Nordic Economic Policy Review 2024

We use data from Sweden to shed some light on the trade-off between policies to make work pay and the size of automatic (fiscal) stabilisers. We use standard methods to estimate the size of Sweden's automatic stabilisers in the period 1998–2022. Taxes on labour income were reduced by about 5% of GDP over the course of this period. We find that the implementation of policies to make work pay did not substantially impair automatic stabilisers. The size of the automatic stabilisers decreased slightly, but mainly in the first half of the sample, and they are currently slightly less than 0.5. Furthermore, we conclude that the stabilisers were unaffected by the COVID-19 pandemic and that if they had been allowed to operate freely during the period 2020-2021, they would have transferred approximately SEK 110 billion to households and firms. The soaring inflation in 2022-2023 is not captured by our estimates. However, we judge that the net of the different effects of inflation is negligible. Therefore, the stabilisers remain the same size.

Keywords: Automatic fiscal stabilisers, earned income tax credit, make work pay, COVID-19 pandemic, inflation.

JEL Classification: E64, E62, H31

The article builds upon the working paper “Automatic Stabilizers in Sweden 1998-2019”, written jointly with Johan Almenberg. The author is grateful for support and comments by Johan Almenberg. Insights from Karl Harmenberg and Erik Höglin at the Nordic Economic Policy Review conference, as well as from the editors Juha-Pekka Junttila and Jouko Vilmunen, improved the article. Sebastian Ankargren, David Domeij, Erik Glans, Jesper Hansson, Martin Flodén, Erika Färnstrand Damsgaard, Rickard Sandberg and seminar participants at the Swedish Ministry of Finance and at the National Institute of Economic Research, as well as colleagues at the Swedish Fiscal Policy Council, have also contributed valuable comments.

We use data from Sweden to shed some light on the trade-off between policies to make work pay and the size of automatic (fiscal) stabilisers. Sweden is an interesting case study because it is a welfare state that has undertaken sizeable reforms to strengthen incentives to work. We use standard methods to estimate the size of Sweden's automatic stabilisers in the years 1998–2022. Taxes on labour income were reduced by about 5% of GDP during the same period. We find that the implementation of policies to make work pay did not substantially impair automatic stabilisers. An important driver of this result is the design of the earned income tax credit. The size of the automatic stabilisers decreased slightly, but mainly in the first half of the sample period, and they are currently slightly lower than 0.5. Furthermore, we conclude that the stabilisers were unaffected by the COVID-19 pandemic and that if they had been allowed to operate freely in 2020–2021, they would have transferred approximately SEK 110 billion to households and firms. However, this was partly replaced with discretionary support. If discretionary fiscal policy is to continue to work as well in the future as it did during the pandemic, the need for large automatic stabilisers might not be as great as previously thought. There is a lag between inflation and its effects on the automatic stabilisers, so the soaring inflation of 2022–2023 has not been captured by our estimates. However, we judge that the net of the different effects of inflation is negligible. Therefore, the stabilisers remain the same size despite the high inflation.

The role of fiscal policy in stabilising the economy has been a topic for discussion for the last decade (see, for example, Furman and Summers (2020) and Blanchard (2022)).

Swedish Fiscal Policy Council (2021) relates this discussion to the Swedish context.

One reason for this is that monetary policy has been constrained by the effective lower bound. More recently, discretionary fiscal policy supported the macroeconomy during the COVID-19 pandemic, as well as when households and firms were confronted with soaring electricity prices following Russia's war on Ukraine. The Swedish central government's discretionary policy during the pandemic amounted to SEK 330 billion for 2020–2022, and support for high energy prices in 2022 and 2023 amounts to approximately SEK 70 billion.

Statistics Sweden and Swedish Fiscal Policy Council (2023).

Fiscal policy can also reduce macroeconomic fluctuations through automatic stabilisers: rules and regulations that determine public sector revenue and expenditure and automatically soften the impact of the business cycle on households and firms. For example, income taxes and social security benefits reduce the volatility of households’ disposable income over the business cycle. When economic activity declines, tax revenue declines, but unemployment may rise, causing an increase in rule-based unemployment-related expenditure. When economic activity increases, the reverse may be the case. These changes in public revenue and expenditure are automatic to the extent that they are rule-based and result in smaller variations in disposable income for households and firms and, therefore, smaller variations in private sector aggregate demand. By contrast, the government’s budget balance will vary more as a result.

Another topic that has been prominent position in debates on labour market policy in recent decades is policies designed to make work pay, for example, through lowering taxes on earned income and reducing benefits to the unemployed (see, for example, Blundell (2006)). These policies aim to improve attachment to the labour market and alleviate poverty. This points to a potential conflict between stabilisation policy – where high automatic stabilisers generally mean high taxes and generous unemployment insurance – and policies designed to make work pay.

The purpose of this article is twofold. First, we use the most recent data from Sweden to shed more light on the trade-off between policies to make work pay and the size of automatic fiscal stabilisers.

Almenberg and Sigonius (2021) focus on this question but analyse the period 1998–2019.

Sweden is an interesting example because it is a welfare state that has undertaken sizeable reforms to improve incentives to work. As shown in Figure 1, taxes on labour income were reduced by about 5% of GDP over two decades. About half of this is due to an earned income tax credit introduced in 2007 and expanded several times, most recently in the budget for 2024. Expenditure on unemployment-related transfer payments has also decreased, from more than 2% of GDP at the beginning of the period to under 1% at the end.

Figure 1. Tax on labour income and unemployment-related transfer payments 1998–2022

Note: The adjustments made are for the temporary discretionary support introduced during the COVID-19 pandemic.
Source: Statistics Sweden and own calculations.

