Kingdom of the Netherlands—Netherlands: Selected Issues and Analytical Notes

Important issues of the Netherlands are discussed. Openness to trade has benefited the Netherlands before the crisis and has supported the recent recovery process. However, both financial openness and trade linkages have also been a transmission channel for the financial crisis. Synchronized fiscal tightening across Europe has important spillover effects for GDP growth. The improvement on the supply side of credit has contributed to a normalization of the credit market. However, the recent increase in the financial stress index indicates that the situation is still fragile.

Abstract

Important issues of the Netherlands are discussed. Openness to trade has benefited the Netherlands before the crisis and has supported the recent recovery process. However, both financial openness and trade linkages have also been a transmission channel for the financial crisis. Synchronized fiscal tightening across Europe has important spillover effects for GDP growth. The improvement on the supply side of credit has contributed to a normalization of the credit market. However, the recent increase in the financial stress index indicates that the situation is still fragile.

Analytical Note 6: Asset Booms, Sectoral Changes, and the Estimation of Dutch Structural Fiscal Balances1

1. The calculation and use of structural fiscal balances has gained in importance in recent years for several reasons. Structural balances provide guidance as to the health and direction of fiscal policy and help determine the size and direction of automatic stabilizers. In addition, they are a key component in the assessment of long-run fiscal sustainability by providing a view of what the fiscal balance is likely to tend towards as temporary factors dissipate. Reflecting these considerations, the calculation of structural fiscal balances has taken a central position in the assessment of fiscal policy in the member countries of the European Union, under the Stability and Growth Pact (SGP). Under the SGP all member states are required to maintain a medium-term fiscal objective defined in terms of the structural fiscal balance.

2. However, there are well known measurement problems in calculating the structural balance. These include the estimation of potential output and output gaps, the adjustment of fiscal revenues for the effect of the business cycle using estimated revenue elasticities, and the question of whether adjustments for asset price cycles, changes in the shares of various components of national income, or other factors are also needed. Several approaches have been proposed for dealing with these challenges. However, the most prominent is the approach developed by the OECD (see Girouard and Andre, 2005), which adjusts for the business cycle but not for asset price cycles or changes in the composition of national income. A variant of the OECD approach was developed by the European Commission (see Larch and Turrini, 2009) to form the basis for calculating the structural balance under the SGP. In contrast, the European Central Bank (ECB) has developed a disaggregated approach which takes changes in the composition of national income into account (Bouthevillain and others, 2001).

3. The IMF’s calculations of the structural balance for the Netherlands have traditionally broadly followed the OECD approach as follows:2

  • Potential output is estimated using a Cobb Douglas production function, and output gaps can thus be derived.

  • Structural revenues are then calculated by using an aggregate elasticity of revenue with respect to the output gap, and the estimated output gap, to extract the cyclical component of revenue.

  • On the expenditure side it is assumed that the only type of expenditure with a cyclical component is unemployment-related benefits. Using data on the unemployment rate and estimates of the NAIRU (obtained using a HP filter on unemployment rate data), we are able to estimate the impact of the cycle on unemployment benefits and thus obtain the estimates of structural expenditure. The structural balance is then the difference between structural revenues and structural expenditures.

4. While this approach works well in many cases, recent events have highlighted its limitations under certain circumstances, such as property or other asset price booms.3 Indeed, European Commission (2008) examines why tax elasticities in several member countries appear to undergo substantial changes during economic booms, and recommends extending the assessment of tax elasticities to incorporate a broader number of explanatory variables. There is also a substantial literature—see Eschenbach and Schuknecht (2002), Girouard and Price (2004), and Morris and Schuknecht (2007), among others—that finds that excluding asset prices from the analysis can lead to serious biases in the estimation of the structural balance.

5. This note expands the methodology by taking explicit account of asset prices and sectoral shifts in the economy. Given the potentially costly policy errors that can result from biases in the calculation of structural balances, we then use this expanded methodology to explore whether the standard approach has been robust to the influence of these factors in the case of Netherlands. Our broad conclusion is that Dutch fiscal revenues have not been substantially influenced by asset price cycles or sectoral shifts, and taking explicit account of these factors leads to relatively modest revisions in the estimates of the structural balance.

