Авторы: 147 А Б В Г Д Е З И Й К Л М Н О П Р С Т У Ф Х Ц Ч Ш Щ Э Ю Я

Книги:  180 А Б В Г Д Е З И Й К Л М Н О П Р С Т У Ф Х Ц Ч Ш Щ Э Ю Я


5.5.6 The Lucas critique of econometric policy evaluation

The final implication of the new classical approach for the formulation of

macroeconomic policy concerns what is popularly known as the ‘Lucas critique’,

after the title of Lucas’s seminal paper in which the proposition first

appeared. Lucas (1976) attacked the established practice of using large-scale

macroeconometric models to evaluate the consequences of alternative policy

scenarios, given that such policy simulations are based on the assumption

that the parameters of the model remain unchanged when there is a change in

policy. The Keynesian macroeconometric models developed during the 1950s

and 1960s consisted of ‘systems of equations’ involving endogenous variables

and exogenous variables. Such models, following Koopmans (1949),

contain four types of equation referred to as ‘structural equations’, namely:

1. identities, equations that are true by definition;

2. equations that embody institutional rules, such as tax schedules;

3. equations that specify the technological constraints, such as production


4. behavioural equations that describe the way in which individuals or

groups will respond to the economic environment; for example, wage

adjustment, consumption, investment and money demand functions.

A good example of this type of ‘system of equation’ model is the famous

FMP model (named after the Federal Reserve–MIT–University of Pennsylvania

model) constructed in the USA by Ando and Modigliani. Such models

were used for forecasting purposes and to test the likely impact of stochastic

or random shocks. The model builders used historical data to estimate the

model, and then utilized the model to analyse the likely consequences of

alternative policies. The typical Keynesian model of the 1960s/early 1970s

was based on the IS–LM–AD–AS framework combined with a Phillips curve

relationship. Obviously the behaviour of this type of model will, among other

things, depend on the estimated value of the coefficients of the variables in

the model. For example, such models typically include a consumption function

as one of the key relationships. Suppose the consumption function takes

the following simple form: C = + (Y – T). That is, consumption is

proportional to disposable (after tax) income (Y – T). However, in this simple

Keynesian consumption function the parameters (, ) will depend on the

optimal decisions that economic agents made in the past relating to how

much to consume and save given their utility function; that is, these parameters

were formed during an earlier optimization process directly influenced

by the particular policy regime prevailing at the time. Lucas argues that we

cannot use equations such as this to construct models for predictive purposes

because their parameters will typically alter as the optimal (consumption)

responses of rational utility-maximizing economic agents to the policy changes

work their way through the model. The parameters of large-scale macroeconometric

models may not remain constant (invariant) in the face of policy

changes, since economic agents may adjust their expectations and behaviour

to the new environment (Sargent, 1999, refers to this as the problem of

‘drifting coefficients’). Expectations play a crucial role in the economy because

of the way in which they influence the behaviour of consumers, firms,

investors, workers and all other economic agents. Moreover, the expectations

of economic agents depend on many things, including the economic policies

being pursued by the government. If expectations are assumed to be rational,

economic agents adjust their expectations when governments change their

economic policies. Macroeconometric models should thus take into account

the fact that any change in policy will systematically alter the structure of the

macroeconometric model. Private sector structural behavioural relationships

are non-invariant when the government policy changes. Thus, estimating the

effect of a policy change requires knowing how economic agents’ expectations

will change in response to the policy change. Lucas (1976) argued that

the traditional (Keynesian-dominated) methods of policy evaluation do not

adequately take into account the impact of policy on expectations. Therefore,

Lucas questioned the use of such models, arguing that:

given that the structure of an econometric model consists of optimal decision rules

of economic agents, and that optimal decision rules vary systematically with

changes in the structure of series relevant to the decision maker, it follows that any

change in policy will systematically alter the structure of econometric models.

