values obtained with the different methods do vary, but not too greatly. For the overall interpretation, the most compelling method is probably the indirect one involving conversion with “today’s” effect at a value of 0.1. The total effect of a 1% change in the wage share fluctuates between +0.15% (when the net export ratio is estimated directly) and +0.23 % of GDP (using the indirect estimation method based on average foreign trade involvement). This effect is thus positive in each case. The qualitative result of the examination is that the euro area has a wage-led demand regime.
16
Wage moderation at EU level thus has negative demand-side effects on growth dynamics. The most plausible estimation of the demand effect in percentage points is 0.2% of GDP. Table 3.7
Consumption effect Investment effect Domestic effect Effect on net exports Total effect
NX ratio, direct 0.37 -0.07 0.30 -0.15 0.15
XM ratio, 2005 0.37 -0.07 0.30 -0.10 0.20
XM ratio, Ø 0.37 -0.07 0.30 -0.07 0.23
17
4. Implications for Wage Policy This chapter will explore the policy implications of the results from the empirical study on how demand is affected by changes in income distribution. First, the results will be summarised in a way that can be understood even without prior knowledge of econometrics. The fact that the euro area has a wage-led demand regime naturally has ramifications for various policy areas. The following section will focus especially on those pertaining to wage policy. Economically, wage policy has a variety of effects: distribution effects, price effects, demand effects and innovation effects. Wage policy can thus help to stabilise income distribution and prices and stimulate demand and innovation activity. This study is devoted primarily to the effects wage policy has on demand.
4.1 Summary of the Findings An increase in the wage share has different effects on the various components of aggregate demand: ¾ It affects private consumption expenditure positively, because recipients of wage income have a greater propensity to consume than recipients of capital income. ¾ It has negative effects on (private) fixed capital formation, because profits (which perforce become relatively less on an increase of the wage share) have a positive effect. ¾ Net exports are also negatively impacted by an increase in the wage share, because unit labour costs rise with the wage share, which in turn reduces competitiveness. Theoretically, it is therefore difficult to provide an non-ambiguous answer to the question of how an increase in the wage share affects demand, because there are positive as well as negative effects. That is why this question was subject to empirical examination for the euro area. The euro area is the prototype of a large, closed economy. The following picture emerges: ¾ In the euro area, the effects on consumption are much stronger than those on investments. The domestic sector is therefore clearly wage-driven. ¾ Wage costs have a minimal impact on net exports (relative to GDP) in the euro area. The euro area thus clearly has a wage-led demand regime. A 1% increase in the wage share consequently results in a 0.2% increase in demand. Different estimation methods yield similar results.
18
4.2 Wage Moderation and Employment in the EU – A Prisoner’s Dilemma The first and most direct conclusion that can be drawn in respect of economic policy is that the wage cuts at EU level undertaken to fight unemployment are unsuitable means for achieving this objective. Wage moderation reduces aggregate demand and will therefore not lead to an expansion in employment, but rather to the opposite. This finding is an important one, as it runs counter to the prevailing policy in Europe. The fact is wage moderation has been practised in the EU for 25 years without bringing about any reduction in unemployment. The EU Commission often talks of “employment-friendly wage-policy” but what it means is wage moderation in the name of competitiveness. As a wage reduction has a greater effect on domestic demand than on foreign trade, wage moderation in this case will not stimulate growth and employment but do the very opposite. Another finding of this study that is presumably just as important is that certain countries, especially small extra-verted countries, may see at least short-term growth if they exercise wage moderation. In small open economies, the effect on foreign trade may be greater than the domestic effect on demand. And in fact, several countries, notably the Netherlands and Ireland, have exercised wage moderation, often in the form of policy packages comprising promises of wage moderation from unions, of investments from companies and of social, educational and tax policy measures from the government. This approach appears to have yielded good economic results at least at first glance. In the Netherlands, a prime example of a country that uses wage pacts, wage moderation has resulted in a persistent erosion of productivity growth, which in turn undermines any competitive advantages that may have been gained (Naastepad 2006). A central conclusion of this study is that generalisations on these kinds of policies cannot be made at EU level. The main trade partners of these small open EU member states are naturally other EU member states. While individual instances of wage moderation may have an expansive effect in many individual (perhaps all) EU member states, the opposite is true at EU level. That may sound contradictory initially but has a simple explanation: The EU member states conduct the bulk of their trade within the EU. A country that practises wage moderation and exports to other countries thus competes domestic production out of the latters’ markets. In other words, it exports not only goods but unemployment. This is merely a version of the “beggar thy neighbour” policy, i.e. of effective devaluation by means of wage moderation. In the 1930s, this policy was pursued using exchange rates and custom duties and led to a collapse of world trade. Although there is no immediate risk of that right now, the present wage policy situation could be described as a prisoner’s dilemma. Even if wage moderation did have an expansive effect for each country, 10 the expansive strategy for 10
This idea is posited as a theoretical argument, not as a description. In fact, there appear to be some EU countries in which wage moderation has not had an expansive effect. The German economy has been stagnating for years even though it has the most moderate growth in unit labour costs of the entire euro area (refer to Figures 5.2 and 5.3) and high net exports. A possible explanation is that Germany has a wage-led demand regime.
19
all of them jointly would be to increase wages. As in all constellations described as prisoner’s dilemmas, cooperation and coordination are the two means of possible escape. In this case, wage coordination could be the key to solving the problem.
4.3 Current Developments in Wage Policy in the EU In all countries in the EU (and in most other OECD countries), there has been a clear trend over the past 25 years away from productivity-oriented wage policy to competition-oriented wage policy (Schulten 2004). The adjusted wage share (and unit labour costs) has consequently declined almost continuously since 1980. In their thorough discussion of labour relations in Europe, Marginson and Sisson (2004) conclude that there is virtually no wage coordination Europe-wide, indeed, not even transnationally (compare also Schulten 2004, Chapter 11). They do note, however, that the effective changes in real wages and unit labour costs have converged and therefore describe this situation as “convergence without coordination”. This might suggest that, as a result of inflation rate convergence in the course of and in preparation for the introduction of the euro, wage growth rates have also converged and that no additional action is needed. On the other hand, the increasing foreign trade imbalances within the EU are attributable not only to differing growth performances but also to German wage moderation (Flassbeck and Spiecker 2005). Have growth in wage rates and unit labour costs in fact converged? An initial glance at the standard deviation of growth rates in nominal wages for the euro area (Figure 4.1) would seem to indicate just that. The introduction of the euro did actually bring about not only a convergence in inflation rates but also a marked decrease in the standard deviation in wage growth rates. The picture is not yet complete, however. Although the range of wage growth fluctuation is narrower, this fact does not exclude the possibility that substantial cumulative divergences exist. The relevant factor for competitiveness is not wage growth (or the change in unit labour costs), but rather the level of unit labour costs.
20
Figure 4.1 Standard Deviation of Growth in Nominal Wages for EU12 0.050 0.045 0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Figure 4.2 shows the development of unit labour costs since 1999. This figure does not indicate convergence. While unit labour costs have barely increased in Germany and risen by a mere 5% in Austria, they have grown in Portugal and Ireland by more than 25%. The trend does not indicate a stabilisation of relative unit labour costs but rather a more extensive divergence. Figure 4.2
Nominelle Lohnstückkosten in EU12 (1999=100) Nominal Unit Labour Costs EU12 (1999=100) 1.3 Belgium
1.25
Germany (linked) Greece
1.2
Spain 1.15
France Ireland
1.1
Italy Luxembourg
1.05
Netherlands 1
Austria Portugal
0.95
Finland
0.9 1999
2000
2001
2002
2003
2004
2005
2006
21
A different trend in unit labour costs relative to a (arbitrarily chosen) base year would not be a problem per se, unless unit labour costs were identical in the base year. Presumably, the different currencies did not accede to the monetary union with the “correct” exchange rates. Part of the divergence is therefore attributable to compensation for the original cost differences. If this compensation however were the main reason for the divergence, one would expect relative unit labour costs to stabilise, but that is not the case. For comparison, Figure 4.3 depicts how unit labour costs have developed since 1994. This is the year following the crisis in the European monetary system and a noticeable readjustment in exchange rates. It is in any case prior to the convergence of inflation rates that occurred as part of the Maastricht Process. Once again, there is no detectable stabilisation in relative unit labour costs. While unit labour costs have barely increased by 5% in Germany and Austria, they have grown in Portugal and Ireland by more than 45%. Figure 4.3
Nominelle Lohnstückkosten in EU12 (1994=100) Nominal Unit Labour Costs EU12 (1994=100) 1.5 Belgium
1.4
Germany (linked) Greece
1.3
Spain France
1.2
Ireland Italy
1.1
Luxembourg Netherlands
1
Austria Portugal
0.9
Finland
0.8 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
The logical conclusion is that wage growth rates did indeed converge, but that several countries, like Germany, have systematically had wage agreements that involved lower increases than in other countries. There has therefore been a divergence in the level of unit labour costs. In effect, this process cannot be described as an implicit wage coordination policy.
