1 Introduction

For many years, there was consensus in the constancy of labour and capital shares in national income. That is why the research interest has shifted from functional to personal income distribution. The labour share (or wage share) is the part of national income that goes to labour. Research reveals that the labour share has been falling for many years in most countries (Bassanini and Manfredi 2014; Guerriero 2012; Sweeney 2013). Although this fact has attracted the interest of scholars, international organisations and politicians worldwide, it remains an actual problem today. It should be mentioned that in recent years, the interest has centred more on the reasons for the decline in labour share than on the consequences of this decline.

This paper investigates the impact of the wage share changes on domestic demand in the countries of the European Union (EU). The decrease in wage share has a positive effect on profits and investment, and in turn, it reduces the labour cost, which has a positive effect on exports. However, consumption is expected to decrease because the marginal propensity to consume out of capital income is lower than that out of labour income. If the overall effect on aggregate demand is positive, the demand regime is deemed to be profit-led; if the effect is negative, it is considered to be wage-led.

Consumption is the largest part of the EU countries’ GDP. It also represents a large part of demand, which also has an impact on investment. Thus, a decline in the wage share can have a positive effect on investment owing to the increase in profits; however, in turn, it can have a negative effect owing to the decrease in demand (consumption). In the case of the EU countries, a positive and significant effect of the declining wage share on exports is unlikely as a much greater proportion of the EU member states’ total trade in goods (over 60%) is with partners inside the EU. If the wage share declines simultaneously for a large number of trading partners, the competitive gains of decreasing labour costs would be cancelled out and the wage share decline would depress domestic demand.

Most of the studies find that domestic demand is wage-led; however, with the impact on net exports having been assessed, the aggregate demand regime could be said to be profit-led. There is a lack of research that assesses the shift of the impact of labour share changes on the components of domestic demand—consumption and investment—in countries with different levels of openness to international trade. The scientific problem of this study is reflected in the following question: Does the impact of labour share changes on domestic demand and its components differ in countries with different levels of openness to international trade? This assumption is justified by theoretical arguments. For example, the decrease in labour share has a positive effect on investment owing to an increase in profits; however, conversely, it has a negative effect owing to a decrease in consumption, which constitutes the majority of demand. However, this negative impact should be smaller in more open countries because the decrease in domestic consumption can be compensated by the increase in foreign demand (exports).

The aim of this research is to assess the impact of wage share changes on domestic demand in the EU-country groups, which are determined by the individual countries’ levels of openness to international trade.

The uniqueness of this paper is that most of the research in this field centres on the effects of changes in the functional income distribution on demand, but they pay little attention to other important factors such as the saving rate or debt. So, the first contribution of this paper is that we compare the impact of labour share changes on domestic demand in the country groups with different levels of openness to international trade. The second contribution is that we follow Stockhammer and Wildauer (2015) and extend the Bhaduri–Marglin model to include private debt. The third contribution is that our regression estimates are based on a panel of 28 EU countries covering the period 1995–2014, whereas most of the research in this field relies on time-series analysis for individual countries (i.e. Onaran and Galanis 2014; Stockhammer et al. 2011; Stockhammer and Ederer 2008). To the best of our knowledge, Hartwig (2014) and Stockhammer and Wildauer (2015) were the first to develop an empirical model using a panel of OECD countries. Hartwig (2014) uses a panel of 31 OECD countries from 1970 to 2011 and finds that a decline in the wage share has a negative impact on demand. Stockhammer and Wildauer (2015) reach the same conclusion using a panel of 18 OECD member countries from 1980 to 2013. Kiefer and Rada (2015) use a panel of 13 OECD countries, but they find that the demand regime is profit-led.

Social and economic processes in the EU countries are relatively independent; however, they are much more integrated than the OECD countries. The EU not only connects the common market, which is established by standardised laws and regulations and assures the free movement of people, capital, goods and services, but it also has common policies for trade, agriculture and regional development. The EU countries participate in the common market and their international trade is regulated by the same rules; however, the level of openness of individual countries to international trade is different and this may affect the impact of the labour share on demand.

