Credibility is of great importance in policy evaluation, and some recent analyses have increased our understanding of this issue. These analyses have built on game-theoretic work on reputation. A common example is the government's temptation to exploit the `surprise Phillips curve' to reduce unemployment by surprising the private sector with an inflationary stimulus. The objectives assumed for governments in these analyses are not, however, grounded in a rational theory of government behaviour. In addition, none of these analyses has been subjected to empirical scrutiny, nor do they seem suitable for estimation.
In Discussion Paper No. 255, Research Fellow Patrick Minford sets out an empirical model of government behaviour derived, like private sector behavi- our, from maximizing principles. Conservative parties are assumed to be supported by `capitalists', who hold financial capital in the form of equities and bonds, while labour parties are supported by `workers', who hold human capital; floating voters are those making a transition between the two `classes', and they hold both. Each party trades off the chances of winning over the floating voters against the need to retain the loyalty of their supporters.
Minford compares policy outcomes for inflation and unemployment under `precommitment' and under `discretion', for both Conservative and Labour governments under a four-year electoral cycle. Precommitment implies a manifesto undertaking, suitably supported by penalties for deviation, to fix the macro policy instrument for four years. Discretionary policy permits the government to choose the value of the macro policy instrument in each of the four years according to the circumstances of the time. Expectations are formed rationally in the model. The public forms expectations before the election of what each party would do if elected and the parties then decide, prior to the election, on either precommitment or discretion. The parties view the policy instrument, an increase in public spending, according to its effects on the values of financial capital (negative, via the Fisher effect on interest rates) and human capital (positive, via the Phillips curve effect on output and the shift in the tax burden from labour to bondholders).
Minford estimates the impact of the policy instrument on financial and human capital for Britain by simulating the Liverpool Model over a four-year horizon. His calculations reveal that the optimal precommitted policies for both Conservative and Labour governments involve effectively zero growth in the money supply, implying near price stability. In contrast, discretionary policies lead to 10-30% inflation, with only temporary and modest falls in unemployment. Both parties achieve lower levels of welfare under discretionary than under precommitted policies.
The reason for this result is clear, Minford argues. Under precommitment each party finds that a decision for long-term reflation reduces the value of outside money, and so of financial wealth; also, by raising interest rates, it both depresses current output and lowers the level of government spending (excluding debt interest) that can be accommodated within the higher PSBR, so reducing by both routes the value of human capital as well. But under discretionary policy the decision whether to reflate is taken excluding such costs, since the past is given and the future is already determined by prior recursion. While Minford's results using the Liverpool Model resemble previous results obtained with the surprise Phillips curve, the difference is that the dominant mechanism is the devaluation of outside money

A Political Model of Credibility Patrick Minford

Discussion Paper No. 255, August 1988 (IM)