Modelling the dynamic macroeconomic effects of tax policy

[John Short]

Since 2010 the (UK) government has committed to increase the transparency and sophistication of its modelling of the effects of policies. As part of this, HMRC has been developing a Computable General Equilibrium (CGE) model, capable of modelling the dynamic macroeconomic effects.

The model has been peer reviewed by leading academics in the relevant field, who found that ‘The basic design of the HMRC model for the UK economy meets at large the key requirements for state-of-the-art applied tax policy analysis’.

In December 2013 HMRC published an Analysis of the dynamic effects of Corporation Tax, allowing (for the first time) the government to publish retrospective policy analysis of the dynamic effects of government policy, based primarily on the CGE model.

A second report (April 2014) shows the results of applying the CGE model to the real terms fuel duty reductions announced since 2010. Modelling suggests that the tax reductions will increase GDP by between 0.3 per cent and 0.5 per cent in the long-term.

The modelling shows increased profits, wages and consumption all add to higher tax revenues. As a result, the cost of the policy falls by between 37 and 56 per cent in the long-term.

[petagny]

But CGE models are not without limitations. This is how John McDermott summarises these in the FT this morning:

'CGE doesn’t incorporate monetary policy or spare capacity, two crucial aspects to any macroeconomic forecasting model. It is a closed model – there is no sophisticated place for exports and imports in its analysis. It doesn’t produce confidence intervals. It doesn’t take into account “externalities”. Its results are sensitive to whichever year it takes as its baseline. It is inherently biased towards the benefits of tax cuts on output. And it is an ex ante study: we won’t know how accurate it is in predicting the long-term effects of various policies until decades later.

We see some of these limitations at work in the case of the fuel duty cut:

No sense of statistical significance. In the paper we are told that GDP will be 0.4 per cent higher in 20 years time because of the fuel duty cut. Great! But not only is this a rather small (it refers to total GDP, not the growth rate) amount of output, the report does not give us any confidence intervals so we can gauge the estimates and their statistical significance. Without such estimates, the simulation is almost useless as an exercise in precision.
No account of congestion and pollution. As every economics student learns, an externality is a cost or a benefit affecting those outside of a given transaction. Congestion and pollution are two textbook examples of negative externalities. The report is open about its limitation here – but this is a reminder that it doesn’t account for all possible market failures.
The model is dynamic but not comprehensive. I don’t know what role assumptions about monetary policy and spare capacity should play in the final results – but their absence means we should be sceptical about any attempt to simulate GDP two decades down the track.
It uses a “poll tax” to make up the rest of the revenue. The model does not deal in a sophisticated way with the fact that at least half of the lost revenue from fuel duty tax cut has to be made up elsewhere. It does this by in effect assuming a “poll tax”, i.e. that the budget deficit is closed by a transfer from all households. This serves to bias all results towards the benefits of tax cuts, since it suggests that any cut reduces distortion – and it suggests a method for making up the deficit that is at odds with experience.'

http://blogs.ft.com/off-message/2014/04/15/why-you-should-care-about-dynamic-modelling/