Kingdom of the Netherlands—Netherlands: Selected Issues

International Monetary Fund

doi:10.5089/9781451829556.002.A003

Jump to Content Jump to Main Navigation

Kingdom of the Netherlands—Netherlands

Selected Issues

Author: International Monetary Fund

Publication Date:: 31 Jul 2006

eISBN:: 9781451829556

Language:: English

Keywords:: ISCR; CR; TFP growth; tax revenue; price; income tax; deductibility; tax deductibility; house price; interest payment; revenue volatility; trend growth; Housing prices; Total factor productivity; Personal income tax; Mortgages; Global; Europe; conditional mean; volatility in the Netherlands; social contribution; tax revenue base; direct tax tax revenue; Social security contributions; Corporate income tax; Tax allowances

Download PDF (474.6 KB)

46. The recent fiscal ROSC for the Netherlands and related aide-mémoire had noted that volatility in tax revenues appeared to have increased over the last few years. 41 One manifestation of this had been in the difficulties encountered in forecasting these revenues, particularly from direct taxes, since the late 1990s. This paper provides a more extensive investigation of tax revenue volatility in Netherlands to assess the robustness of the earlier conclusions. In addition, it explores the extent to which other European countries have experienced similar changes in volatility. 42 The paper then examines some of the determinants of volatility in both personal income tax and corporate tax revenues in the Netherlands, and compares these to the determinants in other EU countries: in particular, it examines the role of changes in the amplitude of the business cycle and in the tax revenue base (specifically wages), as well as factors relating to corporate location decisions and profits.

Abstract

III. Volatility of Tax Revenues in the Netherlands^⁴⁰

A. Introduction

46. The recent fiscal ROSC for the Netherlands and related aide-mémoire had noted that volatility in tax revenues appeared to have increased over the last few years. ^⁴¹ One manifestation of this had been in the difficulties encountered in forecasting these revenues, particularly from direct taxes, since the late 1990s. This paper provides a more extensive investigation of tax revenue volatility in Netherlands to assess the robustness of the earlier conclusions. In addition, it explores the extent to which other European countries have experienced similar changes in volatility. ^⁴² The paper then examines some of the determinants of volatility in both personal income tax and corporate tax revenues in the Netherlands, and compares these to the determinants in other EU countries: in particular, it examines the role of changes in the amplitude of the business cycle and in the tax revenue base (specifically wages), as well as factors relating to corporate location decisions and profits.

47. A noticeable characteristic of the Dutch economy, which sets it apart from most of the other EU countries, is its extensive and well-established private pension system with defined pension benefits and contributions. The contributions are tax-deductible and can have implications for personal and corporate income tax contributions. This is because any marked changes in the value of pension fund assets, which are heavily invested in equity markets, are likely to entail changes in pension fund contributions, by employees or employers, with a direct impact on tax revenues. There are also likely to be indirect effects on tax revenues of changes in these contributions, arising from the impact on employment and corporate profitability. Given the potential magnitude of this channel in explaining revenue volatility, it is also examined in some detail below.

48. The empirical results suggest the following five main conclusions: (i) there is evidence of a substantial increase in tax revenue volatility in the Netherlands over the past decade compared to the preceding periods; (ii) the increase in volatility appears to have occurred during a period when on average there was a decline in tax revenue volatility in the EU-15 countries; (iii) this has led the ranking of the Netherlands in tax revenue volatility relative to other countries to increase over the past decade, particularly with regard to personal income tax and social security contributions; (iv) there is some evidence that business cycle volatility in the Netherlands has increased, while it seems, if anything, to have declined in many EU countries, which may account for some of the increased volatility in tax revenues; (v) the unique features of the Dutch pension system, commendable in many respects, may nonetheless have contributed to higher tax volatility through the tax deductibility of contributions.

49. The rest of the paper is organized as follows: Section B provides systematic empirical evidence on a key measure of volatility in different sources of tax revenues—income tax, corporate tax, indirect taxes—over the past three decades, and examines the extent to which it has changed in recent years. This analysis is undertaken for both the Netherlands, and other EU countries. Section C examines the role of business cycle fluctuations, the impact of the Dutch pension fund system, corporate location decisions, and other specific factors. Section D provides conclusions and notes some policy implications.

B. Tax Revenue Volatility in the Netherlands and in other EU-15 Countries

Experience of the Netherlands

50. There are a variety of indicators that have been used to assess tax revenue volatility in the OECD countries. The analysis below relies on the standard, and widely used, measure that takes into account fluctuations in tax revenues from period to period (for instance, quarterly or annual frequency), adjusted for the average values of revenues in a given period. Formally, the measure—the coefficient of variation (CV)—is defined as the standard deviation relative to the mean of the ratio of tax revenues to GDP. As a robustness check, this measure is supplemented by econometric evidence using regression analysis.