22
4.4 Wage Coordination What form could a system for an EU-wide wage coordination policy take? And what would its effects be within the framework of the European Monetary Union? The answer: a productivity-oriented wage policy at European level. Productivity-oriented wage policy means that the wages grow in step with inflation and with productivity. 11 The question arises whether to use national benchmarks for inflation and productivity or those of the euro area itself. Different constellations are conceivable. The declared strategy of the European Trade Union Confederation (ETUC) is that wages should rise in step with national productivity and the European (or more precisely the European Central Bank’s (ECB’s)) inflation target. That means that in formulating their wage demands, the unions would orient themselves to developments in Europe as a whole and not to national trends. That would greatly limit the unions’ flexibility to respond to current developments in their own countries. Such a restriction is almost inconceivable given current conditions. The necessary conditions are discussed later on in this study. First, the above argument is illustrated by means of the following general equation. Wj = b1.Xj + b2. XEU + b3. Pj + b4. PEU + b5.PEZB + b6. uj + b7. (LQEU - LQj) W .. Nominal wage growth; X … Productivity growth; P …Inflation rate; PECB … Inflation target of the ECB; LQ … wage share; u… unemployment rate; The superscript “EU” stands for the EU and the superscript j stands for the respective country.
In current national wage negotiation systems, only national benchmarks play a role. In the above equation b1 and b3 would therefore have values between zero and one. The European reference variables barely play a part: b2, b4, b5 and b6 are close to zero. A flexible labour market means that wages react strongly to (national) unemployment; b7 would then have a high value. In an EU-wide coordinated wage system of the kind advocated by the ETUC (but not necessarily embraced by the national unions), wages would respond to national productivity growth and the EU inflation target. b1 and b5 would be equal to one; the other parameters would be close to zero. 12 A system of this kind is probably not overly realistic, not even in the medium term. Employers have little interest in European wage coordination and the European Commission 11
The following comments are an attempt to explain the effects and thus potential problem of effective wage coordination in a monetary union. Simplifications are unavoidable. In particular, there is no discussion of the distinction between the homogenization versus the coordination of wage policy. It is also assumed that a national wage policy is describable with national wage agreements. In reality, sectoral wage agreements or even company-level agreements prevail. 12 Moreover, if national income distributions (and thus the unit labour costs) converge, countries with above average wage shares (and thus unit labour costs) would have to curb wage growth and those with below-average wage shares would have to increase wage growth. Otherwise, existing costs differences (and consequently, foreign trade imbalances) could be cemented in place. b7 would not equal zero under a strategy of this kind.
23
is advocating the exact opposite of wage coordination with its demand for the decentralisation of collective agreement negotiations. National unions would also have less latitude to act in a system of this kind. Greek employees, for example, would be compensated for the European inflation target only, but not for the (higher) rate of inflation in Greece. By the same token German employers would have to raise wages in step with the ECB inflation target instead of the (lower) rate of inflation in Germany. Resistance to the idea should come as no surprise. The biggest problem would be forgoing (or having only a limited possibility of) responding to national labour market developments. What makes this so difficult is that the current policy package in the EU puts the burden of adjustment to asymmetric shocks on wage policy. Monetary policy is uniform, the exchange rates are fixed and fiscal policy is greatly constraint (in most countries) because of the Stability and Growth Pact. In other words, the usual economic policy instruments are blocked. Wage policy is the only remaining adjustment mechanism or shock absorber so to speak. Under the current policy regime, countries with a current account deficit have to exercise wage moderation to re-balance their foreign trade position. 13 EU-wide wage coordination would have to entail a concurrent change in the policy package to reduce the burden on wage policy as a shock absorber. By definition, a monetary union excludes exchange rates as an instrument and has a common monetary policy. 14 Therefore, any change in wage policy must be accompanied by a newly defined role for fiscal policy, a role with greater latitude. Wage coordination is desirable with respect to the distribution function, the demand function and the price stability function. In terms of distribution, wage coordination could help end the decline in the wage share. One of the causes for this decline is that the single market has increased competitive pressures between national labour markets and, in turn, wage elasticity in labour demand (Rodrik 1997, Hatzius 2000). Wage coordination would at least counter this effect. Other causes such as globalisation and the weakness of unions would of course not be affected. Based on the findings presented here, even a partially successful stabilisation of the wage share would help to stimulate demand. Although modest and thus scarcely useful for cyclical policy, this effect would be substantial in scope in the medium term. The adjusted wage share has decreased by about 11.6 % over the past 25 years. We estimate the cumulative negative effect on demand to be approximately 2.2 % of GDP.
13
Italy is a prime example of a country that used devaluation in the past to stabilise its foreign trade position. With this tool no longer available, wage policy has to be used to achieve stabilisation. 14 While the refinancing rate (interest rate paid by banks for borrowed funds) is necessarily uniform, the same does not have to hold true for interest rates on loans. A system such as the one suggested by Palley (2006) is conceivable, namely, one in which collateralised loans would be assigned regionally different minimum reserve requirements. This approach would allow lending rates to be effectively raised in regions with overheated real estate markets. Such a system with differentiated reserve requirements would be particularly suitable for real estate loans. It is difficult to say how effectively this system would work in actual practice. The suggestion is in any case well worth pursuing.
24
Wage coordination would in principle be reconcilable with the target of price stability. A productivity-oriented wage policy does not cause inflationary pressures per se, provided employers accept it. If companies reject a stabilisation of the wage share and seek to increase the profit share by raising prices, that would naturally boost inflation. The problem then would be a lack of consensus on distribution policy, not the productivity-oriented wage policy. An inflationary effect would be probable in the short term, however. A coordinated, productivity-oriented wage policy would imply, for example, that the wage agreements in Germany would involve substantially higher wages. A reasonable way to ameliorate the problem, at least while wage coordination is being established, would be to interpret price stability as a monetary policy goal more flexibly instead of rigidly applying the current 2% benchmark, an arbitrary figure anyway from the economic point of view.