The size of the automatic fiscal stabilisers in Sweden is estimated for the period 1998–2022 using the same approach as Girouard and André (2005), which breaks down the elasticity of the fiscal balance to the business cycle into a structural part reflecting tax and benefit rules and a cyclical part reflecting how tax bases and benefit-related aggregates respond to the state of the economy. The structural component can be assessed using the rules that apply in a given year. The cyclical component can be estimated using time series data.

A limitation of this approach is that it does not model the behaviour of agents in the economy and is, therefore, subject to the Lucas critique. Nor does it take into account what type of shock is affecting the economy. Simply put, it is an unconditional expectation of the fiscal balance for a given change in GDP. However, even this simple measure is informative, and it is commonly used as a rule of thumb in fiscal policy.

Flodén (2009) uses the same approach and finds that the automatic stabilisers decreased from close to 0.6 in 1998 to only slightly above 0.5 in 2009. This would imply that a 1 percentage point change in the GDP gap would be expected to change the government fiscal balance (as a share of GDP) by approximately 0.6 percentage points in 1998, and by 0.5 percentage points in 2009. We extend the analysis in Flodén (2009) with over a decade of additional data, during which the earned income tax credit was expanded several times.

The findings show that automatic stabilisers in Sweden declined slightly 1998–2022, but mainly in the first half of the period. This confirms the findings in Almenberg and Sigonius (2021), which looked at 1998–2019. Direct taxes on labour have decreased considerably since 1998, in particular because of the earned income tax that was introduced in 2007 and gradually scaled up. However, the average tax rate declined more than the average marginal tax rate, making the income tax more progressive. This partly offsets the effect of lower taxes on the automatic fiscal stabilisers. Expenditure on unemployment benefit also fell during the first half of the period covered. The contribution to the automatic stabilisers from income tax and unemployment-related transfer payments decreased from 0.26 to 0.16, meaning that a 1 percentage point change in the GDP gap would be expected to change the government fiscal balance (as a share of GDP) by approximately 0.1 percentage points less in 2022 than in 1998. However, the size of the automatic stabilisers also depends on other taxes, which remained relatively unchanged during this period. The total size of the automatic stabilisers remained at 0.46 in 2022, compared to 0.55 in 1998. The reduction in the size of the automatic stabilisers is modest, considering the scope of the reforms carried out. The findings show that it is possible to increase the incentives to work without substantial impairment of the automatic stabilisers.

The article also discusses how recent crises have affected the size of the automatic stabilisers and what role they have played in stabilisation policy. Several support measures were introduced by the government to aid households and firms during the COVID-19 pandemic in 2020 and 2021. Many of those measures were subsequently extended, and the total support is estimated to have been SEK 170 billion in 2020 and SEK 120 billion in 2021. Our calculations show that if the automatic stabilisers had been allowed to operate freely, the public sector would have automatically distributed approximately SEK 110 billion to households and firms in those two years. However, this was in part replaced with discretionary support to protect people and labour market matches. These policies broke the link between the output gap and the usual effects from the automatic stabilisers on the public sector.

The CPI with fixed mortgage interest rate was 7.7% in 2022 and 6.0% in 2023. Unexpected changes in the inflation rate led to a change in the composition of GDP, as the wage share decreased and the profit share increased. Tax on corporate profits has a higher elasticity than direct taxes on wages, so this shift is expected to make the automatic stabilisers bigger. Expenditure on unemployment insurance decreases as a share of primary expenditure, which makes the stabilisers smaller. We find that these effects are similar in size, so cancel out each other. Therefore, the automatic stabilisers in Sweden are still slightly below 0.50.

The rest of this paper is organised as follows: Section 2 presents the baseline results of our study of developments of the size of the automatic stabilisers in Sweden. Section 3 discusses how the stabilisers were affected during the crises of recent years and section 4 presents our conclusions.

As per Girouard and André (2005), we use the budget elasticity, which describes the response of the fiscal balance to fluctuations in GDP around its trend

The budget elasticity can be used along with the GDP gap and the public sector budget balance to calculate the cyclically-adjusted budget balance, which shows the underlying fiscal position when cyclical movements are removed. This approach is used by the EU, IMF, and OECD.

, as a measure of the size of the automatic fiscal stabilisers.

Almenberg and Sigonius (2021) provides more details about the estimation, as well as several robustness tests.

The budget elasticity is calculated using a disaggregated approach

A disaggregated approach is used by Van den Noord (2000); Girouard and André (2005); Flodén (2009); Price et al. (2015). The benefit of using a disaggregated approach is that longer time series can be used to estimate the relationship between the business cycle and the tax bases while elasticities that depend on political decisions, e.g. the elasticity between tax revenue and the tax base, can be modelled explicitly. Another method is micro simulations, as deployed by Auerbach and Feenberg (2000); Auerbach (2009), or macro simulations, as in NIER (2015); McKay and Reis (2016). The advantage of macro simulations is that they can capture how the size of the automatic stabilisers depends on the type of shocks that hit the economy.

, estimating separate elasticities for four categories of tax revenue as well as for primary expenditure. The elasticities are added using GDP shares as weights. Letting  be the budget elasticity with respect to changes in the GDP gap,  the elasticity of revenue from tax with respect to the GDP gap, tax share of GDP,  the elasticity of primary expenditure with respect to the GDP gap and  primary expenditure (expenditure net of interest payments) as a share of GDP, we can write

 

The elasticities  and  show how public revenue and expenditure respond to changes in the GDP gap and can be separated into two constituent parts. On the revenue side, the first part is how tax revenue changes in response to changes in the tax base, ,  and the second part is how tax bases change in response to changes in the GDP gap, . On the expenditure side, a similar calculation is performed by looking at how primary expenditure changes in response to changes in unemployment and how unemployment changes in response to changes in the GDP gap.