51. The basic results of using the CV measure for the Netherlands are provided in Table 1. This shows the magnitude of volatility in revenues from three types of taxes: personal income tax and social security contributions, corporate income tax, and indirect taxes, for the three decades since 1975. As the results using quarterly data indicate, for the most recent period, 1995–2004, the highest level of volatility (15.9 percent) is observed in corporate income tax, followed by indirect taxes, and lastly by personal income tax and social security contributions. ^⁴³ The changes in volatility over time are particularly striking: for direct taxes as a whole (personal income tax including social contributions and corporate taxes), the CV increased from 2.5 percent during 1975–84, to 3.0 percent in the following decade, and to 4.4 percent in the decade beginning 1995. The difference between the first period, which saw significant economic volatility associated with the two oil price shocks, and the most recent period which has also seen considerable economic as well as financial market volatility, is highly significant. There has also been a pronounced increase in volatility in revenues from indirect taxes in the most recent period compared to the preceding period, although in this case, relative to the first period there was a less sharp increase. The above conclusions are robust to the frequency of data—analysis based on annual data (Table 2) yields very similar results.

^{Table 1.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Quarterly

Source: Quarterly data from OECD Economic Outlook 78.

^{Table 1.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Quarterly

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	2.5	3.0	4.4
- Corporate income tax	8.2	5.5	15.9
- Personal income tax and social contributions	2.8	3.1	4.4
Indirect taxes	4.4	3.3	6.2
Memo: Wage bill	3.8	1.7	1.2

Source: Quarterly data from OECD Economic Outlook 78.

^{Table 1.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Quarterly

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	2.5	3.0	4.4
- Corporate income tax	8.2	5.5	15.9
- Personal income tax and social contributions	2.8	3.1	4.4
Indirect taxes	4.4	3.3	6.2
Memo: Wage bill	3.8	1.7	1.2

Source: Quarterly data from OECD Economic Outlook 78.

^{Table 2.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Annual

Source: Annual data from OECD Economic Outlook 78.

^{Table 2.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Annual

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	2.2	2.9	4.5
- Corporate income tax	8.0	5.3	16.3
- Personal income tax and social contributions	2.5	3.0	4.4
Indirect taxes	4.2	3.2	6.4
Memo: Wage bill	3.6	1.5	1.2

Source: Annual data from OECD Economic Outlook 78.

^{Table 2.}

Coefficient of Variation of Tax Revenues (Share of GDP) in the Netherlands, Annual

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	2.2	2.9	4.5
- Corporate income tax	8.0	5.3	16.3
- Personal income tax and social contributions	2.5	3.0	4.4
Indirect taxes	4.2	3.2	6.4
Memo: Wage bill	3.6	1.5	1.2

Source: Annual data from OECD Economic Outlook 78.

52. Looking at the components of direct tax revenues, the most marked increase has been in the volatility of corporate income taxes, which more than doubled from 1975-84 to 1995-2004 despite a noticeable drop in the interim ten-year period. Such an increase is significant, and makes the forecasting of such revenues particularly challenging. However, as corporate taxes account for around 8 percent of the total tax receipts, their impact on the overall revenue volatility of the government is likely to be limited. This is not the case for personal income tax and social security contributions, which account for almost 60 percent of total revenues. The volatility for the latter has increased substantially, rising from 3.1 percent during the 1985–94 period, to 4.4 percent during 1995–2004.

53. The above results using the CV measure are supported by formal econometric analysis, based on a variety of regression models. The first model considers the conditional volatility by estimating the autoregressive pattern in the personal income tax revenues series^⁴⁴, and exploring the extent to which the residuals (taking into account the past history of personal income tax receipts) show any marked changes in recent years. As Figure 1 illustrates, there has indeed been a noticeable increase in the volatility over the past decade, compared to the earlier periods. A second model considers the conditional volatility by examining the extent to which changes in the main tax base, i.e. wages, account for the dynamics in personal income tax revenues. As the second panel of Figure 1 shows, the volatility in personal income tax revenues relative to the tax base—i.e., deviation from the conditional mean—also increased substantially in the period 1995–2004. This issue is examined in more detail below, in the context of the determinants of tax revenue volatility in the Netherlands.

Tax revenue volatility in a European context

54. An immediate question that arises is the extent to which the above developments reflect an EU-wide phenomenon, or whether they are peculiar to the Netherlands. If it is the former, there would be a prima facie case for examining EU-wide factors, as well as specific Dutch factors, that could have led to such a development. The empirical analysis using the CV measures suggests that the experience in the other European countries has not been the same. Indeed, for many countries the opposite appears to have been the case, with tax revenue volatility on average declining over the past decade.

55. For the composite category of direct taxes and social contributions, as Table 3 indicates, the CV for the average EU-15 declined from 6.8 percent in 1975–84 to 4.9 percent in 1985–94, and to 3.2 percent in 1995–2004. There was a corresponding decline in personal income tax and social security contributions from 7.2 percent in 1975-84 to 3.4 percent in 1995–2004. In the corporate taxes, after a small increase during the second decade, there was a marginal decline in the most recent period. Despite the small decline, it is the case, somewhat surprisingly, that on average revenues from the corporate income tax in the EU countries are more volatile than in the Netherlands. For both the Netherlands, and EU in general, this reflects a more volatile tax base than most other taxes and often more flexible rules for tax payments such as loss carry-over, payment lags, etc.

^{Table 3.}

Coefficient of Variation of Tax Revenues (share of GDP) in EU countries, Annual

Source: Annual data from OECD Economic Outlook 78.