In other words, the parameters of large-scale macroeconometric models are

unlikely to remain constant in the face of policy changes, since rational economic

agents may adjust their behaviour to the new environment. Because

the estimated equations in most existing Keynesian-style macroeconometric

models do not change with alternative policies, any advice given from policy

simulations is likely to be misleading. When trying to predict the impact on the

economy of a change in policy it is a mistake, according to Lucas, to take as

given the relations estimated from past data.

This weakness of Keynesian-style macroeconometric models was particularly

exposed during the 1970s as inflation accelerated and unemployment

increased. The experiences of the 1950s and 1960s had led some policy makers

and economic theorists to believe that there was a stable long-run trade-off

between inflation and unemployment. However, once policy makers, influenced

by this idea, shifted the policy regime and allowed unemployment to fall

and inflation to rise, the Phillips curve shifted as the expectations of economic

agents responded to the experience of higher inflation. Thus the predictions of

orthodox Keynesian models turned out to be ‘wildly incorrect’ and a ‘spectacular

failure’, being based on a doctrine that was ‘fundamentally flawed’ (Lucas

and Sargent, 1978). Lucas’s rational expectations version of the Friedman–

Phelps natural rate theory implies that policy makers cannot base policy on the

apparent existence of any short-run Phillips curve trade-off. The monetary

authorities should aim to achieve low inflation, which has significant welfare

gains (see Sargent, 1999; Lucas, 2000a, 2003).

The Lucas critique has profound implications for the formulation of macroeconomic

policy. Since policy makers cannot predict the effects of new and

different economic policies on the parameters of their models, simulations

using existing models cannot in turn be used to predict the consequences of

alternative policy regimes. In Lucas’s view the invariability of parameters in

a model to policy changes cannot be guaranteed in Keynesian-type disequilibrium

models. In contrast, the advantage of equilibrium theorizing is that,

by focusing attention on individuals’ objectives and constraints, it is much

more likely that the resulting model will consist entirely of structural relations

which are invariant to changes in policy. Lucas identified the treatment

of expectations as a major defect of the standard large-scale macroeconometric

models. With rational expectations, agents will react quickly to announced

policy changes. The underprediction of inflation during the late 1960s and

early 1970s seemed to confirm Lucas’s argument. In 1978 Lucas and Sargent

famously declared that ‘existing Keynesian macroeconometric models are

incapable of providing reliable guidance in formulating monetary, fiscal and

other types of policy’.

The Lucas critique implies that the building of macroeconometric models

needs to be wholly reconsidered so that the equations are structural or behavioural

in nature. Lucas and Sargent (1978) claim that equilibrium models are

free of the difficulties associated with the existing Keynesian macroeconometric

models and can account for the main quantitative features of business cycles.

Ultimately the influence of the Lucas critique contributed to the methodoThe

logical approach adopted in the 1980s by modern new classical theorists of

the business cycle, namely ‘Real Business Cycle’ theory (see Figure 5.7).

Such models attempt to derive behavioural relationships within a dynamic

optimization setting.

With respect to macroeconomic stabilization policy, the Lucas critique also

‘directs attention to the necessity of thinking of policy as a choice of stable

“rules of the game”, well understood by economic agents. Only in such a

setting will economic theory help us to predict the actions agents will choose

to take’ (Lucas and Sargent, 1978).

However, some economists, such as Alan Blinder, believe that the ‘Lucas

critique’ had a negative impact on progress in macroeconomics (see Snowdon,

2001a). In addition, direct tests of the Lucas critique have not provided

strong support for the proposition that policy changes lead to shifts of the

coefficients on behavioural equations (see Hoover, 1995a). Blanchard (1984)

has shown that ‘there is no evidence of a major shift of the Phillips curve’

during the change of policy regime adopted during the Volcker disinflation.

Other economists have pointed out that the Volcker disinflation involved a

lower sacrifice ratio than would have been expected before October 1979,

when the policy was implemented (see Sargent, 1999). Finally, it should be

noted that even the structural parameters of new classical ‘equilibrium’ models

may not be invariant to policy changes if economic agents’ tastes and

technology change following a shift in the rules of economic policy. In

practice it would seem that the significance of the Lucas critique depends

upon the stability of the parameters of a model following the particular policy

change under consideration.