Aggregate Demand, Aggregate Supply and Economic Growth

Abstract

While mainstream growth theory in its neoclassical and new growth theory incarnations has no place for aggregate demand, Keynesian growth models in which aggregate demand determines growth neglect the role of aggregate supply. By assuming that the rate of technological change responds to labour market conditions, this paper develops a simple and conventional growth model that integrates the roles of aggregate demand and aggregate supply. The model shows how the long‐run equilibrium growth rate of the economy, at which the unemployment rate is constant, can be affected by aggregate demand.

Key Words:

• Growth
• aggregate demand
• aggregate supply
• technological change
• Keynesian growth models
• hysteresis

JEL Classification:

• O41
• O33
• E12

Acknowledgement

The author is grateful to an anonymous referee for this journal, to Philip Arestis, Amit Bhaduri, Robert Blecker, Marc Lavoie, Jim Rakowski, Jaime Ros, Mark Setterfield, Peter Skott, and Lance Taylor, and to participants at a session at the Eastern Economic Association meetings at New York, March 2005 and at the conference on Understanding Economic Growth: New Directions in Theory and Policy, held at Downing College, Cambridge, 1–2 September 2005, for their useful comments and suggestions.

Notes

1. Just a few new growth theory models, referred to later, have some role for aggregate demand.

2. The difference between the two types of theories lies in which aggregate supply factors affect the long‐run rate of growth of the economy. Thus, the saving rate affects the long‐run growth rate in new growth theories, but not in the Solow model.

3. Unemployment can occur in the economy even without any distinction being made between saving and investment plans, for instance, with the rigidity of the real wage. In this paper the essence of the aggregate demand problem relates to the distinction between planned saving and investment, which brings about adjustments that take the economy away from its supply constrained growth rate.

4. In recent textbooks, including Barro & Sala‐i‐Martin (1995), Aghion & Howitt (1998), and Jones (1998), however, models incorporating aggregate demand are non‐existent. Indeed, the Harrod model is transformed into a Solow model with fixed coefficients!

5. These models have been called neo‐Keynesian (see Marglin, 1984; Dutt, 1990), post‐Keynesian (see Lavoie, 1995), and structuralist (see Taylor, 1991, 2004).

6. See, for instance, Dutt (1984, 1990), Lavoie (1992), Rowthorn (1982) and Taylor (1991, 2004).

7. Some models—see Rowthorn (1982) and Dutt (1990), for instance—assume that investment depends positively on technological change. In these models an increase in the rate of technological change can—but need not, if the response of investment to technological change is small—increase the rate of growth of per capita output. But it does so by increasing aggregate demand, and not because it increases aggregate supply. We abstract from these considerations for simplicity. Our subsequent analysis of the dual‐sided relationship between investment and technological change dynamics would not be qualitatively affected by its incorporation.

8. Lavoie (2003) discusses a variety of other mechanisms by which actual and desired or targeted rates of capacity utilization can be brought into equality while allowing the actual (and desired) rate of capacity utilization to be endogenous in the long run.

9. Take this mechanism to be income distributional changes, and assume, as in Kaldor–Pasinetti models that there are different saving propensities out of wage and profit income. For simplicity, assuming that wage income is entirely spent, and the fraction of profit income saved is s _π, and the profit share is σ, saving is given by S/K = s _π σ u, and investment by I/K = γ + β u. With u being at an exogenously given full capacity utilization level, u*, g = I/K = γ +β β u*, to which S/K adjusts because of variations in σ. This is basically Robinson’s (1962) growth model. Thus output growth depends on the parameters of the investment function, and there is no reason for output growth and hence employment growth to be equal to the rate of growth of labour supply.

10. These do not exhaust possible mechanisms. An alternative mechanism is suggested by Kaldor (1959), who assumes that, in the long run, changes in output and income result in a larger change in investment than in saving, so that excess demand leads to a cumulative increase in output till full employment is reached.

11. A rise in γ reduces dγ/dt, as required for stability.

12. Full employment here can be generalized to a natural rate of unemployment or a NAIRU rate.

13. See for instance, Blanchard & Summers (1987), Price (1988), and Dutt & Ros (2006).

14. An alternative expression of this given by the equation a = ζ(L/N) ^α , where ζ is a positive constant, and where L and N refer to employment and labour supply, which shows that labour productivity growth depends positively on the employment rate. I am grateful to Jaime Ros for this interpretation.

15. Technological change, of course, can be endogenized in a number of alternative ways, including Kaldor’s (1957) technical progress function. The implications, however, are not always the same, as pointed out in note 22 below.

16. Although we do not introduce the wage rate explicitly into the analysis, it is consistent with the notion that labour shortages exert an upward pressure on the wage, which leads to labour saving technological change. For empirical analysis using US data consistent with this view, see Marquetti (2004).

17. I am grateful to Robert Blecker for pointing out that these ideas are supported by the research of Rosenberg (1994).

18. The model can easily be extended to incorporate profit‐seeking research and development expenditures and learning by doing, as in standard new growth theory models, without altering the conclusions. The approach is also consistent with a human capital interpretation. A more rapid rate of growth of labour demand exerts upward pressure on the real wage (which is not explicitly incorporated in the model), inducing workers to obtain more education to take advantage of the higher wage of skilled workers who have more effective units of labour, thereby increasing the rate of technological change.

19. See Gandolfo (1971), for instance. I am grateful to Amit Bhaduri for suggesting this method of analyzing stability.

20. This is clearly not the only zero root model or even growth model available in the literature. Within the post Keynesian tradition, see Dutt (1997) for a discussion of two models, one involving the interaction between changes in short and long period expectations, and one involving the endogeneity of the desired rate of accumulation owing to the threat of entry of firms. These models, however, do not address the issue of the long‐run constancy of the unemployment rate, which is the main issue of the present model.

21. It is implicitly assumed that γ has two (multiplicative) components, one of which is a parameter, and the other adjusts according to (5).

22. Other ways of endogenizing technological change can also have similar implications (see Dutt, 2005). The Kaldorian technical progress function, which makes the rate of labour productivity growth depend on the rate of growth of the capital–labour ratio (that is, on capital deepening), however, does not have the implication that aggregate demand affects the long‐run rate of growth as long as the function has a slope of less than one, as assumed by Kaldor (1957). With a slope equal to one (so that there are no diminishing returns to capital deepening) aggregate demand does affect long‐run growth.

23. The trace of the system is given by f′ k/(s−β β) < 0, and the sum of the principal diagonal minors is given by [γ /(s−β)][h′−f′ (s/(s−β β))]. Since stability requires that this is positive, we require h′ to be small. Note that the determinant is zero, which makes the system lack a unique equilibrium. The model is discussed in more detail in Dutt (2005).

24. Allowing both a and n to adjust (the latter as a result of labour inflow from the subsistence sector) can provide some interesting insights. A large response in n will slow down the response in a, reducing the impact of an increase in aggregate demand on the rate of growth of per capita income.

25. A more complex version of the model presented in Palley (1997) proceeds along essentially similar lines and suffers from the same problems.

26. One version of the model in Palley (1996) attempts to introduce excess demand levels into the analysis by making excess demand affect investment and technological change, but it is unclear what determines output and aggregate demand, and hence excess demand in the first place.

27. Palley’s models also implies that Kaldorian technological change assumptions can imply that aggregate demand affects growth in the long run. Our analysis can be shown to imply that this is not the case—see Dutt (2005).

28. Palley (2003) also discusses a balance of payments constrained model. This model leaves unclear many issues, but because it departs from the closed economy framework of the present paper, is not discussed here.