We find that the average demand regime in the EU is wage-led, as both consumption and investment are negatively affected by the wage share decline. Domestic demand remains wage-led in both the more open and the less open economies, but the impact of the decline in wage share on investment in the more open economies changes from negative to positive.

The remainder of the paper is organised as follows. Section 2 discusses the issues of measuring the labour share. Section 3 describes the impact that the wage share decline has on private domestic demand. Section 4 presents the data sources and research methods, and Sect. 5 presents the findings of the study. Section 6 compares our results with the previous findings in the literature and the final section concludes.

2 The concept of labour share and the measurement methods, including their limitations

Recently, much attention has been paid to labour share studies in the economics literature. The created value added or output (Y) is estimated as a function of the production factors—labour (L) and capital (K)—where Y = f(L, K). The functional income distribution shows how the output is distributed between these factors of production. Labour share is the fraction of income (of the created product) accruing to labour. Discussions have arisen about how to calculate the labour share. In empirical studies, the fraction accruing to labour is often defined as the share of the created product, or value added, paid to employees and it is often referred to as the wage share (Schneider 2011; Mihnenoka and Senfelde 2015).

One of the easiest ways to estimate the labour share is to calculate it as the ratio of the compensation of employees to GDP, but this measurement is not accurate. First of all, income of the self-employed is not included in the compensation of employees; therefore, the method of measurement reduces the labour share and unreasonably attributes the income of the self-employed to capital income. As Schneider (2011) notes, there are many discussions in the scientific literature on how to adjust the labour share, taking into account the income of the self-employed. The consensus is that the income of the self-employed includes components of both labour and capital incomes, but how should these incomes be divided between the two factors? Another question is how we should consider indirect taxes and subsidies. Should we do something to acknowledge that the government gets a share of the value added, or should we distribute this income between labour and capital? Another question for discussion is whether income should be measured on a gross or net value basis.

To answer these questions, scholars have proposed different adjustments to the labour share calculations. One of the proposed labour share calculation options (LS1) is not to divide the compensation of employees by GDP but by the net value added. Net value added refers to the GDP at market prices less taxes on production and imports, plus subsidies, minus the consumption of fixed capital.

$${\text{LS1 = }}\frac{\text{Compensation of employees}}{{{\text{GDP }} - {\text{taxes on production + subsidies }} - {\text{consumption of fixed capital}}}}$$
(1)

In this modification, the created value added is viewed from the position of the producer, i.e. income is estimated at factor costs. Taxes on production and imports are subtracted because they do not reflect any return of capital, whereas subsidies are added, as they accrue to companies. In accordance with the standard definition of capital income, capital income should be estimated and expressed as a net amount; therefore, the consumption of fixed capital is deducted. In other words, in accordance with this adjustment, labour share is calculated as the ratio of the compensation of employees and net value added. According to Bengtsson and Ryner (2014), net domestic product or net value added is not the right choice when compensation of employees is used in the numerator. The compensation of employees includes the employers’ social contributions, which refer to the value of social contributions incurred by employers in order to obtain social benefits for their employees’ needs in case of illness, disability, an accident at work, retirement or redundancy. These transfers represent reproduction costs for labour and, if we include these costs, then the replacement costs for capital (consumption of fixed capital) should be included too.

Labour share, which is calculated in the numerator using the compensation of employees, is referred to as the unadjusted labour or wage share because it does not include income of the self-employed and it is classified as capital income. Hereinafter, we present adjustments of the labour share calculation when calculating the labour share, in which a certain (or even the total) share of income of the self-employed persons is attributed to labour income. Labour share calculated in this way is referred to as the adjusted labour share. It should be noted that sometimes scholars use other adjustments when the denominator is not gross but net value added or GDP.

