Journal Issue

Finland: Selected Issues and Analytical Notes

International Monetary Fund
Published Date:
August 2012
  • ShareShare
Show Summary Details

III. Analytical Note 3: Potential Output Estimates1

A. Introduction

1. Finland has suffered an output collapse in 2009, which was stronger than the drop during the banking crisis in 1991. GDP growth collapsed in 2009, reducing output by 8.4 percent compared to 6 percent in 1991. For the second time in more than three decades, Finland experienced a severe recession: from 2008Q3 until 2009Q2, quarterly growth was negative and the cumulative (peak-to-trough) output loss reached around 10 percent. The crisis in 1991 lasted much longer than the recent crisis, which in turn implied also a higher cumulative output loss. From 1990Q1 until 1993Q1, output contracted by about 12½ percent of GDP.

Finland - Growth Contribution


Sources: National authorities and Fund staff calculations.

2. This time the economy recovered much faster, as the external nature of the shock and the absence of a domestic banking crisis facilitated the fast recovery. The crisis in 1991 was followed by a severe contraction of domestic demand, which the positive net external contribution could not offset. The 2008–09 crisis, instead, was largely due to the collapse of external demand. Domestic demand fell only temporarily and was largely dominated by a drop in investment demand (see also AN1).

3. However, output remains below its pre-crisis trend. If the economy were to continue growing at its 2010 growth rate of 3.7 per annum, it would still need another 2 quarters to merely reach the level of output attained in 2008Q2 (i.e. by 2012Q3). Returning to the output level implied by the pre-crisis trend path would require an acceleration of growth beyond the currently projected growth rate. To the extent that this is not realized, the crisis implies a permanent output loss.

4. Estimates of potential output can help identify the implications for relevant policies going forward. Given the depth of the recession, it is crucial that the economy grows robustly for a sustained period of time to minimize the permanent output loss. This will be more difficult if the economy has also suffered losses to potential output. Restoring demand is insufficient in this case and appropriate supply side measures need to be taken as well. Studies of recoveries of previous recessions arising from financial crises suggest that recoveries are slower, on average, than those following other types of recessions (IMF 2009a) as was also evident in 1991.

B. Methods

5. Estimating the level of potential output is especially difficult in the aftermath of a recession. Estimates of potential output are always subject to considerable uncertainty since potential output is not directly measured. In the immediate aftermath of a recession, this uncertainty is increased even further. The frequently employed filtering techniques are subject to the end-point problems and are therefore less suited for computing output gaps at the end of the sample period.

6. Three different methods are employed to assess prospects for potential output growth. A standard HP filter, a production function approach (PF), and a multivariate approach (MV) are used. While the HP filter is a univariate approach and uses only the information derived from real GDP, the production function approach derives potential output from capital, labor, and total factor productivity (TFP) trends, which, in turn, are determined using an HP filter. Both these approaches are, however, subject to the end-period problem. The multivariate approach instead models the joint behavior of output, unemployment, capacity utilization, inflation, and inflation expectations. The MV approach can be thought of as using a reduced form model, estimated on data for Finland using Bayesian techniques, to infer the levels of potential output and the NAIRU that would be consistent with these observations.2

C. Results

7. The estimates of the output gap following the crisis are larger than for other advanced economies. The output gap is estimated to have been as large as -5 to -7¼ percent in 2009 (Figure 3.1). Estimates for Germany, the Netherlands, Belgium, and France of other studies that apply the MV approach range from -3 to -4 percent (Benes et al. 2010, IMF 2011, IMF 2012). However, it should also be noted that in 2009 the decrease in real GDP has been more severe in Finland (-8.4 percent) than in the other countries (-3½ percent in the U.S., -4.3 percent in EU-27, and -5½ percent in Japan, according to Eurostat).

Figure 3.1.Potential Growth and Output Gap Estimates

Source: Fund staff calculations

8. Potential output growth has been lowered temporarily by about 2 percentage points. Both the MV and PF approaches suggest a drop in the potential output growth of about 1 to 3 percent, while the HP filter suggests a much lower drop.3 The drop is largely accounted for by the fall in the contribution of labor to potential output growth, which fell from 1 percent in 2008 to -½ in 2009. The fall in capital from ¾ to ¼ percent accounts for the remaining ½ percent. Total factor productivity (TFP) instead remained relatively unchanged in 2010 after a trend decline since 2000 from 2¼ percent to below 1 percent contribution to growth.