In sum, for the different tax categories and for primary expenditure, there are three factors that determine their contribution to the overall budget elasticity. First, how tax bases and unemployment respond to changes in the GDP gap. Second, how tax revenue and primary expenditure respond to changes in the tax bases and unemployment. Third, the relative size of the respective tax categories and the primary expenditure in relation to GDP.

We correct for the discretionary policy during the COVID-19 pandemic in several ways. First, we choose to use 2019 as the final year when estimating how tax bases and benefit-related aggregates respond to the state of the economy. Secondly, we exclude temporary tax cuts such as a temporary earned income tax deduction and temporary lower payroll taxes. Similarly, we exclude a temporary increase in unemployment benefit. Thirdly, the primary expenditure is adjusted for pandemic-related discretionary policies, such as support for firms affected by the pandemic. We also take into consideration how such support measures affect the tax bases and adjust the taxes accordingly.

The study uses annual data from the national accounts and published by Statistics Sweden in February 2023. Our macroeconomic time series starts in 1980. To exclude the effects of the COVID-19 pandemic, we use 2019 as the final year when determining firms’ share of the total value added to the economy, as well as when estimating the elasticity between how the tax bases and unemployment rates respond to changes in the GDP gap. For public expenditure and tax revenue, we use data from 1998 to 2022. The wage income distribution is used to calculate the elasticity of direct taxes on labour with respect to its tax base, and we use the distribution from 2016 based on microdata from Statistics Sweden. The distribution for 2016 is used as a proxy for the true distribution for the remaining years but scaled using the median income for each year. Some variables used in the analysis are not observable and are not reported in the national accounts. We use the GDP gap as a measure of the business cycle, i.e. the deviation of GDP from its long run equilibrium trend. We use assessments of this GDP gap and of equilibrium unemployment published by the National Institute of Economic Research (NIER) in March 2023 (NIER, 2023a).

The NIER calculates potential GDP by adding up different components such as potential labour force, equilibrium unemployment and potential productivity. Historically, the GDP gap presented by the NIER has been similar to the GDP gap presented by the OECD, see Figure 2 in NIER (2018).

As outlined above, the budget elasticity is constructed using separate estimates for tax revenue and expenditure. The revenue side is constructed from separate estimates for four tax categories: direct taxes on labour

The tax that households pay on their labour income, net of tax reductions such as the earned income tax credit.

, payroll tax, corporate income tax and indirect taxes

This category mainly includes VAT but also tax on household capital income.

. First, we estimate the elasticity of the tax base to the GDP gap using time series data. Next, we calculate, year by year, the elasticity of tax revenue to changes in the tax base using year-specific tax rules.

The labour cost share of GDP plays an important role in the calculations since it serves as a proxy for the tax bases for direct taxes on labour as well as payroll taxes. We define the labour cost broadly as all output that is not allocated to firms as gross profits. We define profits as the product of (i) gross surplus as a share of the value added by firms, and (ii) firms' share of the total value added to the economy. This measure is broader than in the national accounts.

For an alternative approach, see Price et al. (2015).

As per Girouard and André (2005), we assume, based on their cross-country analysis, that the labour cost share is 72% of GDP in equilibrium and, therefore, that the profit share is 28% of GDP. This is in line with the Swedish data (see Figure 2).

Figure 2. Labour cost share in Sweden 1980–2019

Note. The labour cost share is calculated as the share of GDP that is not allocated to firms as gross profit.
Sources: Statistics Sweden, NIER and own calculations.

The elasticity of the labour cost share of potential GDP with respect to the GDP gap, , is estimated with a regression where changes in the labour cost share are explained by changes in the GDP gap.

See Almenberg and Sigonius (2021) for details about the regression.

The estimates are for the period 1982–2019 and calculate elasticity to be 0.83, which is in line with earlier estimates by the OECD.

See Girouard and André (2005).

A number smaller than one means that the labour cost share decreases and the profit share increases when an economy enters a boom.

The elasticity of direct taxes on labour with respect to the labour cost share depends on both the level and the progressivity of income taxes. Sweden has high taxes on labour income, but the level has been reduced in the last 20 years, mainly through the introduction of a tax reduction for the workers’ contribution to the state pension and the introduction of an earned income tax credit.

The state pension in Sweden is financed by both firms and workers. Workers must contribute and before 2000 their contributions were tax deductible. In 2000, the system changed so 25% of the contribution allows for a tax reduction whereas the remaining share was tax deductible. The share that triggered a tax reduction gradually increased and from 2006 the entire contribution allows for one.

In total, these two reforms, which were similar in magnitude and phased in gradually, reduced taxation of labour income by about 5% of GDP. The reforms targeted low-wage and average-wage earners, lowering marginal tax rates at lower income levels but not at higher income levels. Hence, the effect (on the elasticity) of lower taxes was offset by an increase in progressivity.

If an individual worker earns wage W with the marginal tax rate  and average tax rate  the elasticity between the tax and the wage can be calculated as the ratio between the marginal tax and the average tax, . This relationship is used when calculating the elasticity between direct taxes on labour and the labour cost share, . The calculations are performed year by year in two steps. First, the marginal and average tax rates are evaluated at different income levels. Next, the elasticity is calculated as the ratio of the weighted average of the marginal and average tax rates. As mentioned earlier, the temporary earned income tax credit for 2021 and 2022 is disregarded.

The distribution of wage income is based on micro data from Statistics Sweden for 2016. The distribution has its starting point around the median wage and shows how large a share of the population has income . The income distribution is assumed to have the same shape for all the years but is adjusted with the median wage of each year.