In the second case (LS2), the labour share calculation is adjusted by supplementing the numerator. Johnson (1954) recommends imputing two-thirds of self-employment income to labour income and the rest to capital income.

$${\text{LS2 = }}\frac{{{\text{Compensation of employees + }}\frac{2}{3} {\text{mixed income}}}}{{{\text{GDP }} - {\text{taxes on production + subsidies}}}}$$
(2)

However, it is not necessarily true that two-thirds of the self-employment income (mixed income) attributes to labour and one-third to capital, and these ratios can be different in individual countries.

In the third adjustment (LS3), all of the self-employment income is attributed to labour income (Kravis 1959). This adjustment is explained by the fact that the majority of the self-employed in developing countries provide services and their income may be treated as labour income.

$${\text{LS3 = }}\frac{\text{Compensation of employees + mixed income}}{{{\text{GDP }} - {\text{taxes on production + subsidies}}}}$$
(3)

However, even in developing countries, self-employed people generate and use capital; therefore, such a calculation is considered unrealistic and it overestimates the labour share (Guerriero 2012).

The essence of the LS4 adjustment is that the income of self-employed workers for both labour and capital is distributed in the same proportion as the economy’s total income (Atkinson 1983).

$${\text{LS4 = }}\frac{\text{Compensation of employees}}{{{\text{GDP }} - {\text{taxes on production + subsidies }} - {\text{mixed income}}}}$$
(4)

This adjustment is based on the assumption that the income distribution is the same in both private unincorporated enterprises and large corporations. This assumption, however, is not very realistic, as the labour share calculated in this way exceeds one in some countries (Bernanke and Gurkaynak 2001).

The main problem associated with the proposed adjustments (LS2, LS3 and LS4) is that data are required on self-employed (mixed) income, which is not available for all countries. Considering this situation, another adjustment of the labour share calculation has been recommended.

In the fifth case (LS5), an adjustment of the labour share was carried out by Gollin (2002), who assumes that, even in the absence of information about the income of the self-employed, statistics on the composition of the labour force is available.

$${\text{LS5 = }}\frac{{\frac{\text{Compensation of employees}}{\text{Number of employees}} \times {\text{Total employment}}}}{\text{GDP }}$$
(5)

As Gollin (2002) emphasises, this assumption is appropriate only if earnings of the self-employed and corporate workers are similar. This approach for the labour-share calculation is used by the OECD and European Commission.

In the sixth case (LS6), Guerriero (2012) suggests an adjustment that removes the employers’ income because it is likely that the incomes of employers and employees are not comparable.

$${\text{LS6 = }}\frac{{\frac{\text{Compensation of employees}}{\text{Number of employees}} \times ( {\text{Total employment}} - {\text{number of employers)}}}}{\text{GDP }}$$
(6)

The question of the measurement of labour share is particularly relevant in studies that compare different countries because variations in the calculation of wages or the compensation of employees might exist in the national statistics. In 2013, the European Commission approved the calculation of the adjusted labour share (called the adjusted wage share), which was according to the labour share adjustment recommended by Gollin (2002), i.e. LS5. The adjusted labour share is calculated using GDP at both market prices and factor prices. According to Guerriero (2012), the calculation using GDP at factor prices is more appropriate because taxes do not reflect any return of capital. Since this adjusted wage share has been approved by the European Commission and used in recent studies about the impact of labour share on demand in the EU countries (Stockhammer and Wildauer 2015; Onaran and Obst 2016), it is also used in this study.

