Keynesian and Classical Approaches to Unemployment
Champernowne’s (1936) critical reaction to The General Theory represented an attempt to fill the gap of the determinants of changes in money wages in Keynes’s framework, which he identified as a weak spot in the argument of the book (Joan Robinson also criticized this aspect of Keynes’s analysis; see Boianovsky 2005). Champernowne accepted Keynes’s argument about the relation between the labour market and the market for goods, with its implication that a change in money wages can only affect the level of employment through its indirect effect on the determinants of effective demand, such as the rate of interest and the marginal propensity to consume. Champernowne suggested that Keynes’s assumption that workers are more sensitive to changes in money wage rates than to changes in prices only applies in the short run. As time goes by, workers realize that prices are changing and adjust their money wage claims accordingly. When there are no unanticipated price changes, the economy will settle at the ‘basic rate of unemployment’ decided by the equilibrium between the supply and demand curves for labour written as functions of real wages only. Champernowne pointed out that real wages will move in the same direction as money wages only if aggregate demand changes in the process. Champernowne’s 1936 comparison between the ‘classical’ and the ‘Keynesian’ analyses of unemployment was the first attempt to interpret The General Theory using equations and diagrams.
The explanation given by Champernowne (1936: 202) of money wage stickiness in the short run is based on the existence of contracts made on the expectation of a stationary cost of living and on workers’ habit of thinking in terms of the price level of the past. An unexpected rise in the price level will, therefore, shift the ‘real supply curve for labour’ (ibid.: 215) rightwards at every real wage rate, causing a reduction of real wage rates (since money wages increase by less than the rise in prices) and a higher employment level. An unexpected fall in the price level will have symmetrical short-run effects on real wages and employment (ibid.). But these are temporary effects, since workers will eventually repair their ‘oversights’ (ibid.: 204) of recent changes in the cost of living and will demand a money wage that ensures them the desired real wage at present prices, which means that the real supply curve for labour will shift back to its original position. Champernowne implicitly assumes that the real demand curve for labour is not affected by price changes, since entrepreneurs’ expectations of future prices are supposed to be correct (Keynes 1936 : Chapter 5; see also Champernowne 1964: 180-181).
The real wage rate that workers would demand if they forecast future prices correctly (i.e. under Champernowne’s assumptions, if the price level had been stationary) is called the ‘basic real wage’ (Champernowne 1936: 203) and the corresponding unemployment level is the ‘basic unemployment’ rate (ibid.). Since Champernowne implicitly assumes perfect mobility of labour and the absence of search in the labour market, his basic unemployment is the difference between the given labour force and the effective labour supply at the basic level of real wages, that is, voluntary unemployment.
Actual unemployment will be below or above basic unemployment depending on whether the price level is, respectively, rising or falling. Champernowne (ibid.: 204) calls ‘monetary unemployment’ the excess of actual unemployment over basic unemployment, and symmetrically for ‘monetary employment’. He states in a footnote that the concept of monetary unemployment is ‘copied from Keynes’s “involuntary unemployment,” but differs from that concept’ (ibid.: fn. 1). The difference, of course, is that while Keynes’s (1936 : 15) concept is an attempt to describe movements off the labour supply curve and situations of excess supply in the labour market, Champernowne has in mind points of short-run equilibrium of the labour market corresponding to different positions of the short-run labour supply curve. Furthermore, Keynes’s definition of ‘full employment’ as a fixed upper limit described by the absence of ‘involuntary unemployment’ is incompatible with Champernowne’s concept of ‘monetary employment’, when the economy is above the level given by the basic rate of unemployment (this is explained by rightward shifts of the labour supply curve caused by unanticipated rising prices).
Champernowne (1936) pointed out in his article that the actual rate of unemployment will diverge from the basic rate only temporarily, for workers will tend to repair their oversights of recent changes in the price level. The dynamics of this process is an important feature of his framework:
A period of monetary unemployment is likely to cause falling money-wages and...a period of monetary employment is likely to cause rising money-wages.