A. Dutch Experience

6. GDP contracted sharply in 2009, following years of robust growth, as a result of the global financial crisis and the associated shock to global trade. As a result, the economy contracted 4 percent in 2009. Alongside, the fiscal position deteriorated sharply, moving from a surplus of ½ percent of GDP in 2008 to a deficit of 5½ percent of GDP in 2009. Revenues fell a bit faster than GDP and as a result the revenue-GDP ratio declined by one percent of GDP. However, different revenue categories reacted very differently to global shocks in 2009, with direct taxes increasing slightly in percent of GDP whereas indirect taxes and social contributions showed significant declines in percent of GDP. This suggests that an approach where nominal GDP is essentially the sole determinant of revenues may be prone to bias in times of substantial changes to GDP, which are often the times when policy errors can be most costly.

7. House prices have tended to appreciate faster than nominal GDP, whereas equity prices have tended to appreciate at a slower pace, albeit with greater volatility. In general, the two asset prices that have been found to have influence on fiscal revenues across various countries are house and equity prices. Data for Netherlands indicates that over the past 40 years, house prices have risen at a slightly faster pace on average than nominal GDP, with the increase relatively more pronounced since the mid-1990s. On the other hand, equity prices have tended to grow on average at a pace lower than that of GDP, but with much more volatility than GDP. This then suggests that any revenue source that is significantly influenced by house prices is likely to have been rising in percent of GDP in recent years, whereas a revenue source significantly influenced by equity prices would have shown significant volatility but with a slightly negative trend overall and would have suffered a disproportionately large decline in 2009.

uA06fig01

Nominal GDP vs housing and equity prices 1970=100

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Sources: OECD, Haver, and author’s calculations

8. Asset prices do not appear to have had a major impact on fiscal revenues in recent years. Figure 6-1 presents a breakdown of fiscal revenues over 1970–2006. The overall revenue/GDP ratio rose to a peak in the mid-1908s and subsequently declined, with the bulk of the decline concentrated in social contributions and other current revenue. The most important item under other current revenue is property income, and so the movements in this category could be reflecting changes to earnings from exports related to the gas industry. Direct and indirect taxes have been broadly stable over the period, but indirect taxes do appear to have been on a modestly rising trend in recent years, and also declined in 2009. Overall, however, the impression one gets is that the various revenues appear to have been broadly stable in relation to GDP in the past decade, which would suggest that asset price movements have not had a major influence on revenues. Figure 6-2 generally confirms that the growth rates of the various revenue categories have on average been close to that of nominal GPD, and we do not see evidence of any bubble-induced divergence that could signal the onset of significant bias in the calculation of structural revenues using the standard approach.

Figure 6-1.
Figure 6-1.

Netherlands: Revenue Breakdown, 1970-2009

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Sources: Eurostat, CPB, and Author’s calculations.
Figure 6-2.
Figure 6-2.

Netherlands: Comparison of Revenue and Nominal GDP Growth Rates, 1970-2009

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Sources: Eurostat, CPB, and Author’s calculations.

9. Use of the “eyeball” metric only serves to give broad impressions, however, and more rigorous methods are needed to come to grips with the issue. The next section outlines an estimated model for fiscal revenues that takes into account asset prices and sectoral changes based on a disaggregated approach, in an effort to provide a more complete assessment of what portion of fiscal revenues could be considered structural. The estimated equation could be thought of as a reduced form of a structural model that links revenue and its base to a set of explanatory macro variables.

B. The Model4

10. The estimated equation depends on whether the data are stationary or not, and we present below the model in both cases.

Model with Stationary Data

11. With stationary data, the model is given by equation (1) below:

lnRt=Σi=0ma0ilnMit+Σi=0ma1ilnMit1++Σi=0malilnMitl+εt(1)

where Rt represents revenue in time t;Mit represents the value of the ith explanatory variable in time t; ali represents the elasticity of revenue to the lth lag of the ith explanatory variable; and εt is the error term.