3 The impact of the wage share’s decline on consumption and investment

Most macroeconomic models pay little attention to the effects of income distribution on consumption and investment. Neoclassical macroeconomic models emphasise that income distribution has no impact on consumption, as the marginal propensities to consume out of wages and profits are equal. In a classical closed-economy Kaleckian model, a decline in the wage share always results in a decline in aggregate demand (Blecker 1989) because the marginal propensity to consume out of wages is higher than that out of profits. Wage income is concentrated in lower income households that have a relatively higher propensity to consume, whereas the rich get the major part of capital income and save a relatively large part of it (ILO 2013). Bhaduri and Marglin (1990), Bowles and Boyer (1995) and Stockhammer et al. (2009) estimate that the difference between the marginal propensity to consume out of wages and that out of profits is around 0.4. The decrease in wage share has a negative impact on consumption, as a major part of profits are retained by firms and, hence, cannot be consumed (Stockhammer et al. 2011).

Despite declining wage shares, consumption has been growing in many economies. This is explained by the fact that consumption is fuelled, in large part, by increasing household debt rather than rising wages. Stock market and housing price booms, in addition to changing financial norms and new financial instruments, have made increasing amounts of credit available to low-income households, so debt has become a substitute for higher wages as a source of demand and consumption (ILO 2013).

The negative impact of the wage share’s decline on consumption can be outweighed by its positive impact on investment. Investment is expected to increase when the wage share falls because future profits are expected to rise. In addition, it is often argued that retained earnings are a financial source that can influence investment expenditure (Stockhammer et al. 2011).

The Kaldorian branch of post-Keynesian economics does not agree that investment should be positively affected by a decrease in the wage share. Kaldorian models emphasise the accelerator rather than profitability, believing that firms would not invest more if profits went up and if there is no increase in demand. The level of investment is determined by the adjustment of capacity to exogenous demand (Caldentey and Vernengo 2013).

4 Data and methodology

Domestic demand (Y) is the sum of consumption (C), investment (I) and government expenditure (G). Government expenditure can react to income distribution; however, as a rule, this is ignored in studies on the impact of income distribution on demand.

Our dataset covers 28 EU member countries from 1995 to 2014 on an annual basis. Cluster analysis of the 28 EU countries, according to the trade openness index (the ratio of exports and imports to GDP), reveals two clusters of countries. These clusters are designated ‘more open countries’ and ‘less open countries’.

Consumption (C) and investment (I) are estimated as a function of income (Y), wage share (WS) and some other control variables (summarised as Z). Then, domestic demand is

$${\text{DD }} = {\text{ C(Y}},{\text{ WS}},{\text{ ZC) }} + {\text{ I(Y}},{\text{ WS}},{\text{ ZI)}}$$
(7)

All variables are in real terms. The variables and data sources are provided in the “Appendix: Data definitions and sources”. Household consumption expenditure is determined by the actual and expected changes in the income of consumers, as well as the consumers’ ability to use credit and spend their future income now. According to the permanent income hypothesis, households consume more (save less) in a given period if they expect their future incomes to increase. However, a precautionary response to possible income risks might lead to lower consumption (higher savings). While some consumers might increase their consumption because they save a smaller fraction of their income, others might stimulate consumption because of increased borrowing. An increase in household debt also leads to an increase in income and, thus, consumption. In our research, consumption (C) is estimated as a function of income (Y), wage share (WS) and the saving rate (SR) or household debt (HD).

Private investment is modelled as a function of income or output (Y), wage share (WS), the real interest rate (RIR) and debt. Aggregate demand (or output) and long-term real interest rates are standard in investment functions. The interest rate represents the price of investments and it is expected to have a negative effect. If firms have confidence in future demand and profits, they borrow money and invest, despite an increase in the interest rate. In contrast, negative expectations might slow down investments, even when the interest rate is falling. Corporate lending is a financial source for investment, so we use corporate debt as a control variable in our investment function. As variables of debt, we use CD or PD. CD represents the debt of non-financial corporations and PD is the total private debt of both non-financial corporations and households. We include the debt of households, as total investment consists of both business and residential investments.

