CGE modelling of the impact of policy changes

The Government is publishing a series of analyses of the major tax cuts that it has announced since the 2010 election, using a technique known as ‘computable general equilibrium’ (CGE) modelling. It has already published papers on its cuts in corporation tax and fuel duty rates, and the Chancellor has suggested that increases in the income tax personal allowance would be another area to analyse. But it has not yet published similar analyses of the much larger tax increases and public spending cuts that pay for them and deliver the fiscal consolidation.

CGE models are large-scale stylised representations of the economy. They assume that the economy tends to a state of equilibrium, in which supply and demand for goods, services and factors of production in the economy are balanced (albeit with adjustment costs affecting use of labour and capital). The Government’s model captures the baseline policy environment via a stylised representation of the tax and benefits systems. A series of interlocking equations then specifies the behaviour of the various sectors in the economy and how they adjust to policy changes. These equations embody available evidence and the modeller’s beliefs about how firms and households respond to changes in incentives.^a After a policy change, agents in the model adjust to price changes until equilibrium is restored. By comparing the baseline and policy scenarios, the modeller can evaluate the estimated impact of the policy change.

CGE models are designed to estimate medium and long-term level shifts in GDP caused by policy changes and to help users understand the mechanisms by which the policy effects flow through the economy. As with any model, CGE results reflect the particular parameters and assumptions embodied in the model, so are subject to uncertainty and open to challenge.

CGE models are not designed as forecasting models – they are normally based on trend growth assumptions, so they do not take into account any short-term fluctuations in spare capacity associated with the current economic cycle. Assumptions relating to monetary policy responses to changes in fiscal policy or economic conditions are at the discretion of the modeller. They are typically better suited to analysing specific tax, welfare or infrastructure spending changes. But in principle they can provide insight into how the economy might adjust to any policy shock.

The assumptions underpinning the Government’s recent modelling of corporation tax and fuel duty cuts suggest corresponding increases in economic growth, which in turn reduces their apparent fiscal cost. Most taxes distort consumption, investment, production or saving decisions in some way, and so cutting them typically delivers economic benefits. This beneficial impact is enhanced when – as in the Government’s studies – the modeller adopts a ‘closure rule’ that assumes that any loss of revenue is made up through the imposition of a non-distorting lump-sum poll tax that inflicts no economic damage. Some distortions created by the tax system are of course deliberate, even though they are ‘costly’ in a CGE framework. For example, fuel duty not only raises revenue, but also helps limit the welfare-reducing side effects of motoring. This closure rule assumption can bias CGE estimates upwards, particularly when tax cuts are financed through a corresponding deficit reduction program. Downward biases are also possible. In the case of corporation tax cuts, there is evidence that points to longer-term positive productivity effects that were not fully captured in the Government’s study. Since these tax cuts have coincided with a period in which actual and trend productivity have fallen short of our forecasts, this channel of adjustment would be difficult to observe in outturn data.

The Government’s cuts in corporation tax and fuel duty and its increases in the income tax personal allowance are not in reality being financed by non-distorting lump sum taxes, but rather by other tax increases and cuts in spending on welfare and public services. The tax and spending measures announced by the current Government amount to a significant net fiscal tightening, on top of that put in place by the previous Government between 2008 and 2010.

Although CGE modelling is not designed as a specific forecasting tool, we thought it would be interesting to run a more representative package of measures – including the ‘takeaways’ as well as the ‘giveaways’ – through a CGE model. This might highlight some economic mechanisms we should take into account when preparing our forecasts – which aim to take into account the full economic and fiscal impact of any policy changes.

Unfortunately, the Government said that it was unable to resource this more comprehensive analysis. But we have engaged external consultants to help us understand some of the broad-brush conclusions that such an analysis might generate.^b These include:

• not all recent tax measures are amenable to CGE modelling, for example the large number of anti-avoidance measures. But most measures can be modelled, including not just the cuts to corporation tax, income tax (via the personal allowance) and fuel duty, but also the cut in business rates and the offsetting increases in VAT, National Insurance contributions, various taxes on products (e.g. tobacco and alcohol) and the reductions in capital allowances. By 2018-19, the aggregate impact of these measures primarily changes the composition of the tax take rather than delivering a large change in the tax-to-GDP ratio. As a whole, CGE analysis suggests that these changes would have had a broadly neutral impact on the level of GDP. That is consistent with our recent judgements that when Budget giveaways and takeaways are broadly offsetting, there are unlikely to be large net effects on the economy;
• the model suggests that several of the Government’s tax measures have reduced relatively inefficient taxes while increasing relatively efficient ones. That is true of cutting corporation tax and fuel duty and raising VAT. But it is not uniform across all changes. The model implies capital allowances are relatively efficiency enhancing, so reducing them will have offset some of the positive effects of reducing corporation tax rates. And it suggests increasing the personal allowance has relatively weak efficiency effects; and
• the impact of large reductions in social security and tax credits in the model would flow from the assumed labour supply responses of affected individuals. By reducing out-of-work income relative to in-work income (which at lower incomes is also affected by raising the personal allowance), the incentive to work is increased – and the higher resulting employment would then raise consumption. But for those who do not move into employment, the income effect of lower benefits would reduce consumption. The types of households affected are likely to have little saving and a high marginal propensity to consume. These expected channels of adjustment are consistent with stronger-than-expected employment and the evolution of our judgements about potential labour supply in recent years.

CGE models are less well suited to capturing the impact of sharp cuts in current spending on public services, since the efficiency effects of different types of spending would be difficult to capture. The main channels of adjustment that could be captured reflect the changing flows of income between the private and public sector – essentially, public services spending involves transfers to the private sector via public sector pay and procurement. One channel of adjustment to cuts in public services spending on the scale planned would be a large shift from public to private employment, which is consistent with experience in recent years.

The choice of ‘closure rule’ is also important. Assuming that any changes in the deficit from spending cuts are offset by non-distorting lump sum transfers would imply smaller GDP effects from spending cuts. If you assume that any net fiscal surplus associated with the policies modelled is used to pay down the deficit, this has the effect of transferring resources from the beneficiaries of public sector pay and procurement to the holders of government bonds. As bondholders are likely to be higher in the income distribution, this could shift resources to those with a lower propensity to consume. It is difficult to map this possible channel of adjustment to recent economic developments.

CGE modelling is a valuable tool for assessing the potential medium and long-term economic impact of policy changes in a relatively rounded way. But this exercise has confirmed to us that it is of limited value in making short-term direct adjustments to our economic and fiscal forecasts. We will therefore continue to look at a broader range of information when considering the potential effect of policy measures on the economy. But our examination has also confirmed that it is possible to use CGE models to explore the implications of a wider range of policy changes than the tax cuts that have so far been the subject of the Government’s published analysis.