^{Table 3.}

Coefficient of Variation of Tax Revenues (share of GDP) in EU countries, Annual

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	6.8	4.9	3.2
- Corporate income tax	16.9	19.9	18.3
- Personal income tax and social contributions	7.2	4.8	3.4
Indirect taxes	5.8	4.0	4.0
Memo: Wage bill	2.6	2.2	2.0

Source: Annual data from OECD Economic Outlook 78.

^{Table 3.}

Coefficient of Variation of Tax Revenues (share of GDP) in EU countries, Annual

	1975–84	1985–94	1995–2004
Direct taxes and social contributions	6.8	4.9	3.2
- Corporate income tax	16.9	19.9	18.3
- Personal income tax and social contributions	7.2	4.8	3.4
Indirect taxes	5.8	4.0	4.0
Memo: Wage bill	2.6	2.2	2.0

Source: Annual data from OECD Economic Outlook 78.

56. With regard to indirect taxes, the volatility for the average EU country also appears to have followed a different pattern than that in the Netherlands. During the 1975–84 period, the volatility for the average EU country was higher than that in the Netherlands. As in the Netherlands, the volatility declined in the subsequent period, but then unlike in the Netherlands, it remained unchanged over the subsequent period, and is now appreciably lower.

57. To explore the cross-country variation that underlies the above results, we examined the experience of individual countries in some detail. The results of that exercise are summarized in Figures 2–4. As the first of these Figures indicates, in Netherlands the volatility in personal income tax was almost the lowest in the EU-15 countries during the 1975–84 period. In the following decade, as average volatility declined, Netherlands’ relative position was almost unchanged. It is during the last decade that, while average volatility declined still further, volatility in the Netherlands continued to increase, pushing its ranking up to near the top, exceeded only by Ireland.

58. With regard to corporate income taxes, developments were quite different (Figure 3). The Netherlands—with the third lowest level of corporate income tax revenue volatility—had a relative position at the lower end of the distribution during the first period, and it improved even further during the second period when it had the lowest volatility in EU-15. Even though there has been a rebound in the most recent period, in relative terms, the Netherlands remains in the lower half of the distribution. The experience with regard to indirect taxes is different yet again (Figure 4), with the Netherlands moving from almost the middle of the range in the first two decades, to almost near the top in the most recent period.

C. Determinants of Tax Revenue Volatility in the Netherlands

59. The above evidence suggests clearly that there has been a substantial increase in fluctuations in tax revenues in the Netherlands over the past decade and, particularly with regard to personal income tax, that the increase has been significantly greater than in most other EU countries. This section explores why this might have been so. The empirical analysis, focusing on direct taxes, starts by investigating whether there may have been an increase in the amplitude of the business cycles, and in the volatility in the wage bill. As mentioned in the introduction, the volatility in pension premiums could have had a substantial impact on tax revenues, and that is discussed in next. This is followed by a discussion of other factors that might have had a bearing, such as issues related to corporate location decisions, fluctuations in house prices, and the conduct of fiscal policy.

The role of the business cycle

60. An immediate question related to revenue volatility is how much of it reflects cyclical changes in the economy and how much is due to some basic underlying characteristics of the tax structure. If the cyclical changes are becoming more pronounced, this will directly amplify tax revenue fluctuations, particularly if the elasticity of tax revenues to activity is larger than unity.

61. For an assessment of this factor, we considered fluctuations in the output gap in the Netherlands, as well as in some EU countries. ^⁴⁵ As Table 4 indicates, measured by the average deviations from the mean, fluctuations in the Netherlands’ output gap have widened noticeably over the past decade relative to the preceding one. Indeed, output gap volatility in the past ten years appears to have been even larger than during the 1975-84 period, when economic activity was significantly buffeted by the two oil crises, and ensuing economic and financial market turbulence.

^{Table 4.}

Output Gap Volatility

Source: Quarterly data from OECD Economic Outlook 78.Note: Volatility is measured as the average absolute deviation from mean

^{Table 4.}

Output Gap Volatility

	1975–84	1985–94	1995–2005	2000–05
France	1.2	1.5	0.9	0.9
Ireland	2.8	2.2	2.6	1.8
Italy	1.9	1.4	0.9	0.9
Sweden	2.0	2.9	1.4	0.9
United Kingdom	1.9	2.5	0.5	0.4
Germany	2.3	2.0	1.1	1.5
Average	2.0	2.0	1.3	1.3
The Netherlands	1.8	1.4	1.9	2.4

Source: Quarterly data from OECD Economic Outlook 78.Note: Volatility is measured as the average absolute deviation from mean

^{Table 4.}

Output Gap Volatility

	1975–84	1985–94	1995–2005	2000–05
France	1.2	1.5	0.9	0.9
Ireland	2.8	2.2	2.6	1.8
Italy	1.9	1.4	0.9	0.9
Sweden	2.0	2.9	1.4	0.9
United Kingdom	1.9	2.5	0.5	0.4
Germany	2.3	2.0	1.1	1.5
Average	2.0	2.0	1.3	1.3
The Netherlands	1.8	1.4	1.9	2.4

Source: Quarterly data from OECD Economic Outlook 78.Note: Volatility is measured as the average absolute deviation from mean

62. An increase in economic fluctuations is likely to lead to an increase in tax revenue volatility. But its magnitude depends crucially on the associated short and long-term revenue elasticities with respect to the output gap. As they reflect different attributes of the tax system, these elasticities can differ substantially. In general, tax systems are structured such that, in the absence of discretionary changes, nominal tax revenues follow income growth in the long-term, implying that long-term elasticities of tax revenues to income are close to unity. This attribute of the tax system holds for the Netherlands where cumulative tax revenue growth was about the same as cumulative wage income growth.