Panel data models are used in this study. The advantage of the application of panel models is that we use both cross-sectional and time-series variations in the data. In order to determine whether the non-stochastic and time-unvarying unobservable effects do not correlate with the independent variables, we apply the first differencing (FD) method. When applying this method, the rate values depend on both groups and time, i.e. when t = 1, 2…; N = 1, 2… (in this case, N = 28 countries and t = 20 years). All indicators are not available for all countries in all years, so, in some model specifications, N is smaller. This model is applied to the panel data and, in general, is described by the following equation:

$$\Delta y_{it} = \, \alpha \, + \delta_{3} td3_{t} + \cdots + \, \delta_{T} tdT_{t} + \beta_{1} \Delta x_{it1} + \cdots + \beta_{k} \Delta x_{itk} + \Delta e_{it}$$
(8)

where y is the dependent variable, α is a constant, δ represents the effects (secular changes) that are common for all countries (as cyclical economic shocks, etc.), β is the independent variable’s effect on the dependent variable, td is the time fixed-effect variable, x is an independent variable and e is an error term.

The interaction between the wage share and country openness is used here because we believe that openness shapes the impact of the wage share on C and I. The FD model with an interaction variable is described by the following equation:

$$\begin{aligned} \Delta y_{it} & = \, \alpha \, + \, \delta_{3} td3_{t} + \cdots + \, \delta_{T} tdT_{t} + \beta_{1} \Delta x_{it1} + \beta_{1D} \Delta x_{it1} D_{i} \\ & \quad + \beta_{2} \Delta x_{it2} + \cdots + \beta_{k} \Delta x_{itk} + \Delta e_{it} \\ \end{aligned}$$
(9)

where D is a dummy variable, which is equal to 1 for more open countries and 0 for less open countries. β1 shows the impact of the wage share on the dependent variable (e.g. consumption) in less open countries and β 1D shows the difference of the x impact on y in more open countries compared with less open ones. The effect in the more open-country group is calculated as β 1  + β 1D .

In assessing the impact of the wage share on consumption and investment, the following issues have to be verified:

  • Whether the parameter coefficients and symbols of the model are compatible with economic logic.

  • Whether the model errors comply with the underlying assumptions. The absence of autocorrelation is verified by testing the hypothesis H0: ρ (ΔYi,t; Δui,t−1) = 0. Autocorrelation is eliminated using AR (1), i.e. the model is estimated, including the lagged dependent variable. To test for heteroscedasticity, we use the White test. To increase the robustness of the estimation in the case of heteroscedasticity, we use the weighted least squares (WLS) method. The weight is equal to \(1/\sqrt {\text{squared residual}}\). The weight is set according to the OLS regression estimates error variation.

  • Whether the assumption of data distribution normality is satisfied (Chi square statistics are used).

  • Whether there are no multicollinearity problems. To detect multicollinearity, we apply the variance inflation factor (VIF).

Model specifications with lagged explanatory variables are estimated to test the robustness of our results. Generally, the results from the alternative estimations were similar to those presented in this paper and, in the case of different results, they did not change the overall effect of the wage share on domestic demand by very much.

Calculating the marginal effects, we follow the methodology presented in Stockhammer and Wildauer (2015). The marginal effect of the wage share on domestic demand is computed in the following way:

$$\frac{\partial DD}{\partial WS}\frac{1}{Y} = \beta_{C,WS } \left( {\emptyset \frac{C}{Y}} \right)\frac{1}{\emptyset WS} + \beta_{I,WS} \left( {\emptyset \frac{I}{Y}} \right)\frac{1}{\emptyset WS}$$
(10)

\(\beta_{I,WS}\) and \(\beta_{I,WS}\) are the estimated elasticities of consumption and investment spending with respect to the wage share.

\(\emptyset \frac{C}{Y}\) and \(\emptyset \frac{I}{Y}\) represent the GDP-weighted average (based on PPPs) of C/Y and I/Y of the 28 countries included in the panel. First, we compute the GDP weighted averages of C/Y and I/Y for each year. Next, we compute the simple averages of these yearly averages.