The income distribution for 2016 is used for all years, but it is adjusted according to the evolution of the median income. This means that the same share of the population is assumed to have, for example half the median wage, each year. To ensure that our results are not affected by this assumption we have also calculated the elasticity using the income distributions from 2004 and 2010 without any significant effects on the results; see NIER (2018). The median wage for 2022 is approximated by increasing the median wage in 2021 by the change in the average hourly wage.

The calculations make an implicit assumption that when labour costs increase, all wages increase proportionally. But labour costs are also affected by workers moving in and out of employment. Low-paid workers may be over-represented in this category. Since they face lower marginal and average tax rates, this might affect the elasticity. Also, the introduction of the earned income tax credit has changed the tax for an employed worker compared to an unemployed worker with unemployment insurance, which also affects the elasticities. These concerns are addressed later.

The calculations show that the marginal tax rate was reduced, on average, by 8 percentage points between 1998 and 2014 (see the dark blue line in Figure 3). After 2014, it increased slightly. Meanwhile, the average tax rate was also reduced on average across the wage distribution (see the red line in Figure 3). As a share of the tax rate, the reduction in the average tax rate exceeds the reduction in the marginal tax rate. Hence, the elasticity of direct taxes on labour to the labour cost share increased between 1998 and 2022.

The elasticity of direct taxes on labour was at its lowest in 1999 and has increased by almost 0.3 since then, from 1.24 to 1.51 in 2022 (see the light blue line in Figure 3). The biggest increase between consecutive years was between 2006 and 2007, when the first step of the earned income tax credit was introduced.

Figure 3. The elasticity of direct taxes on labour with respect to the labour cost share ()

Source: Own calculations.

The elasticity of direct taxes on labour with respect to the labour cost share, as shown by the grey line in Figure 3, is multiplied by the elasticity of the labour cost share with respect to the GDP gap, previously estimated to be 0.83, to arrive at the elasticity of direct taxes on labour with respect to the business cycle.

The elasticity of payroll taxes with respect to the business cycle is the product of (i) the elasticity of the labour cost share and the GDP gap (estimated to be 0.83; see above) and (ii) the elasticity of payroll taxes with respect to the payroll (here proxied by the labour cost share), which is 1.0 since payroll taxes are not capped in Sweden. Multiplying (i) and (ii) gives an elasticity of payroll taxes with respect to the business cycle of 0.83.

The elasticity of corporate income tax with respect to the business cycle is the product of (i) the elasticity of corporate profits and the GDP gap, , and (ii) the elasticity of the corporate income tax with respect to corporate profits.

As a proxy for corporate profit's share of GDP, we use the gross profits as the share of GDP, which, as mentioned earlier, is the same as one minus the labour cost share. While in theory this is the part of added value that accrues to firms, it differs from taxable profits that allow for deductions for depreciation, interest and other items. Taxable profits amount to about 10% of GDP, whereas the profit share, defined as above, is about 25–30%. However, we use this approximation because we are estimating elasticities and not levels.

Almenberg and Sigonius (2021) also show that using taxable profits instead provides a slightly higher contribution to the automatic stabilisers from corporate income tax.

This approach, which is in line with previous research, implies that all additional added value that accrues to firms during a boom is taxable, which means that the elasticity of profits with respect to the GDP gap is 1.45.

The formula for calculating the elasticity of the profits with respect to the business cycle is presented in the appendix.

The Swedish corporate income tax is proportional to profits, so as an approximation the corporate income tax revenue responds one-to-one to changes in profits (but since losses can be offset against future profits, this is only an approximation, albeit a reasonable one). As a result, the elasticity of the corporate income tax with respect to the business cycle is also 1.45.

Indirect taxes consist of consumption taxes in the form of value-added taxes and excise duties, as well as taxes on household capital income. It is hard to assess how these tax bases correlate with the GDP gap. As per Girouard and André (2005), we have assumed an elasticity of 1. Since these taxes are largely proportional, the elasticity of tax revenue to the tax bases is also set to 1. Hence, the elasticity between tax revenue from these indirect taxes and the GDP gap, which is the product of the two elasticities, is 1.0.

The calculations above are summarised in Table 1. Column (iii) shows how the elasticity of direct labour taxes to the GDP gap has increased over time, driven by an increase in the elasticity of direct taxes on labour with respect to the labour cost, as shown in column (ii). The other three revenue elasticities (payroll taxes, corporate income tax and indirect taxes) are, by design of the chosen method, constant.

Table 1. Tax elasticities with respect to the GDP gap.

Note. Columns (i) and (ii) show the elasticity of the labour cost share with respect to the GDP gap and the tax income with respect to the labour cost. The product of the two columns gives the elasticity of direct taxes on labour with respect to the GDP gap and is displayed in column (iii). Column (iv) to (vi) shows the elasticities for each tax with respect to the GDP gap.
Source: Own calculations.

The elasticity of the unemployment gap with respect to the GDP gap is estimated with a regression where changes in the unemployment gap are explained by changes in the GDP gap.

See Almenberg and Sigonius (2021) for details about the regression.

  The estimation is for the period 1982–2019, and we use the estimate of equilibrium unemployment published by NIER. The NIER defines equilibrium unemployment as the unemployment rate that would prevail if the GDP gap were zero and the economy progressed along a balanced growth path. The elasticity is –6.08, which implies that when the equilibrium unemployment level is, for example, 7 percent, a 1 percentage point increase in the GDP gap lowers the unemployment rate by 0.4 percentage points.

Primary expenditure, , is divided into two parts: unemployment-related transfers (unemployment insurance and compensation for participants included in labour market programmes) and other expenditure. It is assumed that unemployment-related transfer payments are the only expenditure that varies with the business cycle and proportionally with unemployment. Unemployment-related transfer payments are taxable in Sweden and the variable that matters for the automatic fiscal stabilisers is expenditure net of tax. The elasticity of net primary expenditure with regard to the unemployment gap, , is a function of the net expenditure unemployment-related transfers as a share of primary expenditure, corrected for the unemployment gap.