63. The long-term elasticity, however, does not provide information about the path of tax revenues and the economy during specific periods. Such information can be provided by short-term elasticity estimates as these measure the responsiveness of tax revenues to, say, yearly changes in tax bases. In general, short-term tax revenue elasticities tend to be larger than unity, which in the case of personal income taxes reflects a certain amount of progressivity in the tax structure. The interpretation of a corporate income tax elasticity higher than unity is more ambiguous, but relates to the non-symmetrical tax treatment of profits and losses (a firm pays taxes if it makes a profit, but it does not receive a refund for tax losses) and the provisions for carrying losses forward into other tax years.

64. The estimates for the Netherlands by the OECD suggest the short-term elasticity of personal income taxes and social contributions to the output gap is 1.25 percent (see Andre and Girouard, 2005). ^⁴⁶ This means that tax revenues increase, on average, 25 percent faster than the economy in upturns and decrease by roughly 25 percent more than the economy during downturns. For corporate income taxes, the short-term elasticity is somewhat higher at 1.5 percent, indicating that corporate income taxes are even more responsive to fluctuations in the pace of economic activity.

65. Since the short-term elasticity of tax revenues to the output gap is higher than unity this suggests that not only has nominal tax revenue volatility increased, but also the volatility of tax ratios. The volatility of nominal tax revenues and tax ratios associated with higher output gap volatility is shown in Table 5, indicating that higher business cycle volatility clearly explains part of the observed increase in tax revenue volatility.

^{Table 5.}

Average Absolute Deviation from the Unconditional Mean

(In percentages)

Note: Volatility of nominal revenues is calculated by multiplying the average absolute deviation of the output gap by the short-term elasticities, while volatility of tax ratios is calculated by multiplying the average absolute deviation of the output gap by the short-term elasticities minus 1.

^{Table 5.}

Average Absolute Deviation from the Unconditional Mean

(In percentages)

		Nominal Revenue		Tax Ratio
	Output gap	PIT and social contributions	CIT	PIT and social contributions	CIT
1975–84	1.8	2.3	2.8	0.5	0.9
1985–94	1.4	1.8	2.2	0.4	0.7
1995–2005	1.9	2.4	2.9	0.5	1.0
2000–2005	2.4	3.0	3.6	0.6	1.2

^{Table 5.}

Average Absolute Deviation from the Unconditional Mean

(In percentages)

		Nominal Revenue		Tax Ratio
	Output gap	PIT and social contributions	CIT	PIT and social contributions	CIT
1975–84	1.8	2.3	2.8	0.5	0.9
1985–94	1.4	1.8	2.2	0.4	0.7
1995–2005	1.9	2.4	2.9	0.5	1.0
2000–2005	2.4	3.0	3.6	0.6	1.2

66. As noted earlier, the standard deviation of the difference between personal income tax revenues and its conditional mean—where the conditional mean is determined by the fitted values from a regression of personal income tax revenues on wage income as a percent of GDP—has increased substantially in the last 15 years. This indicates an increasing margin of error imbedded in the forecasting of personal income tax revenues and social contributions. To illustrate the impact of this increased margin on tax revenue projections, consider an increase of 2 percent in wages. Based on an estimated personal income tax revenue elasticity of about 1.25 (from the OECD as mentioned above), we would expect personal income tax revenues and social contributions to grow by 2½ percent. The actual rate of increase in personal income tax revenues and social contributions, using the estimated increased volatility in recent years (that is, the deviation of the rate of growth from its expected value in Table 5), would vary between 1.3 and 3.7 percent—a margin of 2.4 percentage points. The same increase in wage income 20 years ago would have led to an increase in personal income tax revenues and social contributions between 1.8 percent to 3.2 percent, a smaller margin of 1.4 percentage points. ^⁴⁷

67. A formal econometric analysis shows that business cycle fluctuations account for a significant part of the increase in tax revenue volatility. An estimate of the magnitude of this effect, however, varies between 0.05 and 0.08 depending on the specification of the estimating equations. The different specifications imply that, for example, a one percent increase in the output gap volatility would lead to an increase in the volatility of tax revenue that varies between 0.05 and 0.08 percent. (See Appendix for details)

The role of pension funds

68. The unique feature of the Dutch pension system may have contributed to the increased volatility in tax revenues. Thus, while the above analysis suggests that greater economic fluctuations played a role in increasing tax revenue volatility, they may not have been the sole explanations. Other factors that have received at least some recognition in the literature include the dynamics related to the premiums for the Dutch private pension system (see van Ewijk and van de Ven, 2003). The Dutch system, highly acclaimed as it rightly should be, relies on private pension schemes as complements to public pensions. It has two distinct characteristics that have important implications for tax revenues. First, unlike several other European countries—such as Belgium, Denmark, Finland, Sweden, and United Kingdom—that have private pension schemes with “defined contributions,” the Dutch pension system for the most part has a “defined benefit” structure. Pension systems set-up as “defined benefit” schemes are usually more susceptible to swings in asset holdings as all pensioners are in principle guaranteed a certain pension benefit independent of developments in asset holdings. Second, with the sole exception of Denmark, the Netherlands is the only country that allows income tax deduction of pension contributions. This means that changes in pension contributions that may follow changes in the funds’ asset holdings will have a direct impact on tax receipts.