A change in wage share will lead to changes in consumption and investment spending. The total effect of the change in wage share depends on the relative size of the reactions of consumption and investment to the changes in income distribution. If ∂DD/∂WS > 0, the demand regime is called wage-led. If the effect is negative, i.e. ∂DD/∂WS < 0, it is called profit-led.

5 Empirical results

5.1 The impact on consumption

Consumption (C) is estimated as a function of income (Y), wage share (WS) and the saving rate (SR) or household debt (HD). The function takes the following form:

$$\begin{aligned} \Delta {\text{ln(C}}_{{{\text{i}},{\text{t}}}} ) { } & = \, \alpha \, + \, \delta_{ 3} {\text{td1997}}_{\text{t}} + \cdots + \, \delta_{ 20} {\text{td2}}0 1 4_{\text{t}} + \beta_{ 1} \Delta {\text{ln(WS}}_{{{\text{i}},{\text{t}}}} ) { } \\ & \quad + \beta_{ 2} \Delta {\text{ln(Y}}_{{{\text{i}},{\text{t}}}} ) { } + \beta_{ 3} \Delta {\text{ln(SR}}_{{{\text{i}},{\text{t}}}} ) { } + \, \Delta {\text{e}}_{\text{it}} \\ \end{aligned}$$

The estimates are presented in Table 1.

Table 1 Consumption function, dependent variable: Δln(Ci,t)
Full size table

As expected, the decrease in wage share has a negative impact on consumption, with an elasticity in the range of 0.146–0.185. By comparison, Stockhammer and Wildauer (2015) estimate that a 1% decrease in the wage share has a negative effect on consumption by about 0.14% in OECD countries. In our case, a 1% decrease in the wage share lowers consumption by about 0.15% if we do not take the saving rate and household debt into account. However, this model can underestimate the effect of the wage share, as a decrease in consumption due to falling wages can be compensated by an increase in household debt and a decrease in the saving rate. If we take the latter variable into account, a 1% increase in the wage share lowers consumption by approximately 0.17–0.19%.

Table 2 provides the results of our assessment of the impact of the wage share on consumption expenditure, where we enter the variable of the interaction of the wage share and the country’s openness, i.e. it verifies whether the impact of the wage share on consumption differs between more open and less open economies.

Table 2 The impact of wage share on consumption expenditure (ΔlnCi,t) in less open and more open economies
Full size table

We conclude that the impact of wage share changes on consumption expenditure is higher in more open countries. When the wage share declines by 1%, consumption expenditure decreases by 0.115–0.165% in less open countries and by 0.224–0.251% in more open countries. The difference in the impact can be explained by the fact that the difference between the marginal propensity to consume out of wages and that out of profits is higher in more open countries.

5.2 The impact on investment

Private investment is modelled as a function of income or output (Y), wage share (WS), the real interest rate (RIR) and debt. We use CD or PD as the variables of debt. CD represents the debt of non-financial corporations and PD represents the total private debt of both non-financial corporations and households. The investment function takes the following form:

$$\begin{aligned} \Delta {\text{ln(I}}_{{{\text{i}},{\text{t}}}} ) { } & = \, \alpha \, + \, \delta_{ 3} {\text{td1997}}_{\text{t}} + \cdots + \, \delta_{ 20} {\text{td2}}0 1 4_{\text{t}} + \beta_{ 1} \Delta {\text{ln(WS}}_{{{\text{i}},{\text{t}}}} ) { } \\ & \quad + \beta_{ 2} \Delta {\text{ln(Y}}_{{{\text{i}},{\text{t}}}} ) { } + \beta_{ 3} \Delta {\text{ln(RIR}}_{{{\text{i}},{\text{t}}}} ) { } + \beta_{ 4} \Delta {\text{ln(CD}}_{{{\text{i}},{\text{t}}}} ) { } + \, \Delta {\text{e}}_{\text{it}} \\ \end{aligned}$$

The estimates are presented in Table 3.

