On wage- and profit-led demand regimes: learning from the evidence

Abstract

This paper contributes to the empirical research around the “wage-led” or “profit-led” demand regimes. It first reviews how Kalecki, and then Steindl, approached the relationship between economic growth and income distribution. Then, empirical analysis carried out under the probabilistic approach to econometric modeling shows statistical evidence, estimated through cointegration analysis, that in the long run, in three very open economies—Mexico, France, and Korea—the wage share is positively associated with demand and output. It finally discusses the macroeconomic dilemma that almost all countries have to face, i.e., a positive effect of a high-wage policy on demand and employment may diverge from a negative effect on output compatible with external equilibrium.

Keywords:

• demand-led regimes
• economic growth
• empirical models
• real exchange rate
• wage share
• Mexico
• France
• Korea

JEL Classification::

• C5
• F4
• E12

Notes

1 As far as we know, Bowles and Boyer (1995) were also the first authors to empirically study the issue. Bowles S., Boyer R. Wages, Aggregate Demand, and Employment in an Open Economy: An Empirical Investigation. In ‘Macroeconomic Policy after the Conservative Era’ by G. A. Epstein and M. G. Gintis (eds.), Cambridge University Press, 1995.

2 Richard Goodwin, another heterodox economist, is an “outlier” in this debate. He proposed a classical model based on the predator-prey system, where wage flexibility gives rise to cycles rather than full employment (Goodwin 1967). Goodwin latter recollected (Palazzi 1982, 40), “When I discussed for the first time my 1967 paper in a seminar at Cambridge, Joan Robinson … dismissed my work…. She said … that my argument was pre-Keynesian, for the basic hypothesis was ultimately that of full employment…” (We thank Prof. Gugliemo Chiodi for this reference and for the English translation of the paragraph). In our view Robinson was right, because in Goodwin’s model the problem of effective demand is completely absent.

3 To simplify, we omit taxes and overheads, as well as workers’ savings.

4 An algebraic relation exists between the concepts of degree of monopoly, profit margin, and mark-up. Here we will use the three terms indistinctly, because they have an exact, positive association

5 Kalecki knew that the price fall might bring about a reduction in the short-term interest rate, however he considered this effect to be of secondary importance. In his celebrated rebuttal to Pigou he added that falling prices might rather cause a recession (Kalecki 1944 [1990]).

6 He remarked: “The two cases differ only in that in the former the wages decline and the prices of imported raw materials remain unchanged, while in the latter the wages remain unaltered (in terms of domestic currency), and the prices of imported raw materials increase in inverse proportion to the currency depreciation” (Kalecki 1939, 38). De Jesús and López (2020) have empirically shown the negative association between the wage share and the real exchange rate in a paper including the cases of France, Korea and Mexico.

7 Details on this point can be obtained from the authors upon request.

8 The probabilistic approach to econometrics provides a methodology based on statistical tests to represent the empirical regularities present in the data. If equation (4)(4) $$\mathrm{Y}=\frac{\mathrm{I}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+{\mathrm{C}}^{\mathrm{K}}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+\mathrm{T}\left(\mathrm{Y},{\mathrm{Y}}^{\mathrm{*}},\mathrm{ }\mathrm{q},\mathrm{ }\mathrm{I}\right)+{\mathrm{B}}^{\mathrm{d}}\left(\mathrm{G},\mathrm{Y}\right)}{\left(1\hspace{0.17em}-\hspace{0.17em}\mathrm{\omega }\right)}.$$(4) as represented in a VAR model is not reject by the statistical tests, then we can infer that this is a valid representation of the data generation process. Moreover, if cointegration test do not reject, then it follows that there is at least one linear linear combination between the variables represented in the VAR model. Finally, identification tests can be applied to infer which among the stable linear combinations can be normalized for GDP.

9 We define here real exchange rate q as: $q_t=e_t\frac{p_t^{\mathrm{*}}}{p_t},$ where $e_t$ indicates the domestic currency nominal exchange rate vis-à-vis the foreign currency and $p_t^{\mathrm{*}},$ $p_t$ indicate the foreign and domestic price indexes, respectively. Thus, competitiveness improves when the real exchange rate rises.