69. There is empirical evidence that large fluctuations in the value of their assets have led Dutch pension funds to change pension premia with some frequency since the mid-1990s. The share of equities in the pension funds’ investment portfolios increased from 10 percent in 1990 to more than 40 percent in 2005. This higher share of equities led to large increases in pension assets during the late 1990s, while exerting significant downward pressure in the early part of the decade following the stock price declines. Separately, since “defined benefits” are linked to wage growth (albeit conditionally in a number of cases), in a period of rising incomes pension funds have raised contributions with an eye to keeping pension assets above higher pension liabilities, with similar consequences for tax revenues.

70. An analysis for the past thirty years shows a clear negative relationship between tax ratios and pension premia. Such a relationship has been most marked over the past decade, which may in turn reflect increases in coverage of private pension schemes (Table 6). Fluctuations in pension premia affect tax revenues primarily through the deductibility of pension contributions in personal and corporate income taxes. But indirect channels, through the impact of pension premia on households’ disposable income and private consumption and indirect taxes, can also play a role.

^{Table 6.}

Tax Ratios’ Correlation with Pension Premiums in the Netherlands

Source: Quarterly data from OECD Economic Outlook 78.

^{Table 6.}

Tax Ratios’ Correlation with Pension Premiums in the Netherlands

	1975–84	1985–94	1995–03	1975–03
Direct taxes and social contributions	-47.6	-23.9	-79.7	-42.7
- Corporate income tax	19.9	-38.0	-78.4	-15.4
- Personal income tax and social contributions	-52.5	-17.9	-45.5	-33.1
Total taxes	-40.1	-17.9	-80.7	-45.3

Source: Quarterly data from OECD Economic Outlook 78.

^{Table 6.}

Tax Ratios’ Correlation with Pension Premiums in the Netherlands

	1975–84	1985–94	1995–03	1975–03
Direct taxes and social contributions	-47.6	-23.9	-79.7	-42.7
- Corporate income tax	19.9	-38.0	-78.4	-15.4
- Personal income tax and social contributions	-52.5	-17.9	-45.5	-33.1
Total taxes	-40.1	-17.9	-80.7	-45.3

Source: Quarterly data from OECD Economic Outlook 78.

71. A formal econometric analysis shows that the higher volatility in pension premia is important when explaining the increased personal income tax volatility. Firstly, as noted above, the analysis suggests that the conditional mean of the personal income tax revenue can be modeled either as an autoregressive factor or as a lagged function of wage income (see Appendix for details). Secondly, the empirical analysis provides evidence that the conditional variance of personal income tax revenue around its conditional mean is not constant. Against this background, the analysis supports the hypothesis that the time-varying conditional variance is sufficiently explained by pension premia fluctuations. The positive parameter value for pension premia volatility in the conditional variance function of the personal income tax revenue implies that increasing pension premia have contributed to increased tax revenue volatility.

Other factors

72. Closer global integration and the enlargement of EU are likely to have had an impact on the location decisions of corporates, with implications for tax revenue volatility. The impetus to corporates changing locations to book profits in order to minimize tax burdens would have been provided by the increasing tax competition between countries. This factor is of course likely to have had an impact on corporates originating not just from the Netherlands, but also from the other EU countries. In addition, larger and more volatile transaction flows and asset price swings, following in part greater global integration, may amplify tax revenue fluctuations as wealth positions of households and private firms change more rapidly. This is particularly so since the Dutch economy relies heavily on trade—and to a larger extent than most other European countries (see Figure 6). The recent marked boom-and-bust cycle in the stock markets has also likely had an effect on tax revenues from stock options and capital gains.

73. Fluctuations in mortgage interest payment deductions for income taxes may also have affected tax revenue volatility. In the Netherlands, interest payments on mortgage loans are (for the most part) deductible, and, as a result, increased volatility in mortgage interest payments may affect the volatility of tax revenues. The volatility of foregone income tax revenue is affected by the volatility of mortgage debt (and relatedly house prices) and mortgage rates. Of course, increases in house prices are generally associated with decreasing mortgage interest rates, and vice versa. If these two effects do not cancel each other out, the net effect can lead to increased volatility in the income tax revenues relative to the tax base.

74. Changes in house prices also affect wealth, which ultimately can affect the volatility of indirect tax revenues. A CPB (2006) study shows that housing wealth in the Netherlands has been growing steadily since 1995, with the exception of the period between 2000 and 2003, where the growth rate of house prices declined slightly, and even turned slightly negative in 2003. The CPB study also finds that durables consumption is related to housing wealth (and asset wealth). They estimate the marginal propensity to consume out of housing wealth for durable goods to be 0.1 percent. Their findings also show an asymmetry: the impact of house price gains on durables consumption is not significant, while the impact of losses is significant and twice as large. They therefore conclude that households tend to cut down spending on durables when facing a drop in wealth more strongly then they step up spending when they experience a housing wealth gain. This asymmetric affect of housing wealth on consumption may have increased the volatility observed in indirect taxes.