Taking exponents of both sides of equation (1) to eliminate the natural logs, and using a star and hat to represent the structural level of the variable and the estimated value of the coefficient, respectively, then the estimated structural revenue is given by equation (2) as:

Rt*=RtΠi=0m(Mit*Mit)a^0iΠi=0m(Mit1*Mit1)a^1iΠi=0m(Mit1*Mit1)a^lt(2)

Model with Non-stationary Data

12. Where the data are nonstationary, it is often fruitful to first explore if a stationarity inducing transformation is possible. This can substantially simplify the estimation process and the calculation of structural revenues. Division by GDP is a good candidate here, because the ratios of all the macro variables used (and underlying tax bases) to GDP are likely to be bounded, and may therefore be more likely to be stationary. Indeed, we find this to be the case for all the variables used in the estimation in this paper.

13. With all variables divided by GDP, the approach is very similar to that used in the case of stationary data. The main difference is that the variables are interpreted as ratios to GDP. On this basis, dividing both numerator and denominator of the ratio by (Mit*/Mit) by nominal GDP (Yt) and denoting the resulting scaled variable by mt, the ratio becomes (mit*/mit). Thus we have the structural value of the revenue/GDP ratio given by:

rt*=rtΠi=0m(mit*mit)a^0iΠi=0m(mit1*mit1)a^1iΠi=0m(mit1*mit1)a^lt(3)

where rt represents the ratio of revenue to GDP.

14. If scaling by GDP or other such transformation does not work, the model can be estimated in first differences, or an error correction model. For example, equation (4) below, which allows for cointegration, could be used.

ΔlnRt=Σi=0matΔlnMitλ(lnRt1Σi=0mμilnMit1)+εt(4)

In this case, the estimation results are in relation to growth rates rather than levels of the variables, and the structural revenue is determined by the formula:

Rt*=Rt(Rt1*Rt1)(1λ)Πi=0m(Mit*Mit)a^iΠi=0m(Mit1*Mit1)(λ^μ^ia^i)(5)

Since the right-hand side includes the lagged value of structural revenue, pinning down actual values for the level of structural revenue will require additional assumptions to identify the level of structural revenue in at least one year.

C. Estimation Results

15. Log-linear regressions were estimated for all the revenue categories on annual data for 1970–2008. Data used were all divided by GDP. For all variables (in percent of GDP), the unit root hypothesis is rejected using the Ng-Perron test (Table 6-1). Ng and Perron (2001) show that this test generally has superior power and size properties compared to the traditional Dickey-Fuller and Phillips-Perron tests. Division by GDP has the implication that the regression equations seek to explain movements in the ratios of various revenue types to GDP, rather than their levels. In addition, GDP itself does not enter any equation as a stand-alone explanatory variable.

Table 6-1.

Ng Perron Unit Root Tests 1/

article image
Source. Author’s calculations

All variables are in logs and in percent of GDP.

MZa statistic, as in Eviews; *** and ** represent rejection of the unit root hypothesis at the 1 and 5 percent levels, respectively. Critical values: 1% level, -23.8; 5% level, -17.3.

16. A variety of macroeconomic variables were examined for explanatory power in the regression exercise. In each equation, variables and lags were added or dropped on the basis of significance, goodness of fit, and implications for serial correlation of the residuals. At the end, the variables that showed significant explanatory power in the equations were nominal consumption, nominal residential investment, house price index, equity price index, exports, and wages. It does not appear plausible that any of these explanatory variables is significantly influenced by movements in any single revenue type, so OLS—which is what is used—should yield consistent estimates. Figure 6-3 presents these variables and their movement over the estimation period.

Figure 6-3.
Figure 6-3.