9. A moderate recovery of potential output growth is projected. The production function approach indicates that potential output growth gradually recovers to 2 percent. The multivariate approach, which suggests a somewhat lower drop, implies a similar recovery path. The HP filter, which implies the lowest short-term impact on potential output growth suggests that potential growth remains depressed at around 1½ percent—to some extent a result of the end-period problem. Compared to the 1991 crisis, potential output growth has held up fairly well in the recent crisis and the output gap is expected to narrow quicker.

Finland - Contribution to Potential Output Growth


Sources: Fund staff calculations based on production function approach.

10. Results suggest a closing of the output gap by 2017. The multivariate approach provides the fastest narrowing of the output gap, reaching -1 percent already in 2011 (Figure 3.2). The production function approach implies a more gradual closing, reaching -1 percent only by 2015. The HP filter lies between the two. The difference is due to a lower MV potential growth forecast for 2011 compared to the forecast underlying the production function approach.

Figure 3.2.Output Gap, Potential Growth, and NAIRU 1/

Source: Fund staff calculations.

1/ Estimates of the potential output, output gap and NAIRU are based on the multivariate approach.

11. All estimates imply a permanent output loss. An estimate of the permanent output loss can be computed as the difference between the potential output path implied by the estimated potential output growth rates of the three models and a counter-factual output path that could have prevailed in the absence of the crisis. The counter-factual path is chosen by using the 2007 potential output growth value from the production function as starting point and the predicted value for 2017 as end-point. Growth rates for the other years are constructed by linear interpolation. The permanent loss of output under this calculation ranges from 6½ (HP) over 7½ percent (MV) to 8 (PF) percent. While these values are substantial, they are well below the estimate implied by the 1991 crisis using a similar procedure for the production function approach, which yields a loss of 14¾ percent.

D. Prospects

12. A declining labor force and limited advances in TFP growth are likely to be the main obstacles to potential output growth in the medium run. The Ministry of Finance expects a decline in the labor force of 70,000 by 2015 (MoF, 2011). This will depress the contribution from labor to potential output growth. To alleviate the reduction in available workers, better use of the existing work force is needed. Furthermore, there appears to be a trend decline in the contribution from TFP since 2000, which, if trends continue, is likely to weigh down potential output in the years ahead.

13. The contribution of capital to growth is likely to recover only slowly as uncertainty of the—primarily international—economic environment delays investment. Investment is largely driven by the performance of the export sector. The expected trade slowdown for 2012 and the uncertain outlook for Europe, is likely to retard firms’ investments. This implies that the capital stock is likely to recover only slowly, which is reflected in the still depressed gross capital formation to GDP ratio of 21 percent in 2011—close to 2 percentage points of GDP below the pre-crisis peak of 23 percent in 2007. Furthermore, the decline was mostly accounted for by a reduction in investment in machinery and equipment and non-residential construction.

Finland - Employment Rate of Older Workers

(percent of population aged 55–64)

Source:s OECD and Fund staff calculations.

E. Policies to Promote Growth

14. Labor force participation has to be enhanced further to compensate for the declining work force in the medium run. Female labor force participation is relatively high in Finland in comparison with the OECD average, while the male participation rate falls below the OECD and European averages. This is partly due to early retirement. To maintain robust support from labor to potential output growth, the effective retirement age should be increased and employment of older workers made more attractive.

Finland: Unemployment 2007–2011 annual & 2011M1–2012M3 monthly


Sources: Finland Ministry of Employment and the Economy and Fund staff calculations.

15. Incentives to seek work should be improved. The OECD (2010) notes that the implicit tax on further work is still high, which implies that the extra benefits from work relative to using the unemployment or disability benefit “pipeline” are very low. This effect is reinforced by relatively comfortable unemployment and disability benefits. While Finland’s average and marginal labor tax wedges have declined in the last 10 years, both wedges still remain well above the OECD average and discourage work.

Finland - Beveridge Curve

(percent of working force)

Sources: OECD and Fund calculations.