25
5. Literature Baumgartner, J, Breuss F, Kaniovski, S, 2005. WIFO-Macromod – An econometric model of the Austrian economy. In: OeNB (ed): Macroeconomic Models and Forecasts for Austria. Proceedings of OeNB Workshops No. 5 Bhaduri, Amit, Marglin, Stephen, 1990. Unemployment and the Real Wage: The Economic Basis for Contesting Political Ideologies. Cambridge Journal of Economics, 14: 375-93 Bowles, S, Boyer, R, 1995. Wages, aggregate demand, and employment in an open economy: an empirical investigation. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press European Commission 2006. Time to move up a gear. The new partnership for jobs and growth. http://ec.europa.eu/growthandjobs/pdf/illustrated-version_en.pdf Fagan, G, Henry, J, Mestre, R (2001). An area-wide model (AWM) for the Euro area. ECB Working Paper 42 Flassbeck, H, Spiecker, F, 2005. Die deutsche Lohnpolitik sprengt die Europäische Währungsunion. WSI Mitteilungen 12/2005, 707-13 Gordon, David, 1995. Growth distribution, and the rules of the game: social structuralist macro foundations for a democratic economic policy. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press Gordon, David, 1995. Putting the horse (back) before the cart: disentangling the macro relationship between investment and saving. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press Hatzius, Jan, 2000. Foreign direct investment and factor demand elasticities. European Economic Review 44 (1), 77-143 Hein, E, and Ochsen, C., 2003. Regimes of interest rates, income shares, savings, and investment: a Kaleckian model and empirical estimations for some advanced OECD-economies, in: Metroeconomica, Vol. 54, 404-433 Hein, E, Krämer, H, 1997. Income shares and capital formation: patterns of recent developments. Journal of Income Distribution 7, 1: 5-28 Marginson , P, Sisson, K (2004). European integration and industrial relations. Multi-level governance in the making. Houndsmill: Palgrave MacMillan Marglin, S, Bhaduri, A, 1990. Profit Squeeze and Keynesian Theory. In: S. Marglin and J. Schor (eds): The Golden Age of Capitalism. Reinterpreting the Postwar Experience. Oxford: Clarendon Marterbauer, M; Walterskirchen, E., (2003). Bestimmungsgründe der Lohnquote und realen Lohnstückkosten. Wifo-Monatsberichte 2/03, 151-59 Naastepad, R, Storm, S (2006/7), OECD demand regimes (1960-2000), Journal of Post-Keynesian Economics 29 (2) 213-248 Naastepad, Ro (2006). Technology, demand and distribution: a cumulative growth model with an application to the Dutch productivity slowdown. Cambridge Journal of Economics 30 (3) 403434 Palley, Thomas, 2006. Currency Unions, the Phillips Curve, and Stabilization Policy: Some Suggestions for Europe. Intervention 3 (2),351-370 Rodrik, Dani, 1997. Has globalization gone too far? Washington: Institute for International Economics Schulten, Thorsten (2004). Solidarische Lohnpolitik in Europa. Zur Politischen Ökonomie der Gewerkschaften. Hamburg: VSA-Verlag Stockhammer, E. and O. Onaran (2004), 'Accumulation, Distribution and Employment: A Structural VAR Approach to a Kaleckian Macro-model' Structural Change and Economic Dynamics 15 (4), 421-47
26
Data Appendix Appendix A: Data definition EU Model Notation C
Ameco Notation OCPH
E
-
ED
NWTD
ET Y YW
NETD OVGD -
I
OIGT
i
-
M
-
NX R P PM PX ULC RULC W
PVGD PMGS PXGS PLCD ALCD -
X
-
Designation Private final consumption expenditure at constant prices Exchange rate
Dependent employees, private sector Employees, private sector GDP, real Trade partners’ GDP, real
Gross fixed capital formation at constant prices (referred to in the study as “private investment”) Interest rate, long-term, real, based on GDP deflator Imports, real
Net exports, real Gross operating surplus, real GDP Deflator Import price deflator Export price deflator Unit labour costs Real unit labour costs Compensation of employees, real Exports, real
Calculation
Index (1960=100). Growth rate based on the weighted growth rates of euro exchange rates relative to trade partners’ currencies. Weights = Exports as a percentage of total EU exports in 2005 (Source: Eurostat). Trade partners: United States, Switzerland, Japan, Norway, Turkey. Exchange rate in direct quotation (€1 in the respective currency). Index (1960=100). Growth rates based on the weighted growth rates of the trade partners’ GDP. Weights = Exports as a percentage of total EU exports in 2005 (Source: Eurostat). Trade partners: United States, Switzerland, Japan, Norway, Turkey.
Source: OECD Economic Outlook: IRLR Goods imports of Euro12 countries from outside the EU. Belgium, Luxembourg from 1999. NX = X – M R = UOGD/PVGD W = UWCD/PVGD Exports of goods of Euro12 countries outside the EU. Belgium, Luxembourg from 1999.
Kammer für Arbeiter und Angestellte für Wien Prinz-Eugen-Straße 20-22, 1040 Wien, Telefon (01) 501 65
Functional Income Distribution and Aggregate Demand in the Euro Area
Institut für Geld- und Finanzpolitik University of Economics and Business Administration, Vienna (WU) Wien Study commissioned by the Chamber of Labour, Vienna
Functional Income Distribution and Aggregate Demand in the Euro Area
Engelbert Stockhammer [email protected] Institut für Geld- und Finanzpolitik University of Economics and Business Administration, Vienna (WU) Study commissioned by the Chamber of Labour, Vienna Vienna, March 2007
Acknowledgements The author is grateful to Paul Ramskogler and especially Stefan Ederer for their excellent work as project assistants and to Michael Mesch, Özlem Onaran, Werner Raza and Georg Zwiener for their ideas and suggestions. The study profited from synergies with the FWF Project No. P18419-G05.
Erhältlich bei:
Kammer für Arbeiter und Angestellte für Wien Abteilung EU und Internationales Prinz-Eugen-Straße 22-22 1040 Wien Tel: +43(01) 50165 – 2278 Fax: +43(01) 50165 - 2199 Email: [email protected] Internet: http://studien.arbeiterkammer.at
Das Werk ist urheberrechtlich geschützt. Die dadurch begründeten Rechte, insbesondere die Übersetzung, der Nachdruck, die Entnahme von Abbildungen, der Funksendungen, der Wiedergabe auf photometrischem oder ähnlichem Wege und der Speicherung in Datenverarbeitungsanlagen, bleiben, auch nur auszugsweiser Verwertung, vorbehalten. © 2007 Kammer für Arbeiter und Angestellte für Wien, 1040 Wien, Prinz-Eugen-Straße 22-22 Impressum: Medieninhaberin, Herausgeberin: Kammer für Arbeiter und Angestellte für Wien, 1040 Wien, Prinz-Eugen-Straße 22-22 Verlags- und Herstellort: Wien
Executive Summary An increase in the wage share has different effects on the various components of aggregate demand: ¾ It affects private consumption expenditure positively, because recipients of wage income have a greater propensity to consume than recipients of capital income. ¾ It has negative effects on (private) fixed capital formation, because profits (which perforce become relatively less on an increase of the wage share) have a positive effect. ¾ Net exports are also negatively impacted by an increase in the wage share, because unit labour costs rise with the wage share, which in turn reduces competitiveness. Given these positive and negative effects, it is theoretically impossible to provide a clearcut answer to the question of how an increase in the wage share affects demand a priori. Rather, two regimes are possible. If an increase in the wage share results in an increase in aggregate demand, the demand regime is referred to as “wage-led”; if it leads to a reduction in aggregate demand, it is called “profit-led”. Econometric methods are applied in this study to examine this issue for the euro area. The euro area is the prototype of a large, rather closed economy. The following picture emerges: ¾ The effects on consumption are much stronger than those on investments. The domestic sector is therefore clearly wage-led. ¾ Wage costs have a minimal impact in the euro area on net exports (relative to GDP). ¾ A 1% increase in the wage share in the euro area leads to a 0.2% rise in demand. Different methods of estimation yield similar results. However, the findings for the euro area are not directly applicable to its member states. A policy dilemma may arise. Wage moderation may well trigger short-term growth increases for certain (mostly small) countries, but if the same policy is adopted by all (EU) states, it clearly slows down demand and contributes to unemployment instead of reducing it. Wage moderation has dominated functional income distribution in Europe for about 25 years and unemployment has increased in that same period. The adjusted wage share in the euro area has fallen by 11.6 % since 1981, while the unemployment rate has increased by 1.2 %. The findings presented here help to explain this apparent paradox. At EU level, wage moderation curbs demand and therefore does not increase employment. One way out of this policy dilemma would be to coordinate wage policy throughout Europe (or within the euro area). This approach would help to stabilise income distribution. If broadly accepted, a coordinated, productivity-based wage policy could also be reconciled with the goal of price stability. It could have an inflationary effect in the transitional phase, however. 3
A coordinated wage policy, however, is not consistent with the current economic policy regime of the EU. Foreign exchange rates are now gone as an economic policy instrument, while monetary policy is uniform throughout the EU and fiscal policy is greatly limited by the Stability and Growth Pact. As a result, wage policy now bears most of the burden of adjusting the economy to asymmetrical shocks. If coordinated, wage policy would be less available as a tool for pursuing national economic policy goals. A coordination of wage policy would therefore also have to involve a change in the entire policy package in the EU, with fiscal policy taking on a more active role.
4
1. Introduction The wage share in the euro area has seen a sharp decline of 11.6 %-points since 1981, whereas the unemployment rate has increased a further 1.2 % since that year from an already high level (see Figure 1.1). 1 Europe has been exercising wage moderation for 25 years without seeing any resulting decline in unemployment. How is that possible? Doesn’t every textbook on microeconomics say that wage decreases lead to an increase in employment? The current Non-Accelerating Inflation Rate of Unemployment (NAIRU) theory also supports this thesis. But it is not just textbooks, old and new. The EU Commission also joins in the chorus supporting “employment-friendly labour cost developments and wage setting mechanisms” (EC 2006, 40). What it actually means is wage moderation. It is argued that moderate wage increases (coupled with dynamic productivity growth) strengthen Europe’s competitiveness and lead to economic growth and employment. Figure 1.1 Wage moderation without Employment Growth in the Euro Area 70
12
68 10
Adjusted Wage Share (market prices) 66 64
8
Unemployment rate
62 60
6
58 4 56 54 2 52
adj WS unempl.rate
50
0
1960 1962 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006
And who would dare dispute that wage moderation improves competitiveness? No one. But does improved competitiveness automatically mean more growth and employment? Foreign trade is just one part of aggregate demand. The other components are private consumption and private investment expenditures, public consumption and public investment expenditures. To understand the effects of wage moderation on demand in the overall economy, one must determine how all these components react to a change in income distribution.