75. Fiscal policy is another factor that may have affected the volatility of tax revenues. A change in tax regime—for instance a shift from direct taxation to more indirect taxation as occurred in 2001—can lead to a sudden change in the tax revenues as percent of GDP, followed by a period of adjustment with increased volatility. In addition, policy induced volatility in the tax system can also influence volatility of tax revenues as a share of GDP. One possible example of this was the past practice under the fiscal framework whereby the government used ex post deviations of tax revenues from a pre-establish nominal reference levels for discretionary tax changes. As these nominal deviations may have reflected deviations in economic growth from its projected path (i.e., not necessarily a change in the tax ratio), the discretionary tax changes may have increased the volatility of tax revenues as a share of GDP. However, it is difficult to measure the impact of that fiscal policy on tax revenue volatility, as this requires a data series for tax policy changes over the sample period, which currently is not readily available.

D. Concluding Remarks

76. The above analysis shows that tax revenue volatility has increased substantially over the last decade in the Netherlands. This appears particularly to be the case for corporate taxes, but volatility in personal income taxes and indirect taxes has also increased substantially. Increased volatility is generally not reflected in other EU countries, where on average volatility has been more or less constant, or if anything, there has been some decline.

77. There are several reasons why volatility could have increased in the Netherlands and the paper has explored these: the business cycle fluctuations appear to be somewhat greater in the recent periods, reflecting possibly both greater fluctuations in global activity and trade, as well as in asset prices; the tax deductibility of mortgage interest payments could also have imparted some effect; and the location decisions of corporates likely also played some role. The tax deductibility of pension premia also appears to have been a key factor, and the paper presented significant empirical evidence that tends to support this.

78. One policy implication relates to how higher tax volatility might affect the so-called signal value within the fiscal framework. The commendable Dutch fiscal framework includes a fiscal deficit signal value of 2½ percent of GDP, which triggers corrective measures if the fiscal deficit exceeds this value to avoid breaching the 3 percent Maastricht limit. This signal value was adopted at a time when budgetary uncertainties, including those relating to tax revenues, were somewhat less. The higher tax revenue volatility witnessed in recent years suggests that consideration could be given to a somewhat more conservative signal value.

APPENDIX I: Factors Determining Tax Revenue Volatility: Pension Contributions and Business Cycle

79. The objective of our empirical analysis is to determine whether fluctuations in pension premiums and the business cycle explain the increased volatility of personal income tax revenues in the Netherlands. ^⁴⁸ We use a basic linear auto-regressive conditional heteroskedastic (ARCH) framework introduced by Engle (1982) and extended by Bollerslev (1986), which have often been used in the financial literature, to estimate the volatility of the tax revenues and pension premiums. ^⁴⁹ This framework has the advantage that it allows us to simultaneously estimate the parameter values for the explanatory variables for the conditional variance. ^⁵⁰

Empirical framework

80. The general specification of an ARCH-framework is based on an observed time series y_t, which can be written as the sum of a predictable part and a stochastic part,

\begin{matrix} y_{t} = E [y_{t} | Ω_{t - 1}] + ɛ_{t}, & (1) \end{matrix}

with Ω_t-1 denoting the relevant information set available at time t-1. In this framework the conditional variance of the residuals, ε_t, is time varying: $E [ɛ_{t}^{2} | Ω_{t - 1}] = V_{t}$ . This feature of the ARCH-framework enables us to capture any structural element in the conditional variance of our observed variable, the personal income tax revenues. Engle (1982) formulated an ARCH-function for V_t to describe the conditional variance of y_t at time t as^⁵¹

\begin{matrix} V_{t} = β_{1} + β_{2} ɛ_{t - 1}^{2}, & (2) \end{matrix}

where, given the non-negativity condition imposed on V_t, the parameters have to satisfy the following conditions: β₁ > 0 and β₂ ≥ 0. If β₂ = 0, then the conditional variance is constant and the time series is homoskedastic.

81. We analyze the conditional variance of the personal income tax revenue in the Netherlands by assuming two functional forms for the conditional mean of the personal income tax revenue. We model the conditional mean as an autoregressive process of order one, and alternatively, we also model it as a function of wage income as percent of GDP. An important condition in this context that we have to test is whether the parameter value for the conditional mean of the personal income tax revenue is significantly different from one. A parameter value of one would simply imply that all changes in personal income tax revenue as a percent of GDP comes from a stochastic process. In this regard, we empirically test whether the conditional variance of the personal income tax revenue—that is the conditional variance of the stochastic process, as the conditional variance of mean is zero—is homoskedastic or not.

82. If the conditional variance is time-varying, we have to test whether the dynamics of the pension premiums and the business cycle play a significant role in explaining this variance. From a theoretical stand-point we would expect that the variability of the pension premiums to affect the conditional variance of personal income tax revenues, and not the level of it. In addition, we would expect this effect to be positive, i.e. an increase in the variability of the pension premiums would lead to higher volatility in the personal income tax revenues. Since the volatility of the pension premiums is not a measured time series, we model an autoregressive process for the pension premiums to capture its conditional variance. The conditional variance of the pension premiums is then incorporated in the ARCH-specification for the personal income tax revenues. This is an extension of the standard ARCH-specification, as it includes beside the lagged squared residual term another independent variable as an explanatory variable.