Netherlands: Revenue Determinants, 1970-2009

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Sources: Statistics Netherlands and Author’s calculations

17. Most equations were estimated with high precision, with coefficients generally plausible (Table 6-2). As all variables are in percent of GDP a positive or negative sign on a coefficient for a regressor does not translate to a similarly signed relationship between the respective variables in levels. In particular, a positive coefficient simply means that faster-than-GDP increases in the explanatory variable are associated with faster-than-GDP increases the revenue type under consideration, while a negative coefficient means that faster-than-GDP increases in the explanatory variable are associated with slower-than-GDP increases in the revenue type under consideration.

Table 6-2.

Regresssion Estimates 1/2/

article image
Source. Author’s calculations.

All variables are in logs and in percent of GDP.

***, **, and * represent significance at the 1 percent, 5 percent, and 10 percent levels, respectively.

D. Calculating Structural Revenues

18. Estimates of the structural values of the explanatory variables are obtained using the HP filter. Calculating the structural revenue/GDP ratio for each revenue type requires us to obtain estimates for the ratios (mit*/mit), which measure how far each explanatory variable is from its structural or fundamental value. To do so, we need to generate estimates for the structural values mit* of the explanatory variables. One way is to use a smoothing technique, such as the HP filter, to extract the trend value of the variable, which is then treated as the structural value. This is the approach taken by Morris and Schuknecht (2007) among others.

19. Figure 6-4 presents the results of using the HP filter on all the explanatory variables. Our focus is on measuring structural revenues over the historical period up to 2009. However, to minimize end-point problems, projections up to 2015 are added. The extraction of trend appears satisfactory in all cases. Thus, for all explanatory variables we use the HP filtered values as our estimates of the structural values. Overall structural revenue is then the sum of the structural levels of the different revenue types.

Figure 6-4.
Figure 6-4.

Netherlands: Revenue Determinants and HP-Filtered Trends, 1970-2015

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Sources: Statistics Netherlands and Author’s calculations.

20. For all revenue types there has been not been much divergence between actual and structural levels in percent of GDP. Figure 6-5 presents the actual and structural values (from the estimated equations in Section III) of the various revenue types. Thus, the traditional manner of calculating structural revenues does not appear to have led to substantial biases.

Figure 6-5.
Figure 6-5.
Figure 6-5.

Netherlands: Actual and Structural Revenues, 1970-2009

(In percent of GDP)

Citation: IMF Staff Country Reports 2011, 143; 10.5089/9781455286645.002.A006

Source: Eurostat, CPB, and Author’s calculations.

E. Structural Expenditures and Structural Balance

21. As indicated above, unemployment benefits are assumed to be the only component of fiscal expenditure with a significant cyclical component. On this basis, structural expenditure, E,* can be expressed as follows:

E*=Es+(NAIRUUR)U(6)

Structural expenditure is then divided by potential GDP and this ratio is then subtracted from the structural (revenue/GDP) ratio to obtain the estimated structural balance in percent of potential GDP.

22. Differences in the estimates of the structural balance between our approach and the standard one are modest. Table 6-3 presents the estimates of structural revenues, structural expenditure, and the structural balance. Interestingly it indicates that the deterioration in the structural balance in 2007 may have been overstated by the estimates derived using the standard approach. On the other hand, the deterioration of 2009 was very similar under the two approaches, though the estimated structural balance under the disaggregated approach is about ½ percent of GDP better than suggested by the standard approach. For 2010, the disaggregated approach indicates no improvement in the structural position, whereas the standard approach points to modest tightening. Nevertheless differences between the two estimates are generally not very large, suggesting that the standard approach has been basically robust to the effects of asset prices and sectoral changes in the case of Netherlands.

Table 6-3.

Structural Fiscal Position 1/

article image
Sources: Eurostat, CPB, and author’s calculations.

In percent of potential GDP unless otherwise stated.

References

  • Bouthevillain, C., P. Cour-Thimann, G. van den Dool, P. Hernandez de Coz, G. Langenus, M. Mohr, S. Momigliano, and M. Tujula, 2001, “Cyclically Adjusted Budget Balances: An Alternative Approach,” European Central Bank, Working Paper No. 77.

    • Search Google Scholar
    • Export Citation
  • Eschenbach, F., and L. Schuknecht, 2002, “Asset Prices and Fiscal Balances,” European Central Bank, Working Paper No. 141.