1
Data source: AMECO, adjusted wage share at market prices
5
This empirical study focuses on precisely this issue for the euro area, namely how a change in functional income distribution affects aggregate demand. The euro area is a large, relatively closed economy. This study is structured as follows. First, section 2 discusses the theoretical background for the empirical study, namely, a neo-Kaleckian macro model that allows for wage-led and profit-led demand regimes. Then, econometric estimates for the euro area will be presented (in section 3). This section will analyse how a change in the wage share affects private consumption, private investment expenditures and foreign trade. Section 4 will discuss the economic policy implications of the findings. Wage coordination will be a pivotal point throughout. The discussion will explore implications for aggregate wage development and a design for and possible problems associated with wage coordination. Special attention will be paid to the role of wage policy in the economic policy regime of the EU. .
6
2. Theoretical Background: Wage-led and Profit-led Demand Regimes To examine the effect a change in functional income distribution has on aggregate demand, this study will make use of a general neo-Kaleckian model based on Marglin and Bhaduri (1990). In classical Kaleckian models (for closed economies), an increase in the wage share always results in increased demand. The model used here permits profit-led and wage-led growth, because it allows for a positive effect of profits on investment expenditures. The question of the overall effects that a change in functional income distribution has on demand thus becomes an empirical question. This chapter will present the theoretical background for this empirical study. Aggregate demand (Y) is the sum total of consumption (C), capital investment (I), net exports (NX) and government spending (G). All of these variables except government spending are a direct function of the profit share (π). Government spending and revenues can also react to changes in income distribution but this fact is not taken into account in the following discussion. The model (presented in Table 1) is therefore one of a private open economy and embodies several simplifications. Income distribution, namely the profit share (π), 2 is assumed to be exogenous. Feedback from growth affecting income distribution, i.e. regarding lower unemployment and a better negotiating position for employees, are not considered at this stage. 3
Table 2.1: Model Overview GDP
Y = C(π) + I(π) + NX(π) + G
Profit share
π == R/Y
Consumption
C = f (W, R)
Investments
I = f (R, Y, i)
Exports
X = f (ULC, PIMP,YW; E)
Imports
M = f (ULC, PIMP, Y)
Y: GDP; C: Private Consumption; I: Private Investments; NX: Net Exports; G: Government Spending; π: Profit share; W: Total Wages; R: Profits; i: Long-term Interest Rate in Real Terms; X: Exports; M: Imports; ULC: Unit Labour Costs; PIMP: Import Price Deflator; YW: GDP of other countries. 4
The effect an increase in the profit share has on aggregate demand and production is uncertain a priori and depends on the sum of reactions from the components of GDP, namely, consumption, investments and net exports. To determine how a change in functional income 2
Functional income distribution and its measure, the profit share, are used synonymously in this report. The wage share is by definition one minus the profit share. For theoretical modelling, it is therefore immaterial which ratio is used. 3 In terms of econometrics, the problem of simultaneity is ignored. 4 See Appendix A for more detailed definitions of the variables.
7
distribution affects consumption, income is broken down into wages (W) and profits (R) and the marginal propensity to consume is estimated for both. Since the propensity to save investment income is higher than for wage income, consumption can be expected to decline as the profit share rises. The investment function depends on profits (R), output (Y) and the long-term interest rate in real terms (i). Investments will increase as the profit share increases, because the expected profits become greater. In keeping with the standard theory, one expects a positive effect on output and a negative effect on interest rates. In an open economy, exports and imports must be included in the analysis. Exports are a function of unit labour costs (ULC), which determine the competitiveness of the domestic economy on the world market. By definition, unit labour costs are closely tied to the wage share. If unit labour costs rise, demand for domestic products will decline domestically and abroad, because domestic prices rise relative to foreign prices. Exports also depend on the GDP of the trade partner. This factor is therefore incorporated in the export function. By definition, unit labour costs have an inverse relationship to the profit share. Imports also depend on unit labour costs relative to the foreign price level and on domestic GDP. Exports should react positively to an increase in the profit share, because, by definition, this increase causes a decline in unit labour costs. If unit labour costs decline relative to foreign prices, domestic goods become more competitive on the world market. Exports are therefore stimulated. For this same reason, imports will react negatively to an increase in the wage share. As net exports are obtained by subtracting imports from exports, the total effect on net exports will presumably be positive. The expected sub-effects of a change in income distribution on demand 5 in an open economy are summarised below: •
∂C / ∂π < 0
•
∂I / ∂π > 0
•
∂NX / ∂π > 0
•
∂Y / ∂π = ?
The total effect of a change in functional income distribution on aggregate demand depends on the magnitude of the sub-effects and can only be determined empirically. If the effect is positive (∂Y/∂π > 0), the demand regime is said to be profit-led. If it is negative (∂Y/∂π < 0), demand is said to be wage-led. 6
5
To be precise, excess demand (refer also to Bowles and Boyer 1995). The above discussion pertains to the issue of whether total economic demand is wage-led or profit -led. In the introduction, the question was raised whether employment (or demand for labour) is wage-led or profit-led. The two issues are naturally closely connected. If wage changes affect labour productivity, e.g. as postulated by the efficiency wage theory, then a wage-led demand regime is a necessary but not sufficient condition for a wage-led employment regime. 6
8
Given the strength of international trade, one would empirically expect net exports to play a dominant role in determining the overall result. However, although individual countries can boost demand by expanding exports, the world as a whole cannot. It is therefore important to distinguish between the domestic sector of an economy and an open economy. In this case, the domestic sector is defined as the sum of consumption and investments, assuming no change in net exports (as would happen if wages changed in all countries simultaneously). If consumption reacts more sensitively than investments to an increase in the profit share, the domestic sector is said to be wage-led. The strategy applied for econometric estimation is similar to the one in Bowles and Boyer (1995). The model is estimated using the individual equations for saving, investments and net exports. The key divergences from Bowles and Boyer are as follows: First, the econometric specifications differ. Bowles and Boyer do not discuss any time series characteristics of the economic variables and neglect the subject of unit roots. Consequently, they use no difference or error correction models of the kind that form the core of modern time-series econometrics. Second, Bowles and Boyer use the employment rate as a proxy variable for capacity utilisation. For several European countries with long-standing unemployment, that is a misleading indicator. Therefore, output growth is used in this study instead. Brief mention should be made of the empirical literature on neo-Kaleckian models at this juncture. Gordon (1995a) estimates consumption and investments as a function of income distribution for the United States. In a vector autoregressive (VAR) model, numerous exogenous shocks are simulated. Gordon (1995b) expands this model to include the open economy and empirically examines how aggregate demand reacts to changes in functional income distribution in the United States. His conclusion is that the United States has a profitled growth regime. In their studies, both Hein and Krämer (1997) and Hein and Ochsen (2003) use a model for a closed economy based on Marglin and Bhaduri (1990). Hein and Ochsen (2003) expand the model to include the interest rate as an exogenous variable and work out various accumulation regimes depending on the sensitivity of the savings function and the investment function to the interest rate. In the empirical section of their studies, they estimate the savings function and the investment function econometrically and attempt to characterise the accumulation regimes of France, Germany, the United States and Great Britain. Stockhammer and Onaran (2004) estimate a structural VAR model consisting of variables for capital accumulation, capacity utilisation, profit share, unemployment rate and growth in labour productivity for the United States, the United Kingdom and France. They use a VAR model to take into account the reciprocal interaction of the variables. The goods market is simulated with a model based on the work of Marglin and Bhaduri (1990). It is supplemented with an equation for income distribution, a productivity function and a function for unemployment. The conclusion from the findings of the empirical study is that unemployment is determined by the goods market but that the impact of the income distribution on demand
9
and employment is very weak. Technical progress shifts income distribution in favour of profits. Onaran and Stockhammer (2005) apply a similar model to Turkey and Korea. Naastepad (2006) presents and estimates an expanded model for the Netherlands, in which the growth of labour productivity is explicitly modelled. Productivity, savings, investments and exports are estimated as individual equations. One finding is the Dutch demand regime from 1960 to 2000 was wage-led, albeit marginally so. Her overall conclusion is that Dutch demand has been relatively insensitive to changes in real wage growth since the end of World War II.