16. Improving job matching could enhance participation and reduce unemployment. The Beveridge curve has moved out over the last 20 years, as higher vacancies coexist with higher unemployment. At the same time long-term unemployment has been increasing significantly, indicating that there are obstacles in re-employing separated workers. Hynninen et al. (2009) point out that the matching may be improved by scaling up efficiency in the local labor market offices to the level of best practice. The experience with the increase in the eligibility age for the unemployment pipeline from 55 to 57 years in 2005 and the subsequent fall in the long-term unemployment rate for the 55–59 year-olds suggests that the incentive structure plays an important role in reducing matching inefficiencies. Further increases in the eligibility age could increase the employment rate of the older workers. Linking the eligibility age to life expectancy could be a way to implement such a system.

Finland - R&D 2010

(percent of GDP)

Sources: OECD and Fund staff calculations.

17. Initiatives to promote a diversified economy and limit the administrative burden could boost TFP growth. Finland maintains its role as leader in R&D investment, with investment in R&D equivalent to about 4 percent of GDP. The government’s plans to facilitate the administrative process for business and implement a one-stop-shop principle could benefit small and medium size enterprises, promote entrepreneurship and eventually enhance TFP growth.

F. References

    BenesJaromirMarianneJohnsonKevinClintonTroyMathesonPetarManchevRobertoGarcia-Saltos and DouglasLaxton2010Estimating Potential Output with a Multivariate FilterIMF Working Paper 10/285Washington: International Monetary Fund.

    HynninenSanna-MariAkiKangasharju and JaakkoPehkonen2009Matching Inefficiencies, Regional Disparities, and UnemploymentLabour Vol. 23(3) pages 481-50609.

    International Monetary Fund (IMF)2011Kingdom of the Netherlands – Selected Issues and Analytical NotesWashington: International Monetary Fund.

    International Monetary Fund (IMF)2012Belgium – Article IVWashington: International Monetary Fund.

    Ministry of Finance2011Europe 2020 – Strategy: Finland’s National Programme Update Autumn 2011.

    Organization for Economic Co-Operation and Development (OECD)2010Economic Review FinlandParis: Organization for Economic Co-Operation and Development.

Appendix 3.1. Details of the Multivariate Model of Potential Output

The approach of the multivariate model of potential output is to treat the unobserved levels of potential output (Ȳt), the NAIRU (Ut¯), and equilibrium capacity utilization (Ct¯) as latent variables. After specifying a system of economic relationships between observed output, unemployment, capacity utilization, inflation, and long-term inflation expectations, the parameters of the system and the latent variables can be simultaneously estimated using maximum likelihood and the Kalman filter.1

There are four main economic relationships. First, an inflation equation relates the level and the change in the output gap to observed annual inflation, π4:

where y is the output gap and επ4 denotes shocks to inflation expectations. A simple random walk extracts inflation expectation shocks from observed inflation expectations, π4LTE:

The (unobserved) unemployment gap, u, is related to the output gap by an Okun’s law relationship:

where εu is the shock term. Finally, the capacity utilization gap, c, is also related to the output gap:

Given these economic relationships, identification of the gaps is accomplished by relating the gaps to the levels of actual output, unemployment, capacity utilization, and inflation. This is done by estimating equilibrium, or potential, levels for each of output, unemployment, and capacity utilization. The respective laws of motion for potential output, unemployment, and capacity utilization are as follows:

where, in each equation, the G term is a damped autoregressive process, meaning that the trend rate of change itself is stochastic. The system is completed by three measurement equations, which are given by the definitions of the (log) output, unemployment, and capacity utilization gaps:

The following assumptions are made about steady-state levels:

Table Appendix 1.1:Steady-State Calibration Values
Trend growthGssY¯2.5
Long-run equilibrium unemploymentUss7.7
Labor share of incomeθ0.5

After estimation on Finnish data, the values of the dynamic parameters are:

Table Appendix 1.2.Estimated Parameter Values

Prepared by Sebastian Weber and Mika Kortelainen.

Details of the approach can be found in Benes et al (2010). The technical details of the model and its assumptions are presented in an appendix.

Results are based on a smoother value of 100 for annual data.

In practice, Bayesian methods are used to aid the estimation. The system is completed by adding the following equation:

Other Resources Citing This Publication