In so far as we can assume that rising and falling money-wages will respectively cause rising and falling real wages, we may conclude that a period of monetary employment contains the seeds of its own destruction in the form of a tendency for real wages to rise, whereas a period of monetary unemployment has in it the seeds of its own destruction, in the shape of a tendency for real wages to fall (ibid.: 204; italics added).
He endorsed Keynes’s (1936 : 262) view that a change in money wages, for a given state of effective demand, leads to a corresponding change in prices without any effect on employment and real wages. Champernowne eventually played a key role in convincing Pigou that this was an essential element of Keynes’s framework (see Pigou 1938: 134). In the context of his 1936 article, ‘the demand of labour for a certain real wage can only make itself effective in so far as it influences the attitude of the monetary authority and its manipulation of the rate of interest’ (Champernowne 1936: 204). Starting from a position of monetary employment—caused, as we saw above, by a temporary inability of workers to realize that prices have gone up— Champernowne argues that workers will demand higher money wages as their oversight is gradually repaired, which will lead to a rise in the price level in the same proportion. Workers will react by demanding another rise in money wages, which will be accompanied by another rise in prices. If real wage aspirations by wage earners are not attained in the short period, his model allows us to predict what may happen in the next short period, and so on (see also Harcourt 2001: 440). According to Champernowne (ibid.: 205), this steady rate of inflation will accelerate, since the ‘bargaining power’ of the representative worker increases as he becomes more confident in his ability to raise his money wage and, more importantly,
[t]he speed with which he will revise his demands in the face of increases in the cost of living will become greater the more accustomed he becomes to the danger of his real wage being reduced by the rise in the cost of living. We see that a period of monetary employment will be accompanied not merely by rising money-wages and prices, but moreover by money-wages and prices rising at a rapidly increasing rate (ibid.; italics added).
This is close to the famous accelerationist result of the ‘natural rate of unemployment hypothesis’ formulated by Milton Friedman (1968) and Edmund Phelps (1968). The only way to keep the rate of unemployment below its equilibrium or ‘basic’ level is by increasing the rate of inflation. As in Friedman and Phelps, the mechanism behind acceleration is the (not fully developed) assumption that workers form their expectation of price increases based on past rises in prices, that is, adaptive expectations. Furthermore, Champernowne suggests that the length of wage contracts is reduced as workers become more aware of the price increase process, which contributes to the acceleration of the inflation rate.
The acceleration process will eventually and indirectly bring about an increase in real wages, since the monetary authority ‘would be forced to put a stop to it’ (Champernowne 1936: 205) by means of a higher rate of interest. However, since the action of the central bank is not anticipated, there would probably be an overshoot, with real wages rising ‘sharply’ and the amount of employment falling below the level corresponding to basic unemployment (cf. Friedman 1968: 8-9 for essentially the same remarks about overshooting). The argument for the case of monetary unemployment is similar: the process of falling money wages and falling prices ‘is likely to become accelerated as labour becomes more disorganised by the depression, and as employers get more desperate and more confident in their power to cut money-wages’ until eventually ‘the lunacy of the situation will be realised’ (Champernowne ibid.: 205) and the monetary authority puts an end to the deflationary process, which leads to a reduction of the real wage rate.
The upshot is that periods of monetary employment or of monetary unemployment are not likely ‘to last for very long’, so that actual unemployment oscillates around the level of basic unemployment, which corresponds to the ‘trend value’ of unemployment (ibid.: 206). The length of depression periods tends, however, to be longer than for boom periods, since people are usually more sensitive to the threat of hyperinflation than to the prospect of a slump, which is reflected in the slower effects of deflations on the monetary policy of central banks (ibid.: 205-206). Champernowne (ibid.: 206) suggests that the ‘ordinary tools of classical analysis’ could be used when trend variables are the object of investigation and the assumption can be made that the amount of employment is not affected by the price level (ibid.: 206-209). The main tool of classical analysis is the ‘real supply curve of labour’ which may be a useful concept to estimate the trend of unemployment and of the interest rate, ‘provided that the monetary authority does not allow labour to be misled by too long periods of rising or falling cost of living’ (ibid.: 216). Classical analysis is not sufficient, however, if the rate of interest—because of scarce investment opportunities or increased liquidity preference caused by the fragile expectations of entrepreneurs—cannot be reduced by enough to avoid falling prices and monetary unemployment, which is Champernowne’s (ibid.) version of the so-called liquidity trap that would be later associated with Hicks’s (1937) investment saving-liquidity preference money supply (IS-LM) analysis (cf. Champernowne 1964: 191), where he takes into account the real balance effect but dismisses it as empirically small.
