Nordic Economic Policy Review 2024

One of the primary objectives of fiscal policy is to enhance welfare by mitigating aggregate volatility, i.e. by dampening the business cycle. Whether this actually improves welfare remains uncertain, as indicated in the literature on real business cycles. The timing of discretionary fiscal policy proves challenging due to implementation lags, (ex post biased) real-time estimates of the output gap and cyclical balance

See, for example, Cimadomo, J. (2012), ’Fiscal Policy in Real Time.’ Scand. J. of Economics, 114: 440-465. https://doi.org/10.1111/j.1467-9442.2012.01697.x

. Ultimately, fiscal policy often becomes pro-cyclical and, as a result, amplifies business cycles. Automatic stabilisers, on the other hand, do not suffer from implementation lags or estimation or projection errors and may address the timing issue effectively.

Kellokumpu, Savolainen and Pesola examine the impact of changes in the output gap on the cyclical balance, which represents the business cycle component of fiscal policy. They assess the sensitivity of various government revenue and expenditure components to variations in the output gap. The individual components are then aggregated to arrive at the overall budgetary semi-elasticity. The estimated semi-elasticity is approximately ½, indicating that a one percentage point change in the output gap results in a ½ percentage point change in the cyclical fiscal balance (cyclical budget deficit relative to output).

In 2020, the output gap measure was -3 percent, and based on their estimate, 1.5 percentage points of the 5.5 percent general government budget deficit were attributed to the automatic response of government revenue and expenditure items.

To estimate revenue elasticities, they employ a step-wise procedure. Initially, they regress the (first difference) detrended tax base on the (first difference) of the output gap, with potential output serving as the trending variable on both sides of the regression. The elasticity of tax revenues on the tax base is then either estimated or evaluated on the basis of the actual tax rates.

Covering the period 1985–2021, which includes multiple economic cycles, including two financial crises and significant changes in the taxation and unemployment benefit systems, the study demonstrates that the aggregate impact of automatic stabilisers is relatively stable.

The examination of the relationship between the output gap yt − y⋆ and the tax base of income taxes, nominal wage sum wt, relies on the following regression

The authors use the same methodology as Almenberg and Sigonius (2021) in ”Automatic fiscal stabilizers in Sweden 1998–2019,” NIES Working Paper 155.

: 

 

where y⋆ is the estimated potential output based on the common EU methodology (production function approach) and ∆ is the first difference operator. The study does not reveal whether the compensation is deflated or not, but it should be. The correct equation would be:

,

or

,

where is (GDP) inflation. Clearly, inflation should be correlated with LHS and, due to the Phillips curve, positively with the output gap:

The estimated regression coefficient in (1) would then be . Hence, the estimate is upward biased. Normally, the estimates of κ are fairly small, in the ballpark of 0.1 such that the bias would not distort the results qualitatively.

The estimated elasticity of nonlinear taxation relies on the simulation of the microsimulation model (SISU), incorporating a broad, representative dataset of individuals and their detailed budget constraints based on tax and social security details. The authors estimate a 5 percent general increase in wage income for all individuals and calculate average and marginal tax rates.

The elasticity of indirect taxes is calibrated assuming unit tax base elasticity with respect to the output gap and applying VAT tax rates to the tax base. Analysing corporate income taxes is challenging due to the lack of information on the tax base. The closest empirical counterpart is 1− labour share. Tax evasion, highly varying risk premia, etc., make estimating output gap elasticity virtually impossible.

The relationship between the unemployment rate and output gap is estimated using the following regression:

,

where is the NAWRU consistent with the potential output measure . Sub-sample estimation of γu reveals sensitivity to the particular sample period, reaching the upper limit of −6 in a sample dominated by the 1990s crisis and −2 in a sample dominated by the stagnant 2010s. This implies that a one percentage point decrease in the output gap could result in a variation of between a two and six percentage points increase in the unemployment rate. The authors then link the unemployment rate with public unemployment-related expenditure and find that the overall elasticity of this expenditure on the output gap is relatively small.

McKay and Reis (2016) identify four channels for automatic stabilisers: 1) the disposable income channel; 2) the marginal incentive channel which is related, for example, to tax progressivity that evens out the work done over boom and bust periods; 3) the redistribution channel, which involves taxing low marginal propensity consumers and paying transfers to high marginal propensity consumers, i.e. from wealthy to poor; 4) the social insurance channel, which reduces the risks that agents face and discourages precautionary savings that make them likely to face liquidity constraints during large aggregate shocks. I employ these channels to classify the results below.

Let us approximate household disposable income dt as

where wt is the nominal wage income, wttaxes paid out of it and utbenefits unemployment benefit net of taxes. The paper estimates the output gap elasticity of each component such that (using the same notation as the paper)

.

Plugging in the parameter estimates gives . This means that the variance of disposable income is substantially smaller than the variance of the output gap.

Income tax progressivity evens out the work effort. Quantitative evaluation requires a model with a heterogeneous agent environment. The same holds for evaluating the redistribution channel and social insurance channel. McKay and Reis (2016) combine an incomplete-markets model with a sticky price model to study the quantitative role of fiscal stabilisers in stabilising output volatility. Unsurprisingly, the demand effects, in the form of the disposable income channel, are not present in their model. Their finding is that the redistribution and social insurance channels play a stabilising role, particularly when monetary policy is constrained by the zero lower bound. The positive welfare effect is driven by the insurance aspect of social insurance.

Kellokumpu, Savolainen, and Pesola provide useful estimates of various government expenditure and revenue elasticities. The in-depth discussion and exploration of the effects of average and marginal tax rates and their evolution are valuable sources of information for macro-modelers. The same applies to unemployment benefits. Their dependence on the output gap is potentially less robust due to the usual uncertainties related to output gap estimation, such as real-time information and underlying trend and growth assumptions. Finally, the evaluation of the stabilising role of automatic stabilisers requires a quantitative, model-based approach, as emphasised by McKay and Reis (2016).

Almenberg, J. & Sigonius, M. (2021). Automatic fiscal stabilizers in Sweden 1998–2019. NIES Working Paper 155.

Cimadomo, J. (2012). Fiscal Policy in Real Time. Scandinavian Journal of Economics, 114(2), 440-465. https://doi.org/10.1111/j.1467-9442.2012.01697.x

McKay, A. & Reis, R. (2016). The Role of Automatic Stabilizers in the U.S. Business Cycle. Econometrica, 84(1), 141-194. https://doi.org/10.3982/ECTA11574