83. To empirically asses the role of business cycle in the dynamics of the personal income tax revenue in the Netherlands we model the conditional variance of the personal income tax revenue as a function of the output gap volatility as an explanatory variable. The output gap volatility series is calculated as the absolute deviation from its conditional mean.

84. As a result, our empirical specification for the personal income tax revenues, y_t, in an ARCH-framework, can be formulated as follows:

\begin{matrix} y_{t} = α_{1} + α_{2} y_{t - 1} + ɛ_{t} & (3) \end{matrix}

and, alternatively,

\begin{matrix} y_{t} = α_{1} + α_{2} w_{t} + ɛ_{t}, & (4) \end{matrix}

where the conditional variance are modeled as follows:

\begin{matrix} V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P}, & (5) \end{matrix}

\begin{matrix} V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{G}, & (6) \end{matrix}

and,

\begin{matrix} V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}, & (7) \end{matrix}

with w_t denoting wage income as a share of GDP, $V_{t}^{T}$ and $V_{t}^{P}$ representing the conditional variances of, respectively, the personal income tax revenues and pension premiums, and $V_{t}^{G}$ representing the output gap volatility.

Time series properties of the data

85. Our sample covers the period of 1970—2005 and includes quarterly data on GDP, output gap, personal income tax revenues, pension premiums, and wage income from the OECD and Statistics Netherlands. The descriptive and time series statistics for the variables used in this empirical analysis are presented in Table 1 and Figure 1. Both series, the log of personal income tax revenues (as percent of GDP) and the pension premiums (as percent of wages), show some structural events around 1980 and 1990. ^⁵² Time series analysis of the variables indicates a high degree of non-normality and non-stationarity in their levels. First differencing makes both series stationary, as shown by the ADF-statistics.

^{Table 1.}

Time Series Properties of Personal income tax Revenues, Pension Premiums, and Wage Income

^{^1/}The LM-test is asymptotically χ²(q) distributed.

^{Table 1.}

Time Series Properties of Personal income tax Revenues, Pension Premiums, and Wage Income

		Log of personal income tax revenues as percent of GDP	Log of pension premium (as percent of wages)	Log of wage income (as percent of GDP)
Normality tests
	Skewness	-0.46	-0.09	-0.09
	Kurtosis	1.69	2.14	1.99
	Jarque-Bera	15.31	4.34	6.23
Stationarity test: Augmented Dickey-Fuller (ADF)
Level:
	5% level	-2.88	-2.88	-2.88
	10% level	-2.58	-2.58	-2.58
	ADF-statistic	-1.36	-1.68	-1.38
First difference
	1% level	-2.48	-3.48	-3.48
	5% level	-2.88	-2.88	-2.88
	ADF-statistic	-6.05	-4.14	-17.10
ARCH LM-test
	χ²(1) ^1/	5.69	4.87	26.49

^{^1/}The LM-test is asymptotically χ²(q) distributed.

^{Table 1.}

Time Series Properties of Personal income tax Revenues, Pension Premiums, and Wage Income

		Log of personal income tax revenues as percent of GDP	Log of pension premium (as percent of wages)	Log of wage income (as percent of GDP)
Normality tests
	Skewness	-0.46	-0.09	-0.09
	Kurtosis	1.69	2.14	1.99
	Jarque-Bera	15.31	4.34	6.23
Stationarity test: Augmented Dickey-Fuller (ADF)
Level:
	5% level	-2.88	-2.88	-2.88
	10% level	-2.58	-2.58	-2.58
	ADF-statistic	-1.36	-1.68	-1.38
First difference
	1% level	-2.48	-3.48	-3.48
	5% level	-2.88	-2.88	-2.88
	ADF-statistic	-6.05	-4.14	-17.10
ARCH LM-test
	χ²(1) ^1/	5.69	4.87	26.49

^{^1/}The LM-test is asymptotically χ²(q) distributed.

86. We conduct a formal test for conditional heteroskedasticity in the context of ARCH models based on the Lagrange Multiplier (LM) principle for both series. For this purpose we assume that the conditional mean of the corresponding time series are adequately described by an autoregressive model, where the autoregressive order is determine by the Akaike criterion. The corresponding LM-test is obtained from a regression of the squared residuals as an autoregressive model, where the null hypothesis of conditional homoskedasticity is formulated as the parameter values for the autoregressive terms being zero. We also applied the LM-test for conditional homoskedasticity on the residuals from the regression of the personal income tax revenues on wage income as a share of GDP. The LM-test results for both regressions show that there is evidence for the presence of ARCH(1) in the series for personal income tax revenues. ^⁵³, ^⁵⁴ Note that the hypothesis of conditional homoskedasticity for pension premiums is not rejected.

Estimation results^⁵⁵

87. The formal estimation results support the hypothesis that the volatility of personal income tax revenue is affected by developments in pension premiums and the business cycle. Tables 2–7 provides the estimation results for the ARCH-specification for the personal income tax revenue as specified in Equations (3)—(6). The results show that the conditional mean of personal income tax revenues can be model as, both, an autoregressive process or as a function of wage income, with elasticities varying between 0. 9—1.0 and 1.0—1.10, respectively. This means that the mean of the personal income tax revenues can be projected based on its lagged value or on the projected wage income. The larger than unity tax revenue elasticity in the case of personal income tax revenue, reflects the progressive nature of the tax system in the Netherlands.