See Almenberg and Sigonius (2021) for details about how to derive the formula for the elasticity of primary expenditure with respect to the unemployment gap.

Expenditure on unemployment-related transfer payments has varied considerably over time and has declined as a share of GDP and as a share of primary expenditure (see Figure 4). Three factors drive this trend: (i) unemployment declined, (ii) unemployment benefits increased at a slower rate than nominal GDP, and (iii) a declining proportion of the workforce has been eligible for these benefits.

OECD reports in their tax-benefit data portal that the Swedish net replacement rate in unemployment has fallen from 82% in 2001 to 70% in 2019 for a worker unemployed for two months with a previous wage 67% of the average wage. According to the Swedish Public Employment Service, the proportion of unemployed people who received unemployment benefit fell from 69% in 1999 to 45% in 2019.

The calculations are corrected for the temporary increase in unemployment-related transfer payments, as well as primary expenditure, during the COVID-19 pandemic. We subtract SEK 10, 6 and 1 billion during 2020–2022 for the unemployment-related transfer payments and SEK 97, 91 and 33 billion from the primary expenditure.

Figure 4. Expenditure on unemployment-related transfer payments 1998–2022

Note. The expenditure in the figure consists of gross expenditure, i.e. before tax. The adjustments made are for the temporary discretionary support introduced during the COVID-19 pandemic.
Source: Statistics Sweden and own calculations.

The elasticity of expenditure with respect to the unemployment gap is shown in Figure 5. This shows that the elasticity of primary expenditure with respect to the unemployment gap has decreased over time. The driver is the decrease in unemployment-related transfer payments, from about 4.5% of primary expenditure at the outset of the period to about 1.5% at the end. The elasticity is small, approximately 0.03 at the beginning of the period and 0.01 at the end, reflecting the fact that most primary expenditure is not affected by the unemployment rate.

Figure 5. Elasticity of primary expenditure 1998–2022

Note. Table 4 in the appendix shows the different factors that determine the elasticity of expenditure with respect to the unemployment gap, as well as the elasticity of primary expenditure with respect to the GDP gap.
Source: Own calculations.

To arrive at the elasticity between primary expenditure and the GDP gap, the elasticity of the unemployment gap and the GDP gap (estimated to be –6.06; see above) is multiplied by the elasticity of primary expenditure with respect to the unemployment gap. The results are reported in Figure 5. As shown, the elasticity of expenditure to the GDP gap is about one-third of the size in 2022 compared to 1998, namely -0.06 compared to -0.15. The driver for this, as mentioned earlier, is the decrease in unemployment-related transfer payments. A negative elasticity means that the public sector supports the private sector during a downturn, but the fact that it is close to zero means that the effect is minor. However, that the vast majority of public spending is not affected by the business cycle is in itself something that stabilises the economy.

To arrive at an estimate of the overall budget elasticity – and hence of the automatic stabilisers –the estimated elasticities reported above are aggregated and weighted for GDP shares. Figure 6 shows that direct taxes on labour have decreased as a share of GDP between 1998 and 2010 and have remained flat since 2010. The lower share is mainly a consequence of the tax credit for the employee pension contribution phased in from 1998 and the introduction of the earned income tax credit from 2007 onward. Other taxes have remained relatively unchanged as a share of GDP. Primary expenditure has varied more over time with a modest downward trend. The numbers in the calculations are corrected for the discretionary support 2020–2022. The taxes are adjusted for the tax cuts (both the temporary earned income tax credits and the temporary cuts in the payroll tax amounting to SEK 46 billion during the period) but also for the fact that some forms of support, such as the furlough schemes, increased the tax base. For direct tax on labour, the tax rate for the median income in Sweden is used to deduct the expected tax from the furlough scheme. The corporate income tax rate is used to calculate how much of the support to firms is paid back as taxes.

Figure 6. GDP shares of taxes and primary expenditure 1998–2022

Note: The series are adjusted for the temporary discretionary support introduced during the COVID-19 pandemic.
Sources: Statistics Sweden and own calculations.

Payroll taxes, corporate income taxes and indirect taxes have been relatively stable as a share of GDP. This, combined with the method we use, in which their elasticities with respect to the GDP gap are constant over time, means that their contributions to the automatic stabilisers are about the same throughout the period.

Meanwhile, the contributions to automatic stabilisers from direct taxes on labour and primary expenditure have changed (see Table 2). The elasticity of direct taxes on labour with respect to the GDP gap has increased during the period studied, in particular during the first half of it. At the same time, direct taxes on labour as a share of GDP have decreased. The latter effect dominates, leading to an overall reduction in the contribution made by direct taxes on labour to the automatic stabilisers. This reduction occurred prior to 2010, and since then, there has not been any significant change, even though a slight downward shift can be observed.

The contribution from primary expenditure to the automatic stabilisers has decreased, mainly due to a decrease in the elasticity during the first half of the period studied, but also because primary expenditure as a share of GDP has shown a modest downward trend (see Table 2).

Table 2. Weighted elasticities and contributions to automatic stabilisers. Elasticity and share, respectively

Note. The contributions to the total size of the automatic stabilisers are calculated by multiplying each elasticity for a given year with its weight (GDP share) in the same year.
Sources: Statistics Sweden and own calculations.

Table 3 summarises how the different taxes and forms of primary expenditure contribute to the overall budget elasticity (i.e., to the size of the automatic stabilisers), and how these contributions have evolved over time.