10
3. Empirical Study for the EU This chapter will empirically investigate the effects unleashed by changes in functional income distribution. To this end, functions for consumption, investment, and foreign trade will be econometrically estimated for the euro area utilising the model presented in Chapter 2. An attempt was made to keep modelling pragmatic and where possible, to assume broadly accepted behavioural functions and then expand them to include distribution variables. Unfortunately, there are hardly any reference models for the estimated behavioural equations. There is currently still a lack of macroeconomic studies on the central behavioural equations for the euro area (covering a prolonged period with annual data). One obvious reference point would be the ECB Area-Wide Model (AWM; Fagan, Henry and Mestre 2001). It is heavily theoretical, however, and contains a number of (theoretically motivated) restrictions that render the coefficients incomparable with the estimates presented here. For this reason, the WIFO Macro Model (Baumgartner, Breuss and Kaniovski 2005) was selected as the reference framework for estimations on the euro area. A pragmatic approach was also taken with regard to time-series specification. Where possible, use was made of error correction models (ECM), which currently constitute the most common time-series specification. Other specifications were used in cases where diagnostic statistics suggested their suitability; in practice, most of them were difference specifications. Income distribution was assumed to be exogenous in all estimations. Of course, this assumption is not completely accurate. In other words, the estimations do not take into account pro-cyclical mark-up prices (e.g. resulting from constant mark-ups to the normal unit costs) or the negative effects that unemployment has on wages. This approach is justified not by theory but merely by the limited scope of this study. Endogenizing income distribution would be the subject of a more extensive research project. The AMECO database (on the euro area) served as the database for estimations and contains the largest number of relevant time series of (aggregate) values for the Euro 12. The definitions of the variables can be found in the Data Base Appendix. The euro area is viewed as an entity for the entire period (approximately 1960 to 2004), even before the monetary union. Close attention was therefore paid to test for structural breaks.
3.1 Consumption Function The consumption function was estimated in the first differences (of the logarithms) after the autoregressive distributed lag (ADL) specification clearly suggested a difference specification. The Augmented Dickey-Fullter (ADF) test for residues did not suggest a cointegration correlation (at log levels). This finding is somewhat surprising, as the consumption function can usually be modelled effectively with error correction mechanism (ECM) models. 11
The hypothesis of different propensities to consume is clearly supported (see Table 3.1). It was estimated at 0.49 (statistically significant at the 1 % level) for wages and 0.11 (statistically significant just above the 10 % level) for profits. The coefficients represent elasticities and correspond to marginal propensities to consume of 0.52 (for wages) and 0.15 (for profits). The consumption differential between wages and profits thus amounts to 0.37. 7 A 1% redistribution from wages to profits therefore reduces consumption spending by 0.37 % of GDP. Table 3.1 Dependent variable: Variable C ln (W) ln (W(-1)) ln (R) ln (R(-1)) ln (C(-1)) Δln (W) Δln (R) Adj, R² DW stat.
ADL ln (C) Coeff. Prob. 0.16 0.68 0.38 0.00 -0.39 0.00 0.12 0.12 -0.11 0.12 0.98 0.00 1.00 1.62
Differences Δln (C) Coeff. Prob. 0.01 0.00
0.49 0.11 0.78 1.45
0.00 0.10
3.2 Investments The investment function was estimated as ECM (in keeping with the WIFO specification) by imposing the restriction that the investment ratio is constant in the long run, i.e. technically speaking that the investment and output coefficients are identical. In addition, profits were included as is customary in various recent investment equations (however, not in the WIFO macro model). Table 3.2 summarises the regression results. Besides the usual ECM specification, a specification with a more generous lag structure was estimated. This increased the profit coefficient and improved its statistical significance, which nonetheless occurs only at a level of 10 %. The elasticity of investments relative to profits is modest, with a long-term value of 0.15 (=0.03/0.21). The marginal effect is 0.07. An increase in profits by 1% of GDP leads ceteris paribus to an increase in investments of 0.07 % of GDP.
7
Taking the logarithms of the variables yields an unpleasant difference relative to the theoretical model. The latter is additive in levels, while the estimated model is additive in log levels. Implicitly, thus the consumption function is subject to the Cobb-Douglas functional form. The calculated consumption difference can therefore be viewed only as an approximation.
12
Table 3.2 Dependent variable: Variable C Δln (Y) Δln (Y(-1)) Δln (R) Δln (R(-1)) ln (I(-1)/Y(-1)) ln (R(-1)) i (-1)) Adj, R² DW stat.
ECM Δln (I) Coeff. Prob. -0.30 0.00 1.89 0.00 0.13
0.49
-0.14 0.01
0.01 0.61
0.72 1.73
ECM with lags Δln (I) Coeff. Prob. -0.59 0.00 1.59 0.00 0.84 0.03 0.38 0.06 -0.18 0.38 -0.21 0.00 0.03 0.07 -0.00 0.43 0.73 1.75
3.3 Net Exports Estimating the effect on net exports is probably the most sensitive part of determining the overall effect of a change in income distribution. It was therefore done carefully utilising a dual strategy. First, a net export function was estimated directly as a function of domestic GDP growth, of key trade partners’ GDP growth, of (nominal) exchange rates and of unit labour costs. The advantage of this method is two-fold: it is simple and it answers the relevant question. Its disadvantage is that it is not compatible with the foreign trade functions in current macro econometric models (such as the WIFO Macro Model). Wage costs are generally not integrated in the import and export functions directly but rather indirectly via prices. As a second method, exports and imports were estimated separately in a first step as a function of the price level (and other control variables). This procedure follows the WIFO model, which can be considered typical of modern modelling. In a second step, prices were regressed to the nominal unit labour costs (and import prices). The price equation implicitly contains information on how high a nominal wage increase must be to increase real unit labour costs by one percent. This latter fact is relevant here. Data definitions: There are data available on the exports and imports of goods from the euro area and non-euro countries, but unfortunately none for services. For this reason, all estimates on goods exports and imports and the subsequent results were weighted with a factor reflecting the relative magnitude of service exports (in 2005). The resulting error is presumably small. For the United States, there are data for goods and services covering a long period but no trend is evident. The value averages 1.25. GDP growth of the trade partners was calculated from the weighted GDP growth rates of the US, Switzerland, Japan, Norway and Turkey. The weights correspond to the relative shares in foreign trade involved (in recent years). In an analogous step, the exchange rate was calculated from the weighted growth rates of exchange rates.
13
Table 3.3 summarises the results of the direct estimation of the net export ratio, i.e. the first estimation strategy. All variables have the signs expected with the exception of the exchange rate, which is statistically significant at the 5 % level. Owing to the high autocorrelation in the residues from the original estimation equation, the estimation was repeated with an AR(1) correction. There are hardly any substantial changes in the results. Autocorrelation problems remain in spite of a marked improvement in the DW statistic. A one per cent increase in real unit labour costs reduces net exports by about 0.1 % of GDP. After the necessary transformation to take into account service exports and imports, this corresponds to an effect of -0.15 % of GDP with a 1 % increase in the wage share 8 . Table 3.3 Dependent variable: Variable C Δln (Y) Δln (YW) Δln (E) ULC Adj, R² DW stat.
Without AR1 corr. NX / Y Coeff. Prob. 0.06 0.00 -0.15 0.01 0.16 0.01 0.01 0.58 -0.11 0.00 0.43 0.68
With AR1 corr. NX / Y Coeff. Prob. 0.05 0.06 -0.18 0.00 0.12 0.01 0.01 0.28 -0.10 0.03 0.70 1.42
The second method for estimating the effect of a change in functional income distribution on net exports is more elaborate. In a first step, the exports are regressed to the relative prices (export prices relative to import prices) and to the output of the trade partners and exchange rate (relative to the main trade partners) (Table 3.4). As there were no indications of cointegration, a difference specification (with autocorrelation correction) was selected. The coefficients have the signs expected, but at a level of only low statistical significance (at 10 % for the GDP of the trade partners and at a level of less than 10 percent for the exchange rates). The coefficient for the relative prices is statistically significant at the 1% level and has an elasticity of -1.1. A difference specification also had to be used in the import regression. A dummy variable for the years following 1995 was also included to capture a structural break (albeit on an ad hoc basis). 9 The equation contains the GDP of the Euro 12 countries as well as the exchange rates and the relative prices. While economic growth has the sign expected and is statistically significant (at the 1% level), the exchange rate has a perverse sign (and is statistically significant at the 5% level). As neither specifications with different lag structures nor specifications without exchange rates brought about any improvement, the value from the original specification was taken although it is not statistically significant. The coefficient of the relative prices is small, at 0.3. Imports therefore react weakly to changes in the price level. 8
Unit labour costs in real terms were corrected to take into account the changes in the structure of employed persons. As the ratio of total employees to dependent employees is approximately 1.25, this figure must serve as the multiplier. Refer to Marterbauer and Walterskirchen (2003) for the larger discussion. 9 The inclusion of the dummy variable is primarily statistically motivated. A possible explanation for this break is the opening up of Eastern Europe markets, which increased imports.
