Testing the Bhaduri–Marglin model with OECD panel data

Abstract

The Bhaduri–Marglin model is a post-Kaleckian model that allows one to study the impact of a functional income distribution on the growth in demand. Over recent years, a number of empirical studies based on this model have aimed at determining whether a redistribution towards profits harms or fosters demand growth. The focus so far has been on a very limited number of countries. This paper is the first to test the Bhaduri–Marglin model with panel data. It finds that demand growth is reduced by a redistribution towards profits in the average OECD country. Productivity growth is also impaired.

Keywords:

• Distribution
• demand growth
• productivity growth
• Bhaduri–Marglin model
• OECD panel data

JEL Classifications:

• C23
• E12
• E20
• O30
• O40

Acknowledgements

I would like to thank the two reviewers for their excellent suggestions. Any remaining errors are mine.

Notes

1. ‘In fact, the long-run trend is but a slowly changing component of a chain of short-period situations; it has no independent entity …’ (Kalecki 1968, 263).

2. This study also examines Italy and Spain.

3. Only Stockhammer and Stehrer (2011) work with quarterly data in order to examine higher than annual frequency effects of changes in distribution on demand. The rest of the literature uses annual data.

4. The emphasis is in the original. See Baltagi (2008, 6–11) for more advantages – as well as limitations – of panel estimation methods.

5. Government consumption is assumed away because it should not be affected by changes in the income distribution.

6. I follow Naastepad (2006) in assuming that foreign real unit labour cost () is equal to 1 and that animal spirits do not change, i.e. .

7. Growth rates are denoted by hats (^) over the symbols.

8. Remember that the growth rate of the wage share – or of real unit labour cost, respectively, ( ) – is equal to the difference between real wage growth () and productivity growth (.

9. If we want to determine the impact of real wage variations on demand ‘it is necessary to perform at least “thought experiments” based on exogenous variations in the real wage rate’ (Bhaduri and Marglin 1990, 376). In reality, however, the real wage rate is likely to be an endogenous variable. Schütz (2012) endogenises the real wage rate and shows that the basic conclusion on whether the demand regime of a country is wage-led or profit-led remains unaffected by whether the real wage rate is treated as exogenous or endogenous except when productivity growth increases strongly in response to growing real wages. An elasticity greater than one, necessary to produce what Schütz (2012, 315f.) calls a ‘paradoxical result’ according to which ‘an increase in the real wage rate leads to so much labor rationalization to cause an increase in the profit share’ has never been documented in the literature. A referee has pointed out that Lavoie (1992, 252f.) already makes a very similar point.

10. Stockhammer and Onaran (2004) include labour productivity in a structural vector autoregression (SVAR) model focusing on (amongst others) the effects of changes in functional income distribution on unemployment.

11. Referring to Cassetti (2003); Hein and Tarassow (2010) criticise this equation arguing that the wage share rather than the real wage should be used as cost push variable. In Cassetti’s (2003, 461) model, however, variations in the wage (profit) share come as a result of changes in the real wage.

12. See Naastepad (2006, 414f.) for a discussion of the profit-led case and her Appendix A for an examination of the necessary and sufficient conditions for a positive and stable growth equilibrium. A positive and stable equilibrium productivity growth rate always exists in a wage-led system. In a profit-led system, however, this requires that β₁* > – C and that

13. Arguably, Schütz’ proposal to replace output by the rate of capacity utilisation does not make much of a difference as he – in line with Bhaduri and Marglin (1990) – defines the rate of capacity utilisation as the ratio of output and potential (full capacity) output and holds the latter constant. In this case, variations in the rate of capacity utilisation are the same thing as variations in output.

14. Bhaduri and Marglin (1990, 384) state that ‘systematic analysis of these longer-term implications of wage-led economic expansion lies outside the scope of the present paper. Our focus here is entirely on the short period’.

15. Naastepad (2006, 410) argues that the parameter β₂ in equation (10(10) ) is a measure for ‘wage-induced technological progress’: it ‘measures the extent to which more expensive labour induces firms to intensify their search for and adoption of labour productivity-raising techniques’.

16. The sample covers all current 34 OECD member countries except Chile, Iceland and Turkey for which I found no employment data in the OECD’s databases. Information on the data sources is given in the appendix.

17. Recall that I follow Naastepad (2006) in assuming that = 1 and that .

18. The highest value for the propensity to save out of wages found in the single-country estimations of equation (15(15) ) was 0.18 (for France), see Bowles and Boyer (1995).

19. This means that the propensity to save out of wages varies between 0.114 (US) and 0.374 (Luxembourg).

20. The lowest value found in a single-country study was 0.34 (for the US), see Naastepad and Storm (2006–7).

21. ‘Real’ always means: in millions of national currency at constant prices, OECD base year. Since all real data enter the model in growth rates, a purchasing power parity (PPP) conversion is neither necessary nor desirable (see Ahmad et al. 2003).

22. Cross-section random effects were rejected by the Hausman test for correlated random effects. This makes sense given that the sample of countries can hardly be regarded as a random draw from the full sample of OECD countries.

23. This point is made very clear in Capaldo and Izurieta (2013).

24. I also estimated equation (10(10) ) using the hour-based productivity measure. The coefficients on and are also significant at the 1% level, albeit a bit smaller.

25. The real wage is defined as nominal wage income deflated by the GDP deflator and divided by the number of persons employed.

26. See Fair (1984, 210–214) for details of this procedure. As in Hartwig (2013), and u₋₁ (see Table 1) are used as instruments. The lagged dependent variable and lagged regressors are automatically added to the instrument list.

27. Against their own expectation, Hein and Tarassow (2010) find profit-led productivity regimes in four OECD countries for the period from 1960 to the early/mid 1980s.

28. Onaran and Galanis (2012, 8) suggest that ‘neoliberal policies have been implemented simultaneously in many developed and developing countries in the post-1980s period’.

29. Based on single-country regressions Onaran and Galanis (2012) find wage-led domestic demand regimes in all four non-OECD countries (Argentina, China, India and South Africa) they include. (These countries become profit-led when external demand is taken into account.)