Table 3. Automatic stabilisers 1998–2022. Elasticity

Note. Column (vi) shows the size of the automatic stabilisers calculated as the sum of columns (i)-(iv) minus column (v).
Source: Own calculations.

The calculations show that the size of the automatic stabilisers decreased slightly up until 2011, from 0.55 in 1998 to 0.47 in 2011, and has remained relatively unchanged since then. The reduction in the automatic stabilisers prior to 2011 was due to a lower contribution from direct taxes on labour and from primary expenditure. The lower contribution from direct taxes on labour is due to a reduction in these taxes as a share of GDP. The effect of lower taxes on labour on the automatic stabilisers has, in part, been counteracted by greater progressivity in the taxation of labour income, in particular because of the way the earned income tax credit is designed.

If the elasticity between direct taxes on labour with respect to the labour cost had been the same in 2022 as in 1998, but direct tax on labour as share of GDP had decreased from 17% to 10%, the size of the automatic stabilisers would have been 0.43 instead of 0.46 in 2022.

The lower contribution from primary expenditure is mainly due to lower unemployment benefits as a share of GDP.

Although the contribution to the automatic stabilisers from direct taxes on labour and primary expenditure has decreased, from 0.26 in 1998 to 0.16 in 2022, the overall reduction in the size of the automatic stabilisers is modest since the other taxes were unaffected by the reforms; the automatic stabilisers have been around 0.5 throughout the period. The findings show that it is possible to increase the incentives to work without substantial impairment of the automatic stabilisers. Furthermore, the findings show that the discretionary policy during the COVID-19 pandemic did not affect the size of the automatic stabilisers. The stabilisers were 0.47 in 2019, before the pandemic, and 0.46-0.47 during the pandemic years.

Almenberg and Sigonius (2021), who use the same method as above to calculate the size of the automatic stabilisers for 1998–2019, apply two important robustness tests to validate the method used. The earned income tax credit is designed in a way that increases the incentives for low-paid workers to take jobs. Many of those workers are in sectors that are sensitive to the business cycles, such as construction. Therefore, Almenberg and Sigonius assessed whether the assumption that the entire income distribution is affected by the business cycle affects the result. They concluded that truncating the income distribution that is affected by the business cycle only has a marginal effect on the automatic stabilisers. Assuming that only those workers who have an income up to 50% of the median income are affected by the business cycle, their calculations show that the stabilisers are slightly lower in the first years of the period studied, compared to the results presented above, and slightly higher in the final years. For 2019, this means that the automatic stabilisers are 0.50, compared to the 0.47 that we reported above. In addition, the size of the automatic stabiliser is virtually identical throughout the period studied.

The earned income tax credit creates a tax shield when individuals are made redundant that decreases the tax take once they find new jobs. Almenberg and Sigonius (2021) allow these workers to be unemployed for part of the year. After the introduction of the earned income tax credit, the marginal tax rate decreases with the assumed duration of the unemployment because the tax credit, in relation to the wage earned, is larger for low wages. Hence, an unemployed worker who receives a lower wage and higher unemployment benefit, faces a lower marginal tax rate. This results in a smaller elasticity of direct taxes on labour with respect to the labour cost share. However, the difference compared to the baseline estimates is relatively small. Assuming that the change to the direct tax on labour income comes from unemployed workers with an income up to 50% of the median income finding new jobs after six months, we arrive at automatic stabilisers of 0.44 in 2019 instead of 0.47.

Almenberg and Sigonius (2021) use four additional robustness tests. They explore how assessments of the automatic stabilisers are affected by (i) shortening the sample to only include data from 1998 onwards, (ii) a different definition of wage sum and profit share as well as unemployment-related transfer payments, (iii) the inclusion of expenditure that may (rightly or wrongly) be deemed to function as semi-automatic stabilisers, and (iv) the uncertainty that stems from the regressions used when estimating how the labour cost share and the unemployment rate respond to the state of the economy. The overall conclusion from the extensions and robustness test is that the method used in this article gives a reliable estimate of the current size of the automatic stabilisers. Hence, the result presented above, where the current size of the automatic stabilisers in Sweden is slightly less than 0.5, holds true.

In recent years, Sweden, like the other Nordic countries, has been hit by two crises. The COVID-19 pandemic in 2020–2022, together with the measures introduced to prevent the spread of the infection, led to a sharp decline in GDP, which prompted discretionary fiscal support measures on a scale never seen before. The inflationary crisis, which started in 2022 and is not over yet, has meant a sharp tightening of monetary policy and that the Swedish economy has entered a recession. The fiscal stance has been fairly moderate in order to not counteract monetary policy. We discuss how the two crises affected the size of the automatic stabilisers and how the automatic stabilisers functioned during them.

When the COVID-19 pandemic hit Sweden in the spring of 2020, the forecasts were gloomy, to say the least. For example, the NIER predicted that GDP would fall by 7% that year.

NIER (2020).

Due to the spread of the infection and the difficult macroeconomic situation, the government brought in a number of support measures to support businesses and households. Several of these measures were subsequently extended several times. During 2020–2022, pandemic measures are estimated to have cost SEK 330 billion.

Statistics Sweden.

Furlough schemes, under which an employee works fewer hours, wages are reduced and the government pays part of the wage, cost a total of SEK 68 billion. Payroll taxes were temporarily reduced at a cost of SEK 46 billion. The unemployment insurance ceiling was temporarily raised, and it became easier to qualify for unemployment insurance, which cost SEK 17 billion. Direct support for firms amounted to approximately SEK 64 billion. Those were the biggest expenses.

The size of the discretionary measures can be contrasted with the support the automatic stabilisers provided. The calculations in Section 2 show that the automatic stabilisers were slightly less than 0.5 during the years studied, which means that the risk in the event of an economic downturn is shared approximately equally between the public sector and the private sector if the stabilisers are allowed to operate freely. The GDP gap in 2020 was estimated to have been -4%.