14
Table 3.4 Exports Dependent variable: Δln X Variable Coeff. Prob. C 0.03 0.22 Δln (YW) 0.92 0.10 Δln (E) -0.28 0.07 Δln (PX/PM) -1.14 0.00 AR(1) 0.37 0.02 Adj. R² DW stat.
0.37 1.69
Imports Dependent var.: Δln M Variable Coeff. Prob. C -0.03 0.11 Δln Y 2.32 0.00 Δln E -0.33 0.02 Δln E (-1) Δln (P(-1)/PM(-1)) 0.29 0.13 T95 0.06 0.00 Adj. R² 0.46 DW stat. 1.74
In two auxiliary estimations, the price level was explained in terms of nominal unit labour costs, import prices and GDP growth, while export prices were explained in terms of domestic price level and import prices. Both estimations were conducted in differences and included autocorrelation correction. A one per cent increase in nominal unit labour costs increases domestic inflation by 0.54; a one per cent increase in domestic inflation increases export prices by 0.47. From the price equation, one can conclude that a 2.2% increase in nominal unit labour costs is required to increase real unit labour costs by 1 %. Prices would then rise by 1.2%. Table 3.5 Dependent variable: Variable C Δln Y Δln PM Δln ULC AR(1) Adj. R² DW stat.
Δln P Dependent variable: Coeff. Prob. Variable 0.02 0.01 C 0.16 0.04 Δln PM 0.03 0.11 Δln P 0.54 0.00 AR(1) 0.85 0.00 0.95 Adj. R² 2.12 DW stat.
Δln PX Coeff. Prob. 0.00 0.36 0.51 0.00 0.47 0.00 0.44 0.00 0.94 2.15
As the calculation of the effects of income distribution on exports and imports is composed of multiple steps, Table 3.6 summarises the relevant parameters to provide an overview. The export and import ratio display a trend, so that elasticities can be converted to marginal effects in various ways. This can be done either using the mean, which corresponds to the average effect of the period, or using the value of the final year (2005), which corresponds to the effect at the end of the period. The overall effect of functional income distribution on exports comprises the effect of real unit labour costs on prices and the effects of prices on export prices. This elasticity value is then converted to a marginal value and lastly, the unit wage costs are converted to the wage share and the result is weighted with the relative share of services as a percentage of total exports. In sum, a 1% increase in the wage share results in a decrease in exports equal to 0.07% of GDP. On average the effect over the observed period was only about 0.05. The analogous values for imports amount to 0.04 and 0.03 % of GDP, respectively. 15
Table 3.6 Exports ∂X/∂Px ∂Px/∂P ∂P/∂RULC 1/RULC X/Y 2005 -1.14 0.47 1.19 0.63 0.47 1.19 0.63 X/Y Ø -1.14 Imports ∂M/∂P ∂P/∂RULC 1/RULC M/Y 2005 0.29 1.19 0.63 1.19 0.63 M/Y Ø 0.29
X/Y 0.10 0.07 M/Y 0.11 0.08
ET/ED G&S / G 1.27 1.25 1.27 1.25 ET/ED G&S / G 1.27 1.25 1.27 1.25
Effect -0.07 -0.05 Effect 0.04 0.03
The effect of a one per cent increase in the wage share on the net export ratio is thus -0.11 in the final year and -0.08 on average. The difference between the two values indicates the increase that has occurred as a result of growth in foreign trade, i.e. globalisation. Both values are lower than the one obtained in the first direct estimation method. There is no obvious explanation for this fact. As will become apparent, however, this difference elicits no qualitative change in the overall result.
3.4 Total Effect The sub-results can now be brought together. The different methods used for calculating the effect on net exports are indicated separately. The (positive) effect an increase in the wage share has on private consumption is substantially greater (0.37% of GDP) than the (negative) effect it has on investments (0.07 % of GDP). The domestic sector of the euro area is therefore clearly wage-led. The net export effect is greater than that of investments (ranging from 0.07% to 0.15% of GDP depending on the estimation method employed). Table 3.7 presents three methods for estimating or calculating the net export effect. The first is a direct method, in which the net export ratio was regressed to real unit labour costs and other control variables. This approach yields the strongest effect on net exports. The second is an indirect procedure, in which exports and imports were regressed to the price level (and other variables) and in a second step, the price level was expressed in terms of nominal unit labour costs (and other variables). This method tends to correspond more closely to the customary modelling strategy currently in use. This method yields elasticities, which are then converted to marginal effects with the current export and import ratios (column “XM ratio 2005”) or the average (column “XM ratio Ø”) export and import ratios. The former indicates the net export effect “today” (0.1); the latter, the average effect (0.07) for the period as a whole. The values obtained with the different methods do vary, but not too greatly. For the overall interpretation, the most compelling method is probably the indirect one involving conversion with “today’s” effect at a value of 0.1. The total effect of a 1% change in the wage share fluctuates between +0.15% (when the net export ratio is estimated directly) and +0.23 % of GDP (using the indirect estimation method based on average foreign trade involvement). This effect is thus positive in each case. The qualitative result of the examination is that the euro area has a wage-led demand regime.
16
Wage moderation at EU level thus has negative demand-side effects on growth dynamics. The most plausible estimation of the demand effect in percentage points is 0.2% of GDP. Table 3.7
Consumption effect Investment effect Domestic effect Effect on net exports Total effect
NX ratio, direct 0.37 -0.07 0.30 -0.15 0.15
XM ratio, 2005 0.37 -0.07 0.30 -0.10 0.20
XM ratio, Ø 0.37 -0.07 0.30 -0.07 0.23
17
4. Implications for Wage Policy This chapter will explore the policy implications of the results from the empirical study on how demand is affected by changes in income distribution. First, the results will be summarised in a way that can be understood even without prior knowledge of econometrics. The fact that the euro area has a wage-led demand regime naturally has ramifications for various policy areas. The following section will focus especially on those pertaining to wage policy. Economically, wage policy has a variety of effects: distribution effects, price effects, demand effects and innovation effects. Wage policy can thus help to stabilise income distribution and prices and stimulate demand and innovation activity. This study is devoted primarily to the effects wage policy has on demand.
4.1 Summary of the Findings An increase in the wage share has different effects on the various components of aggregate demand: ¾ It affects private consumption expenditure positively, because recipients of wage income have a greater propensity to consume than recipients of capital income. ¾ It has negative effects on (private) fixed capital formation, because profits (which perforce become relatively less on an increase of the wage share) have a positive effect. ¾ Net exports are also negatively impacted by an increase in the wage share, because unit labour costs rise with the wage share, which in turn reduces competitiveness. Theoretically, it is therefore difficult to provide an non-ambiguous answer to the question of how an increase in the wage share affects demand, because there are positive as well as negative effects. That is why this question was subject to empirical examination for the euro area. The euro area is the prototype of a large, closed economy. The following picture emerges: ¾ In the euro area, the effects on consumption are much stronger than those on investments. The domestic sector is therefore clearly wage-driven. ¾ Wage costs have a minimal impact on net exports (relative to GDP) in the euro area. The euro area thus clearly has a wage-led demand regime. A 1% increase in the wage share consequently results in a 0.2% increase in demand. Different estimation methods yield similar results.