10 It may be asked why we do not include workers’s savings in equation (4)(4) $$\mathrm{Y}=\frac{\mathrm{I}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+{\mathrm{C}}^{\mathrm{K}}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+\mathrm{T}\left(\mathrm{Y},{\mathrm{Y}}^{\mathrm{*}},\mathrm{ }\mathrm{q},\mathrm{ }\mathrm{I}\right)+{\mathrm{B}}^{\mathrm{d}}\left(\mathrm{G},\mathrm{Y}\right)}{\left(1\hspace{0.17em}-\hspace{0.17em}\mathrm{\omega }\right)}.$$(4) . But this variable can be omitted because we may assume that workers’s savings depend on GDP and on the wage-share, two variables that are already included in equation (4)(4) $$\mathrm{Y}=\frac{\mathrm{I}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+{\mathrm{C}}^{\mathrm{K}}\left(\mathrm{\omega },\mathrm{Y},\mathrm{R},\mathrm{M}\right)+\mathrm{T}\left(\mathrm{Y},{\mathrm{Y}}^{\mathrm{*}},\mathrm{ }\mathrm{q},\mathrm{ }\mathrm{I}\right)+{\mathrm{B}}^{\mathrm{d}}\left(\mathrm{G},\mathrm{Y}\right)}{\left(1\hspace{0.17em}-\hspace{0.17em}\mathrm{\omega }\right)}.$$(4) .

11 We carried out unit-root tests, as well as single-equation and vector misspecification tests. To save space, below we only report vector misspecification tests and misspecification tests for the GDP equations. But all tests, together with our database, are available to interested readers upon request.

12 The identification of the long-term structures of all the models was carried out according to Johansen and Juselius (1994) and Doornik and Hendry (2009); and they were estimated in the Pc-Give 14 module of OxMetrics 7.

13 Due to lack of space we do not report here the estimated Vector Error Correction Model. We point out however that for the three countries under study we found that the estimated systems were stable, and that in equations (4) variables in the right-hand side Granger-cause GDP.

14 In our model for France we did not use natural logarithm for the real exchange rate. In the source we are using most of its values were lower than one, and using natural logarithms might have induced a sign change in the estimated coefficient.

15 Our choice of the selected period was conditioned by the availability of long time series for the variables required for this study.

16 In most cases our results differ from those from other studies dealing with the countries considered here. We may mention Onaran and Galanis (2014); who estimate models including Mexico (profit led), France(wage-led), and Korea(wage-led). For France, Bowles and Boyer (1995); Ederer and Stockhammer (2007); Naastepad and Storm (2006); Hein and Vogel (2008); Stockhammer and Stehrer (2011); Onaran and Obst (2016). Finally, for Korea we must mention Onaran and Stockhammer (2005), who allegedly found a wage-led growth. We do not discuss the results from those papers because, as previously mentioned, we found they had statistical problems that imply that their results cannot be taken at face value.

17 Note, Keynes, added to this the difficulty of modifying the real exchange rate in the modern world, where wages are closely linked with the cost of living.

18 López (2013) carried out econometric research with a methodology similar to ours for the US, France, the United Kingdom, Spain, and Germany, and also found a positive association between GDP and the wage share.

19 The consequences of currency depreciation on the wage share are analogous to a fall in the nominal wage. In the text, we use the case of a wage fall as an example, even though this is rather infrequent situation.

20 In the following paragraphs we adapt to the case of a wage fall the reasoning from López and Malagamba-Moran (2017). In formal terms, the authors found that in equation (2a), $k_{\overline{\omega }}<0$ and $j_{\overline{\omega }}<0,$ where $k_{\overline{\omega }}$ and $j_{\overline{\omega }}$ are the partial derivatives of k and j with respect to $\overline{\omega },$ respectively.

21 For our estimates, we defined output at external equilibrium $Y^{\zeta }$ as: $Y^{\zeta }=\frac{X}{\mu };$ $\mu =\frac{M}{Y},$ where $X$ is total exports, $M$ is total imports and $\mu$ the import coefficient.

22 As mentioned, higher wages may also cause inflation.

23 An anonymous referee offers us the following comment: “In Korea, President Moon Jae-in has been trying to increase the minimum wage since 2017. Nonetheless, Korea now faces a declining growth rate and an increase in the unemployment rate. This fact seems to relate to your argument[s]”. Unfortunately, we do not have sufficient knowledge of the Korean economy to go deeper into the issue. But we note that the wage share is only one among the many variables influencing output and employment in our model. If it so happens that in Korea a wage rise has been accompanied with declining output this, we believe, must be the consequence of the negative impact of other variables on effective demand.

Additional information

Notes on contributors

Verónica De Jesús

Verónica De Jesús is at División Académica de Ciencias Básicas, Universidad Juárez Autónoma de Tabasco, Cuanduacán, Tabasco, México.

Julio López

Julio López is at División de Estudios de Posgrado Facultad de Economía, Universidad Nacional Autónoma de México, Mexico City, Mexico.