The first assessment of the GDP gap for 2020 after the actual GDP for 2020 had been presented (NIER, 2021).

Along with GDP in current prices of SEK 5,039 billion and an automatic stabiliser of 0.47, this means that SEK 95 billion was transferred to households and firms. The corresponding figures for 2021 are -0.7%

The first assessment of the GDP gap for 2021 after the actual GDP for 2021 had been presented (NIER, 2022).

in GDP gap and a GDP in current prices of SEK 5,487 billion, which means that SEK 18 billion was transferred to households and firms. This means that the automatic stabilisers would have added slightly more than SEK 110 billion to the private sector in 2020 and 2021 if they had been allowed to operate freely.

The economy was expected to expand slightly in 2022 and the year is therefore disregarded in the calculations (NIER, 2022).

The calculation above shows the maximum amount the automatic stabilisers could contribute. However,  for this to happen,  the unemployment rate needed to increase, while the labour income, direct tax on labour and payroll taxes needed to decrease. Firm profits and the corporate income tax needed to be lower, as well as the indirect taxes. However, this scenario was prevented by the discretionary support designed to protect people and labour market matches. These policies broke the link between the output gap and the usual effects on the public sector through the automatic stabilisers. In other words, the discretionary policy prevented the automatic stabilisers from operating.

For stabilisation policy to be effective, it needs to be implemented promptly. Otherwise, there is a risk of the business cycle changing, and it needs to target the parts of the economy where it is expected to have the greatest impact. In addition, the measures should – if they are introduced exclusively to stabilise the economy – be temporary and reversed when the economy recovers. These principles are sometimes summarised by saying that the discretionary measures should be timely, targeted and temporary.

It has been a common belief – but not a self-evident truth – that there is a risk of the political process around fiscal policy going awry since it can take a relatively long time to decide on and implement measures, the risk of short-term political considerations leading to the wrong measures, as well as the temptation to make temporary measures permanent. It is against this background that automatic fiscal stabilisers have been considered the safest way for fiscal policy to conduct stabilisation policy.

However, the COVID-19 pandemic showed that discretionary fiscal policy could be timely, targeted and temporary. Many support schemes were implemented just weeks after the pandemic started in Sweden. The aim was to protect matches in the labour market, and they were prolonged as the pandemic continued but eventually abolished again. A conclusion is that discretionary fiscal policy is better at stabilising the economy than its reputation suggests. This implies that the need for large automatic stabilisers might be exaggerated.

A change that took place after the pandemic, and which in the long run could affect the size of the automatic stabilisers, is the introduction of a new furlough scheme for individual firms when they face a temporary downturn.

As a contrast to the system used during the pandemic which was introduced when the entire economy was in a downturn. In order for a firm to be able to apply for the new support, it must demonstrate (i) temporary and serious financial difficulties, (ii) that the difficulties have been caused by circumstances beyond the firm’s control, (iii) that the difficulties could not have been foreseen or avoided, and (iv) that the firm has done everything possible to reduce the cost of labour. Firms can only be approved for support on the condition that they are competitive in the long run.

These firms can then choose to lower the hours worked by employees instead of making redundancies. So far, the uptake has been negligible and has only cost SEK three million.

This can be set against the cost of the unemployment-related transfer payments, which amounted to SEK 34 billion in 2022.

If firms were to start using the new furlough scheme to a greater extent, it would stop workers ineligible for unemployment insurance from becoming unemployed, and the scheme would increase the size of the automatic stabilisers. Evaluating the effect of the new furlough scheme on the automatic stabilisers is a job for future research.

Inflation was high in Sweden in 2022 and 2023. The CPI with fixed mortgage interest rate was 7.8% in 2022 and 6.0% in 2023. The Swedish central bank, the Riksbank, tightened monetary policy to bring down inflation, and the tightening contributed to the Swedish economy entering a recession in 2023 which will last for the next few years.

According to the forecast by the NIER (2023b).

The fiscal stance has been fairly moderate in order not to counteract the Riksbank's interest rate increases.

See, for example, the Swedish Fiscal Policy Council (2023) for a discussion of the budget for 2023.

The effects of inflation on the size of the automatic stabilisers lag slightly. Parts of the tax system and welfare system are automatically adjusted according to the inflation rate, but the adjustments for 2022 depend on inflation between June 2020 and June 2021. As mentioned earlier, the automatic stabilisers depend on rules and regulations that determine public sector revenue and expenditure, but the high inflation in 2022 did not affect the tax system and welfare system for 2022. Hence, we are not able to use the calculations from earlier periods to understand quantitatively the effects of inflation. Instead, we will discuss a few effects qualitatively.

Unexpected changes to inflation can lead to changes in the composition of GDP. In 2022, the wage share decreased, and the profit share increased. The forecast is that this will continue in 2023

See NIER (2023b).

and since corporate profits have a higher elasticity than taxes on wages, this shift is expected to increase the automatic stabilisers. Higher inflation means higher nominal GDP and more tax revenue. At the same time, many forms of public expenditure increase. However, since expenditure on unemployment insurance depends on the wage level and the ceiling for unemployment insurance, which is not increased with the inflation rate, unemployment insurance will account for a slightly smaller share of primary expenditure, and this will decrease the size of the stabilisers. In our judgement, these two effects will be similar in size and cancel out each other. Therefore, the automatic stabilisers in Sweden will still be at the same level as in 2022, i.e. 0.46, or as we stated earlier, slightly less than 0.50.