18
4.2 Wage Moderation and Employment in the EU – A Prisoner’s Dilemma The first and most direct conclusion that can be drawn in respect of economic policy is that the wage cuts at EU level undertaken to fight unemployment are unsuitable means for achieving this objective. Wage moderation reduces aggregate demand and will therefore not lead to an expansion in employment, but rather to the opposite. This finding is an important one, as it runs counter to the prevailing policy in Europe. The fact is wage moderation has been practised in the EU for 25 years without bringing about any reduction in unemployment. The EU Commission often talks of “employment-friendly wage-policy” but what it means is wage moderation in the name of competitiveness. As a wage reduction has a greater effect on domestic demand than on foreign trade, wage moderation in this case will not stimulate growth and employment but do the very opposite. Another finding of this study that is presumably just as important is that certain countries, especially small extra-verted countries, may see at least short-term growth if they exercise wage moderation. In small open economies, the effect on foreign trade may be greater than the domestic effect on demand. And in fact, several countries, notably the Netherlands and Ireland, have exercised wage moderation, often in the form of policy packages comprising promises of wage moderation from unions, of investments from companies and of social, educational and tax policy measures from the government. This approach appears to have yielded good economic results at least at first glance. In the Netherlands, a prime example of a country that uses wage pacts, wage moderation has resulted in a persistent erosion of productivity growth, which in turn undermines any competitive advantages that may have been gained (Naastepad 2006). A central conclusion of this study is that generalisations on these kinds of policies cannot be made at EU level. The main trade partners of these small open EU member states are naturally other EU member states. While individual instances of wage moderation may have an expansive effect in many individual (perhaps all) EU member states, the opposite is true at EU level. That may sound contradictory initially but has a simple explanation: The EU member states conduct the bulk of their trade within the EU. A country that practises wage moderation and exports to other countries thus competes domestic production out of the latters’ markets. In other words, it exports not only goods but unemployment. This is merely a version of the “beggar thy neighbour” policy, i.e. of effective devaluation by means of wage moderation. In the 1930s, this policy was pursued using exchange rates and custom duties and led to a collapse of world trade. Although there is no immediate risk of that right now, the present wage policy situation could be described as a prisoner’s dilemma. Even if wage moderation did have an expansive effect for each country, 10 the expansive strategy for 10
This idea is posited as a theoretical argument, not as a description. In fact, there appear to be some EU countries in which wage moderation has not had an expansive effect. The German economy has been stagnating for years even though it has the most moderate growth in unit labour costs of the entire euro area (refer to Figures 5.2 and 5.3) and high net exports. A possible explanation is that Germany has a wage-led demand regime.
19
all of them jointly would be to increase wages. As in all constellations described as prisoner’s dilemmas, cooperation and coordination are the two means of possible escape. In this case, wage coordination could be the key to solving the problem.
4.3 Current Developments in Wage Policy in the EU In all countries in the EU (and in most other OECD countries), there has been a clear trend over the past 25 years away from productivity-oriented wage policy to competition-oriented wage policy (Schulten 2004). The adjusted wage share (and unit labour costs) has consequently declined almost continuously since 1980. In their thorough discussion of labour relations in Europe, Marginson and Sisson (2004) conclude that there is virtually no wage coordination Europe-wide, indeed, not even transnationally (compare also Schulten 2004, Chapter 11). They do note, however, that the effective changes in real wages and unit labour costs have converged and therefore describe this situation as “convergence without coordination”. This might suggest that, as a result of inflation rate convergence in the course of and in preparation for the introduction of the euro, wage growth rates have also converged and that no additional action is needed. On the other hand, the increasing foreign trade imbalances within the EU are attributable not only to differing growth performances but also to German wage moderation (Flassbeck and Spiecker 2005). Have growth in wage rates and unit labour costs in fact converged? An initial glance at the standard deviation of growth rates in nominal wages for the euro area (Figure 4.1) would seem to indicate just that. The introduction of the euro did actually bring about not only a convergence in inflation rates but also a marked decrease in the standard deviation in wage growth rates. The picture is not yet complete, however. Although the range of wage growth fluctuation is narrower, this fact does not exclude the possibility that substantial cumulative divergences exist. The relevant factor for competitiveness is not wage growth (or the change in unit labour costs), but rather the level of unit labour costs.
20
Figure 4.1 Standard Deviation of Growth in Nominal Wages for EU12 0.050 0.045 0.040 0.035 0.030 0.025 0.020 0.015 0.010 0.005 0.000 1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
Figure 4.2 shows the development of unit labour costs since 1999. This figure does not indicate convergence. While unit labour costs have barely increased in Germany and risen by a mere 5% in Austria, they have grown in Portugal and Ireland by more than 25%. The trend does not indicate a stabilisation of relative unit labour costs but rather a more extensive divergence. Figure 4.2
Nominelle Lohnstückkosten in EU12 (1999=100) Nominal Unit Labour Costs EU12 (1999=100) 1.3 Belgium
1.25
Germany (linked) Greece
1.2
Spain 1.15
France Ireland
1.1
Italy Luxembourg
1.05
Netherlands 1
Austria Portugal
0.95
Finland
0.9 1999
2000
2001
2002
2003
2004
2005
2006
21
A different trend in unit labour costs relative to a (arbitrarily chosen) base year would not be a problem per se, unless unit labour costs were identical in the base year. Presumably, the different currencies did not accede to the monetary union with the “correct” exchange rates. Part of the divergence is therefore attributable to compensation for the original cost differences. If this compensation however were the main reason for the divergence, one would expect relative unit labour costs to stabilise, but that is not the case. For comparison, Figure 4.3 depicts how unit labour costs have developed since 1994. This is the year following the crisis in the European monetary system and a noticeable readjustment in exchange rates. It is in any case prior to the convergence of inflation rates that occurred as part of the Maastricht Process. Once again, there is no detectable stabilisation in relative unit labour costs. While unit labour costs have barely increased by 5% in Germany and Austria, they have grown in Portugal and Ireland by more than 45%. Figure 4.3
Nominelle Lohnstückkosten in EU12 (1994=100) Nominal Unit Labour Costs EU12 (1994=100) 1.5 Belgium
1.4
Germany (linked) Greece
1.3
Spain France
1.2
Ireland Italy
1.1
Luxembourg Netherlands
1
Austria Portugal
0.9
Finland
0.8 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006
The logical conclusion is that wage growth rates did indeed converge, but that several countries, like Germany, have systematically had wage agreements that involved lower increases than in other countries. There has therefore been a divergence in the level of unit labour costs. In effect, this process cannot be described as an implicit wage coordination policy.
22
4.4 Wage Coordination What form could a system for an EU-wide wage coordination policy take? And what would its effects be within the framework of the European Monetary Union? The answer: a productivity-oriented wage policy at European level. Productivity-oriented wage policy means that the wages grow in step with inflation and with productivity. 11 The question arises whether to use national benchmarks for inflation and productivity or those of the euro area itself. Different constellations are conceivable. The declared strategy of the European Trade Union Confederation (ETUC) is that wages should rise in step with national productivity and the European (or more precisely the European Central Bank’s (ECB’s)) inflation target. That means that in formulating their wage demands, the unions would orient themselves to developments in Europe as a whole and not to national trends. That would greatly limit the unions’ flexibility to respond to current developments in their own countries. Such a restriction is almost inconceivable given current conditions. The necessary conditions are discussed later on in this study. First, the above argument is illustrated by means of the following general equation. Wj = b1.Xj + b2. XEU + b3. Pj + b4. PEU + b5.PEZB + b6. uj + b7. (LQEU - LQj) W .. Nominal wage growth; X … Productivity growth; P …Inflation rate; PECB … Inflation target of the ECB; LQ … wage share; u… unemployment rate; The superscript “EU” stands for the EU and the superscript j stands for the respective country.
In current national wage negotiation systems, only national benchmarks play a role. In the above equation b1 and b3 would therefore have values between zero and one. The European reference variables barely play a part: b2, b4, b5 and b6 are close to zero. A flexible labour market means that wages react strongly to (national) unemployment; b7 would then have a high value. In an EU-wide coordinated wage system of the kind advocated by the ETUC (but not necessarily embraced by the national unions), wages would respond to national productivity growth and the EU inflation target. b1 and b5 would be equal to one; the other parameters would be close to zero. 12 A system of this kind is probably not overly realistic, not even in the medium term. Employers have little interest in European wage coordination and the European Commission 11
The following comments are an attempt to explain the effects and thus potential problem of effective wage coordination in a monetary union. Simplifications are unavoidable. In particular, there is no discussion of the distinction between the homogenization versus the coordination of wage policy. It is also assumed that a national wage policy is describable with national wage agreements. In reality, sectoral wage agreements or even company-level agreements prevail. 12 Moreover, if national income distributions (and thus the unit labour costs) converge, countries with above average wage shares (and thus unit labour costs) would have to curb wage growth and those with below-average wage shares would have to increase wage growth. Otherwise, existing costs differences (and consequently, foreign trade imbalances) could be cemented in place. b7 would not equal zero under a strategy of this kind.