Table 3 Investment function, dependent variable: Δln(Ii,t)
Full size table

First, the model confirms the assumption that the real interest rate has a negative impact on investment. The elasticity is similar to that estimated by Stockhammer and Wildauer (2015): a 1% increase in the real interest rate has a negative 0.424% effect on investment. In all model specifications, income has a very strong impact on investment spending, with an elasticity above 2. These results are in line with those of Hein and Vogel (2008), Onaran and Galanis (2012) and Stockhammer and Wildauer (2015).

The results show that, if the wage share decreases by 1%, investment spending decreases by 0.17–0.195%, contrary to the theoretical assumptions of the Kaleckian models, which emphasise that a decrease in the wage share means an increase in profits and, thus, that it stimulates future investments (Stockhammer and Stehrer 2011). Our findings support the Kaldorian approach, which places the emphasis on demand rather than profitability, believing that firms would not invest more if profits went up and there is no increase in demand (Caldentey and Vernengo 2013). If the corporate and household debt is added (specification 3), the wage share impact on investment is reduced from 0.195 to 0.17%.

Data presented in Table 4 show how the investment reaction to labour share and demand changes differ between the countries of more open and less open economies.

Table 4 The impact of wage share on investment (ΔlnI) in less open and more open economies
Full size table

We find that the impact of wage share changes on investment is stronger in the country group of less open economies. It is interesting that it has been established in a single sample that, when the wage share decreases, investment decreases too; however, the direction of impact changes in the countries with more open economies (except the first specification of the model). If the wage share declines by 1%, investment is reduced by 0.213–0.236% in less open economies, whereas, in more open economies, when we control for the impact of debt, investment increases by 0.028–0.046%. Investment reaction to the wage share changes is associated with the fact that a decline in the wage share increases the profit share, which encourages investment; but, on the other hand, the declining wage share reduces consumption and demand, creating a negative impact on investment. The final result depends on which effect is crucial. It seems likely that the investments in more open economies are less responsive to the changes in domestic consumption because a relatively (as compared with less open economies) greater part of production is realised in foreign markets and, therefore, the negative effect of decreasing profits becomes critical. On the contrary, demand in less open economies is more dependent on consumption; therefore, the demand factor and not the profit factor is more important in the group of less open countries.

5.3 The total effect of changes in the wage share on domestic demand

Table 5 summarises the partial effects of a 1 percentage point decrease in the wage share on consumption and investment based on the coefficients from specification (3) in Tables 1 and 2 for EU-28 and Tables 3 and 4 for more open and less open countries.

Table 5 Marginal effect of a 1 percentage point decrease in the wage share on domestic demand (% of GDP)
Full size table

For the EU-28 sample, the negative effect of a decrease in the wage share by one percentage point on private consumption is 0.168 percentage points of the GDP. The effect is substantially larger than that on investment (0.056 percentage points of the GDP). So, it becomes clear that the domestic sector of the EU economy is wage-led. Our results suggest that a 1 percentage point decrease in the wage share (assuming no changes in government expenditure and income) leads to a decrease in domestic demand of 0.225 percentage points of the GDP. The effect in the countries of less open economies is higher than that in the countries of more open economies; but, in both groups of countries, domestic demand is wage-led. Although the impact of the wage share decline on investment becomes positive in more open economies, it is not sufficient to offset the negative impact on consumption, so the domestic demand regime remains wage-led.