  • European Commission, 2008, “Public Finances in EMU—2008,” European Commission, Directorate-General for Economic and Financial Affairs, European Economy 4/2008.

    • Search Google Scholar
    • Export Citation
  • European Commission, 2009, “Public Finances in EMU—2009,” European Commission, Directorate-General for Economic and Financial Affairs, European Economy 5/2009.

    • Search Google Scholar
    • Export Citation
  • Fedelino, A., A. Ivanova, and M. Horton, 2009, “Computing Cyclically-Adjusted Balances and Automatic Stabilizers,” International Monetary Fund, Fiscal Affairs Department, Technical Notes and Manuals 09/05.

    • Search Google Scholar
    • Export Citation
  • Girouard, N., and C. Andre, 2005, “Measuring Cyclically-adjusted Budget Balances for OECD Countries,” OECD Economics Department Working Papers, No. 434, OECD Publishing.

    • Search Google Scholar
    • Export Citation
  • Girouard, N., and R. Price, 2004, “Asset Price Cycles, “One-off” Factors and Structural Budget Balances,” OECD Economics Department Working Papers, No. 391, OECD Publishing.

    • Search Google Scholar
    • Export Citation
  • Horton, M., M. Kumar, and P. Mauro, 2009, “The State of Public Finances: A Cross-Country Fiscal Monitor,” IMF Staff Position Note 09/21, http://www.imf.org/external/pubs/ft/spn/2009/spn0921.pdf.

    • Search Google Scholar
    • Export Citation
  • IMF, 2009, “The State of Public Finances: Outlook and Medium-Term Policies after the 2008 Crisis,” http://www.imf.org/external/np/pp/eng/2009/030609.pdf.

    • Search Google Scholar
    • Export Citation
  • Joumard, I., and C. Andre, 2008, “Revenue Buoyancy and its Fiscal Policy Implications,” OECD Economics Department Working Papers, No. 598, OECD Publishing.

    • Search Google Scholar
    • Export Citation
  • Kanda, D., 2010, “Asset Booms and Structural Fiscal Positions: The Case of Ireland,” International Monetary Fund, Working Paper No. WP/10/57.

    • Search Google Scholar
    • Export Citation
  • Larch, M. and A. Turrini, 2009, “The Cyclically-adjusted Fiscal Balance in EU Fiscal Policy making: A Love at First Sight Turned into a Mature Relationship,” European Commission, Directorate-General Economic and Financial Affairs, European Economy Economic Papers 374.

    • Search Google Scholar
    • Export Citation
  • Morris, R., and L. Schuknecht, 2007, “Structural Balances and Revenue Windfalls: The Role of Asset Prices Revisited,” European Central Bank, Working Paper No. 737.

    • Search Google Scholar
    • Export Citation
  • Ng, S., and P. Perron, 2001, “Lag Length Selection and the Construction of Unit Root Tests with Good Size and Power,” Econometrica, Vol. 69, pp. 151954.

    • Search Google Scholar
    • Export Citation
1

Prepared by Daniel Kanda.

2

This approach is similar to that used by the IMF’s Fiscal Affairs Department—described in Fedelino, Ivanova, and Horton (2009)—in recent publications such as Horton, Kumar, and Mauro (2009) and IMF (2009).

3

An important example of this being the case of Ireland, where a property boom led to substantial distortions in the calculation of the structural balance, which continued to signal healthy fiscal positions even as underlying vulnerabilities mounted.

4

This note abstracts from consideration of the effects of discretionary measures on tax elasticities. European Commission (2009) attempts to adjust revenue data for discretionary measures in a number of countries (excluding Netherlands), and finds that overall this has a relatively small impact on the tax elasticity estimates, though in some cases substantial divergences are observed in certain years. In any case, the bias is likely to be small in our exercise for a sufficiently long time series that allows positive and negative errors arising from discretionary measures to mostly cancel each other out.

Kingdom of Netherlands: Netherlands: Selected Issues and Analytical Notes
Author: International Monetary Fund