23
is advocating the exact opposite of wage coordination with its demand for the decentralisation of collective agreement negotiations. National unions would also have less latitude to act in a system of this kind. Greek employees, for example, would be compensated for the European inflation target only, but not for the (higher) rate of inflation in Greece. By the same token German employers would have to raise wages in step with the ECB inflation target instead of the (lower) rate of inflation in Germany. Resistance to the idea should come as no surprise. The biggest problem would be forgoing (or having only a limited possibility of) responding to national labour market developments. What makes this so difficult is that the current policy package in the EU puts the burden of adjustment to asymmetric shocks on wage policy. Monetary policy is uniform, the exchange rates are fixed and fiscal policy is greatly constraint (in most countries) because of the Stability and Growth Pact. In other words, the usual economic policy instruments are blocked. Wage policy is the only remaining adjustment mechanism or shock absorber so to speak. Under the current policy regime, countries with a current account deficit have to exercise wage moderation to re-balance their foreign trade position. 13 EU-wide wage coordination would have to entail a concurrent change in the policy package to reduce the burden on wage policy as a shock absorber. By definition, a monetary union excludes exchange rates as an instrument and has a common monetary policy. 14 Therefore, any change in wage policy must be accompanied by a newly defined role for fiscal policy, a role with greater latitude. Wage coordination is desirable with respect to the distribution function, the demand function and the price stability function. In terms of distribution, wage coordination could help end the decline in the wage share. One of the causes for this decline is that the single market has increased competitive pressures between national labour markets and, in turn, wage elasticity in labour demand (Rodrik 1997, Hatzius 2000). Wage coordination would at least counter this effect. Other causes such as globalisation and the weakness of unions would of course not be affected. Based on the findings presented here, even a partially successful stabilisation of the wage share would help to stimulate demand. Although modest and thus scarcely useful for cyclical policy, this effect would be substantial in scope in the medium term. The adjusted wage share has decreased by about 11.6 % over the past 25 years. We estimate the cumulative negative effect on demand to be approximately 2.2 % of GDP.
13
Italy is a prime example of a country that used devaluation in the past to stabilise its foreign trade position. With this tool no longer available, wage policy has to be used to achieve stabilisation. 14 While the refinancing rate (interest rate paid by banks for borrowed funds) is necessarily uniform, the same does not have to hold true for interest rates on loans. A system such as the one suggested by Palley (2006) is conceivable, namely, one in which collateralised loans would be assigned regionally different minimum reserve requirements. This approach would allow lending rates to be effectively raised in regions with overheated real estate markets. Such a system with differentiated reserve requirements would be particularly suitable for real estate loans. It is difficult to say how effectively this system would work in actual practice. The suggestion is in any case well worth pursuing.
24
Wage coordination would in principle be reconcilable with the target of price stability. A productivity-oriented wage policy does not cause inflationary pressures per se, provided employers accept it. If companies reject a stabilisation of the wage share and seek to increase the profit share by raising prices, that would naturally boost inflation. The problem then would be a lack of consensus on distribution policy, not the productivity-oriented wage policy. An inflationary effect would be probable in the short term, however. A coordinated, productivity-oriented wage policy would imply, for example, that the wage agreements in Germany would involve substantially higher wages. A reasonable way to ameliorate the problem, at least while wage coordination is being established, would be to interpret price stability as a monetary policy goal more flexibly instead of rigidly applying the current 2% benchmark, an arbitrary figure anyway from the economic point of view.
25
5. Literature Baumgartner, J, Breuss F, Kaniovski, S, 2005. WIFO-Macromod – An econometric model of the Austrian economy. In: OeNB (ed): Macroeconomic Models and Forecasts for Austria. Proceedings of OeNB Workshops No. 5 Bhaduri, Amit, Marglin, Stephen, 1990. Unemployment and the Real Wage: The Economic Basis for Contesting Political Ideologies. Cambridge Journal of Economics, 14: 375-93 Bowles, S, Boyer, R, 1995. Wages, aggregate demand, and employment in an open economy: an empirical investigation. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press European Commission 2006. Time to move up a gear. The new partnership for jobs and growth. http://ec.europa.eu/growthandjobs/pdf/illustrated-version_en.pdf Fagan, G, Henry, J, Mestre, R (2001). An area-wide model (AWM) for the Euro area. ECB Working Paper 42 Flassbeck, H, Spiecker, F, 2005. Die deutsche Lohnpolitik sprengt die Europäische Währungsunion. WSI Mitteilungen 12/2005, 707-13 Gordon, David, 1995. Growth distribution, and the rules of the game: social structuralist macro foundations for a democratic economic policy. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press Gordon, David, 1995. Putting the horse (back) before the cart: disentangling the macro relationship between investment and saving. In: G Epstein and H Gintis (eds): Macroeconomic policy after the conservative era. Studies in investment, saving and finance. Cambridge: University Press Hatzius, Jan, 2000. Foreign direct investment and factor demand elasticities. European Economic Review 44 (1), 77-143 Hein, E, and Ochsen, C., 2003. Regimes of interest rates, income shares, savings, and investment: a Kaleckian model and empirical estimations for some advanced OECD-economies, in: Metroeconomica, Vol. 54, 404-433 Hein, E, Krämer, H, 1997. Income shares and capital formation: patterns of recent developments. Journal of Income Distribution 7, 1: 5-28 Marginson , P, Sisson, K (2004). European integration and industrial relations. Multi-level governance in the making. Houndsmill: Palgrave MacMillan Marglin, S, Bhaduri, A, 1990. Profit Squeeze and Keynesian Theory. In: S. Marglin and J. Schor (eds): The Golden Age of Capitalism. Reinterpreting the Postwar Experience. Oxford: Clarendon Marterbauer, M; Walterskirchen, E., (2003). Bestimmungsgründe der Lohnquote und realen Lohnstückkosten. Wifo-Monatsberichte 2/03, 151-59 Naastepad, R, Storm, S (2006/7), OECD demand regimes (1960-2000), Journal of Post-Keynesian Economics 29 (2) 213-248 Naastepad, Ro (2006). Technology, demand and distribution: a cumulative growth model with an application to the Dutch productivity slowdown. Cambridge Journal of Economics 30 (3) 403434 Palley, Thomas, 2006. Currency Unions, the Phillips Curve, and Stabilization Policy: Some Suggestions for Europe. Intervention 3 (2),351-370 Rodrik, Dani, 1997. Has globalization gone too far? Washington: Institute for International Economics Schulten, Thorsten (2004). Solidarische Lohnpolitik in Europa. Zur Politischen Ökonomie der Gewerkschaften. Hamburg: VSA-Verlag Stockhammer, E. and O. Onaran (2004), 'Accumulation, Distribution and Employment: A Structural VAR Approach to a Kaleckian Macro-model' Structural Change and Economic Dynamics 15 (4), 421-47
26
Data Appendix Appendix A: Data definition EU Model Notation C
Ameco Notation OCPH
E
-
ED
NWTD
ET Y YW
NETD OVGD -
I
OIGT
i
-
M
-
NX R P PM PX ULC RULC W
PVGD PMGS PXGS PLCD ALCD -
X
-
Designation Private final consumption expenditure at constant prices Exchange rate
Dependent employees, private sector Employees, private sector GDP, real Trade partners’ GDP, real
Gross fixed capital formation at constant prices (referred to in the study as “private investment”) Interest rate, long-term, real, based on GDP deflator Imports, real
Net exports, real Gross operating surplus, real GDP Deflator Import price deflator Export price deflator Unit labour costs Real unit labour costs Compensation of employees, real Exports, real
Calculation
Index (1960=100). Growth rate based on the weighted growth rates of euro exchange rates relative to trade partners’ currencies. Weights = Exports as a percentage of total EU exports in 2005 (Source: Eurostat). Trade partners: United States, Switzerland, Japan, Norway, Turkey. Exchange rate in direct quotation (€1 in the respective currency). Index (1960=100). Growth rates based on the weighted growth rates of the trade partners’ GDP. Weights = Exports as a percentage of total EU exports in 2005 (Source: Eurostat). Trade partners: United States, Switzerland, Japan, Norway, Turkey.
Source: OECD Economic Outlook: IRLR Goods imports of Euro12 countries from outside the EU. Belgium, Luxembourg from 1999. NX = X – M R = UOGD/PVGD W = UWCD/PVGD Exports of goods of Euro12 countries outside the EU. Belgium, Luxembourg from 1999.
Kammer für Arbeiter und Angestellte für Wien Prinz-Eugen-Straße 20-22, 1040 Wien, Telefon (01) 501 65