6 Comparison with the literature

Our results are in line with most studies that conclude that domestic demand is wage-led, i.e. the effect of a pro-capital redistribution of income on domestic demand is negative. Hartwig (2014) and Stockhammer and Wildauer (2015) estimate the average demand regime in a panel of OECD countries and find it to be wage-led. Research covering several individual OECD countries also finds wage-led domestic demand regimes for most countries. Naastepad and Storm (2007) use data for eight OECD countries, Hein and Vogel (2008) analyse six OECD countries, and Stockhammer and Stehrer (2011) use data for 12 OECD countries. The results of these studies show that most OECD countries have wage-led demand regimes. But the results by Naastepad and Storm (2007) show that two major nations, Japan and the United States, are profit-led domestically. Nevertheless, these results can be considered as exceptions among many others that confirm the wage-led nature of domestic demand in Japan (Bowles and Boyer 1995; Stockhammer and Stehrer 2011; Onaran and Galanis 2012, 2014) and the United States (Bowles and Boyer 1995; Hein and Vogel 2008; Stockhammer and Stehrer 2011; Onaran and Galanis 2012, 2014).

The results for the EU countries also conclude that the domestic demand regime in most countries is wage-led. The Euro area, in aggregate, is wage-led (Onaran and Galanis 2012; Stockhammer et al. 2009), as well as the EU-15 countries (Onaran and Obst 2016).

As regards investment, our results are in alignment with those of Onaran and Galanis (2014) and Onaran and Obst (2016), who conclude that private investment is not very sensitive to an increase in profits, but it responds strongly to demand. For the panel of 28 EU countries, we do not find evidence that a decrease in the wage share has a positive impact on investment. On the contrary, the effect is negative. These results are in line with Stockhammer and Wildauer’s (2015) evidence for the panel of OECD countries. Hein and Vogel (2008) also conclude that there were insignificant and/or negative effects of a profit-share increase on investment in a few countries under investigation. However, in the group of more open countries, the assumptions of the Kaleckian models confirm that a decline in labour share promotes investment.

7 Conclusions and discussion

We analysed the impact of changes in income distribution on private domestic demand. The analysis was based on panel data for 28 EU countries and inspired by the Bhaduri and Marglin (1990) model, which allows for profit- or wage-led demand regimes. Overall, our results clearly show that the domestic sector has a wage-led nature in the EU. This conclusion is valid in both the more open and the less open groups of countries. Most of the studies reviewed in our paper also find that domestic demand in most of the countries under investigation is wage-led. In this paper, we focus on domestic demand for two reasons. First, consumption is an engine of economic growth and constitutes a major part of the countries’ GDP. When the wage share decreases, consumption also decreases and it can have a negative impact on investment, which is sensitive to demand. We find that output (or demand) has a very strong impact on investment spending in the EU. Another problem arising from a decrease in the wage share is that consumption becomes debt-led instead of wage-led, and an economy may turn towards an unsustainable growth pattern based on debt growth.

We find that the impact of a wage share decline on investment in more open and less open countries is significantly different. Declining wage shares have a negative effect on investment, both in the case of the overall EU and in less open countries. Therefore, we conclude that investment is more responsive to the changes in domestic demand, not in the profit share or foreign demand. In the group of more open countries, the results are opposite: decreasing wage shares promote growth in investment. This can be explained by the fact that, in more open economies, investment is less responsive to the changes in domestic consumption because a relatively (as compared with the less open economies) greater part of production is realised in foreign markets and, therefore, the positive impact of capital (profit) growth becomes crucial.

The demand regime can change to profit-led when the effects of the distribution on foreign trade are taken into account. The conclusions from other research indicate that this is only true for small open economies and not for larger, less open economies. The second reason (why we want to emphasise domestic demand) is that the EU, as a single entity, is a very large, relatively closed, economy. A much greater proportion of the member states’ total trade in goods (over 60%) is with partners inside the EU-28. If the wage share declines simultaneously for a large number of trading partners, no one has a competitive advantage owing to a decrease in labour costs, but domestic demand is depressed. The recent studies, however, present mixed results when considering the effects of the distribution on net exports. Before drawing economic policy conclusions, further investigations should be carried out for each EU member state to gain a more complete understanding of how a wage share decline affects aggregate demand in each country. Taking into consideration the results of this paper, it looks like wage-led strategies are more promising.