One effect of inflation is that it has a detrimental effect on the budget balance of Swedish local authorities. Their income is mainly direct tax on wages and support from central government, two sources not directly affected by inflation. Municipal expenditure, such as purchasing goods and paying rent, do increase with inflation, which, all other things being equal, forces municipalities to cut costs to stick to the fiscal framework. There is a risk that municipalities will cut consumption during the downturn in 2024, which will counteract the effects of the automatic stabilisers. A decade ago, a new system, which allows municipalities to even out their budget balance and consumption over the business cycle came into force. As pointed out by Portes and Wren-Lewis (2015), the way fiscal frameworks are designed can affect automatic stabilisers. Hence, the option for municipalities to even out their results might affect the fiscal stabilisers even though it does not affect the budget elasticity. The slight decline in the automatic stabilisers indicated by the budget elasticity is partly offset by this new system. To what extent the system hinders the municipalities from cutting costs will probably become clear in 2024 and will be another job for future research to evaluate.

The calculations above show that the automatic stabilisers in Sweden were 0.55 in 1998 and 0.46 in 2022. One finding is that policies to make work pay have not impaired automatic fiscal stabilisers to any great extent, reflecting the way these reforms, such as the earned income tax credit, were designed and the fact that many taxes and their contributions to the automatic stabilisers were not affected by the reforms. The method used treats the budget elasticity with respect to the GDP gap as a measure of the size of automatic stabilisers. In addition, the discretionary support during the COVID-19 pandemic prevented the automatic stabilisers from working freely. If not, they would have contributed approximately SEK 110 billion SEK to households and firms. The recent surge in inflation may affect the automatic stabilisers, but in our judgement, the effect will be negligible.

Almenberg, J. and Sigonius, M. (2021). Automatic fiscal stabilizers in Sweden 1998–2019. Working paper No 155. National Institute of Economic Research.

Auerbach, A. J. (2009). Implementing the New Fiscal Policy Activism. American Economic Review 99(2), 543–549.

Auerbach, A. J. and D. Feenberg (2000). The Significance of Federal Taxes as Automatic Stabilizers. Journal of Economic Perspectives, 14(3), 37–56.

Blanchard, O. (2022). Fiscal Policy under Low Interest rates. Cambridge: The MIT Press.

Blundell, R. (2006). Earned Income Tax Credit Policies: Impact and Optimality. Labour Economics, 14(3), 423-443.

Flodén, M. (2009). Automatic fiscal stabilizers in Sweden 1998–2009. Studier i finanspolitik 2009/2, Finanspolitiska rådet.

Furman, J. and Summers, L. (2020). A Reconsideration of Fiscal Policy in the Era of Low Interest Rates. Discussion draft.

Girouard, N. and André, C. (2005). Measuring cyclically-adjusted budget balances for OECD countries. Economics Department Working Paper No. 434, OECD.

McKay, A. and Reis, R. (2016). The role of automatic stabilizers in the U.S. business cycle. Econometrica, 84 (1), 141–194.

NIER (2015). Konsekvenser av att införa ett balansmål för finansiellt sparande i offentlig sektor. Specialstudie 45, National Institute of Economic Research.

NIER (2018). Automatiska stabilisatorer 1998–2018. Specialstudie, National Institute of Economic Research.

NIER (2020). Updated Economic Outlook, April 2020. National Institute of Economic Research.

NIER (2021). The Swedish Economy, March 2021. National Institute of Economic Research.

NIER (2022). The Swedish Economy, March 2022. National Institute of Economic Research.

NIER (2023a). The Swedish Economy, March 2023. National Institute of Economic Research

NIER (2023b). The Swedish Economy, December 2023. National Institute of Economic Research

Portes, J. and Wren-Lewis, S. (2015). Issues in the Design of Fiscal Policy Rules. The Manchester School 83(3), 56–86.

Price, R. W. R., Dang, T., and Botev, J. (2015). Adjusting fiscal balances for the business cycle. Economics Department Working Paper No. 1275, OECD.

Swedish Fiscal Policy Council (2021). Swedish Fiscal Policy 2021.

Swedish Fiscal Policy Council (2023). Swedish Fiscal Policy 2023.

Van den Noord, P. (2000). The size and role of automatic fiscal stabilizers in the 1990s and beyond. Economics Department Working Paper No. 230, OECD.

Using gross profits as a proxy for corporate profits makes it possible to calculate the elasticity of profits with respect to the business cycle by using the profit share, the elasticity of the labour cost share and the GDP gap, as per this equation:

 

where is the profit share in the economy. As mentioned above, we assume the profit share to be 0.28 in equilibrium, and the elasticity of the labour cost share to the GDP gap, , is estimated to be 0.83. Plugging these values into equation (2) gives an elasticity of profits to the GDP gap, , of 1.45.

Table 4 shows the different factors that determine the elasticity of expenditure with respect to the unemployment gap, as well as the elasticity of primary expenditure with respect to the GDP gap.

Table 4. Elasticity of expenditure to the GDP gap. Elasticity and %, respectively

Note. The elasticity of the unemployment gap with respect to the GDP gap is reported in column (i). Column (ii) shows the average tax rate at the median income. Column (iii) shows the unemployment-related transfer payments as a proportion of primary expenditure. Column (iv) shows the unemployment-related transfer payments net of taxes as a proportion of primary expenditure. Column (v) shows unemployment and column (vi) equilibrium unemployment. Column (vii) shows the inverted unemployment gap. Column (viii) shows the elasticity of primary expenditure net of taxes and the unemployment gap, which is obtained by multiplying columns (iv) and (vii), i.e. correcting the net expenditure for the unemployment gap. The expenditure elasticity with respect to the GDP gap is reported in column (ix) and obtained by multiplying columns (i) and (viii).
Sources: Statistics Sweden, NIER, and own calculation