Champernowne used diagrams to illustrate his concepts, but he restricted diagrammatic analysis to the discussion of comparative statics, without any attempt to incorporate a diagrammatic representation of dynamic processes involving endogenous changes in expectations. Nevertheless, diagrams are especially useful to illustrate Champernowne’s suggestion (1936: 212) that the key difference between Keynes and the classics is that the former postulates that the supply of labour is also influenced by the money wage rate, not just by the real wage (see also Darity and Young (1995: 15-19), for a discussion of Champernowne’s equations and diagrams, and a comparison to the Hicksian IS-LM formulation). Another important difference between the classical and the Keynesian systems is, according to Champernowne (1936: 211-212), the impact of ‘general nervousness, the state of the news’ and price-level expectations on the demand for money and for investment in The General Theory, which he would stress even more in his 1964 reassessment. In the classical scheme, the starting point is the labour market with supply and demand curves in real terms, while the Keynesian scheme starts with the determination of the rate of interest, which decides the position of the ‘demand for saving’. Given the employment level and the corresponding real wage, ‘labour’s demand’ curve determines the money wage together with the vertical ‘employers’ offer’ curve. The slope of labour’s demand curve reflects the degree of money illusion of workers; it would be vertical if labour supply depended only on real wages (see ibid.: 215).
Champernowne’s diagrams bring out the different ‘logical structures’ underlying the classical and Keynesian analysis (ibid.: 209). Hence, ‘whereas the classical system of analysis was to deduce the levels’ of the endogenous variables ‘by considering in turn the demand and supply of labour, saving, and money, the Keynesian system is just the opposite, namely to consider the demand curves and supply curves for money, saving, and labour’ (ibid.: 212). Like Pasinetti (1974: 43-46), Champernowne deployed a system of equations of a causal type as opposed to a completely interdependent system of simultaneous equations in the manner of IS-LM. In particular, Champernowne (1936: 211), in contrast to Hicks (1937), described the influence of income on the money demand function as an ‘indirect effect’ only, which allowed him to emphasize the ‘direct effect’ represented by the influence of the rate of interest and expectations.
In his attempt to clarify Keynes’s theory of employment, Champernowne put forward the notion of a short-term trade-off between inflation and unemployment that would later be associated with the Phillips curve literature. As pointed out by Champernowne (1936: 201), under conditions of falling money wages in a depression, ‘a situation must eventually arise in which the monetary authority takes action to check any resultant fall in prices, and so makes effective the attempt of the unemployed to accept a lower real wage’. However, as Champernowne (ibid.: 216) was aware, it might be impossible ‘to lower the rate of interest sufficiently to cause sufficient investment to keep prices and money-wages from falling’ if pessimistic expectations of returns and general uncertainty prevailed. Instead of Keynes’s (1936 ) taxonomy of voluntary, frictional, and involuntary types of unemployment, Champernowne introduced the notions of monetary employment and monetary unemployment. In Champernowne’s short term, the labour market clears only temporarily, since money wages will change in the next period as workers adjust their price expectations, until real wages correspond to their expected level in the long term and unemployment is accordingly at its basic or trend level. Champernowne’s interpretation of labour market dynamics in the 1930s has clear implications for the formulation of economic policy. Although Champernowne (1936) did not explore this, one of the corollaries of his analysis is that a rate of unemployment below the basic rate is not a feasible goal of economic policy because it is associated with accelerating inflation.