^{Table 2.}

ARCH-Specification Personal Income Tax Volatility: Estimation Results

^{Table 2.}

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.30	[-66.93]
y_t-1	α₂	0.93	[23.48]
Part II, Conditional variance of the ARCH-specification, Equation (5):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P}$
Constant	β₁	0.001	[7.58]
$ɛ_{t - 1}^{2}$	β₂	0.50	[3.24]
$V_{t}^{P}$	β₃	0.07	[6.45]
Adj. R²		0.91
ARCH-LM		4.58
Variable descriptions:
y_t- = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.

^{Table 2.}

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.30	[-66.93]
y_t-1	α₂	0.93	[23.48]
Part II, Conditional variance of the ARCH-specification, Equation (5):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P}$
Constant	β₁	0.001	[7.58]
$ɛ_{t - 1}^{2}$	β₂	0.50	[3.24]
$V_{t}^{P}$	β₃	0.07	[6.45]
Adj. R²		0.91
ARCH-LM		4.58
Variable descriptions:
y_t- = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.

Table 3.

ARCH-Specification Personal income tax Volatility: Estimation Results

Table 3.

ARCH-Specification Personal income tax Volatility: Estimation Results

Variables		Parameter values	z-Statistic
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.78	[-21.71]
w_t	α₂	1.02	[14.28]
Part II, Conditional variance of the ARCH-specification, Equation (5):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P}$
Constant	β₁	0.001	[7.93]
$ɛ_{t - 1}^{2}$	β₂	0.95	[15.69]
$V_{t}^{P}$	β₃	0.06	[12.04]
Adj. R²		0.16
ARCH-LM		4.35
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.

Table 3.

ARCH-Specification Personal income tax Volatility: Estimation Results

Variables		Parameter values	z-Statistic
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.78	[-21.71]
w_t	α₂	1.02	[14.28]
Part II, Conditional variance of the ARCH-specification, Equation (5):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P}$
Constant	β₁	0.001	[7.93]
$ɛ_{t - 1}^{2}$	β₂	0.95	[15.69]
$V_{t}^{P}$	β₃	0.06	[12.04]
Adj. R²		0.16
ARCH-LM		4.35
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.

Table 4.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Table 4.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.29	[-10.46]
y_t-1	α₂	0.99	[76.35]
Part II, Conditional variance of the ARCH-specification, Equation (6):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{G}$
Constant	β₁	0.01	[5.19]
$ɛ_{t - 1}^{2}$	β₂	0.60	[1.99]
$V_{t}^{G}$	β₃	0.04	[3.89]
Adj. R²		0.97
ARCH-LM		4.65
Variable descriptions:
y_t-1 = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{G}$ = the “volatility” of the output gap.

Table 4.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.29	[-10.46]
y_t-1	α₂	0.99	[76.35]
Part II, Conditional variance of the ARCH-specification, Equation (6):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{G}$
Constant	β₁	0.01	[5.19]
$ɛ_{t - 1}^{2}$	β₂	0.60	[1.99]
$V_{t}^{G}$	β₃	0.04	[3.89]
Adj. R²		0.97
ARCH-LM		4.65
Variable descriptions:
y_t-1 = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{G}$ = the “volatility” of the output gap.

Table 5.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Table 5.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.44	[-1.95]
w_t	α₂	1.10	[4.00]
Part II, Conditional variance of the ARCH-specification, Equation (6):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{G}$
Constant	β₁	0.02	[1.97]
$ɛ_{t - 1}^{2}$	β₂	0.74	[4.81]
$V_{t}^{G}$	β₃	0.08	[3.12]
Adj. R²		0.19
ARCH-LM		4.40
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{G}$ = the “volatility” of the output gap.

Table 5.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.44	[-1.95]
w_t	α₂	1.10	[4.00]
Part II, Conditional variance of the ARCH-specification, Equation (6):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{G}$
Constant	β₁	0.02	[1.97]
$ɛ_{t - 1}^{2}$	β₂	0.74	[4.81]
$V_{t}^{G}$	β₃	0.08	[3.12]
Adj. R²		0.19
ARCH-LM		4.40
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{G}$ = the “volatility” of the output gap.

Table 6.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Table 6.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.01	[-0.43]
y_t-1	α₂	0.99	[64.38]
Part II, Conditional variance of the ARCH-specification, Equation (7):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}$
Constant	β₁	0.01	[7.74]
$ɛ_{t - 1}^{2}$	β₂	0.39	[4.19]
$V_{t}^{P}$	β₃	0.07	[3.82]
$V_{t}^{G}$	β₄	0.05	[7.90]
Adj. R²		0.97
ARCH-LM		4.62
Variable descriptions:
y_t-1 = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.
$V_{t}^{G}$ = the “volatility” of the output gap.

Table 6.

ARCH-Specification Personal Income Tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (3):
y_t = α₁ + α₂y_t-1 + ε_t
Constant	α₁	-1.01	[-0.43]
y_t-1	α₂	0.99	[64.38]
Part II, Conditional variance of the ARCH-specification, Equation (7):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}$
Constant	β₁	0.01	[7.74]
$ɛ_{t - 1}^{2}$	β₂	0.39	[4.19]
$V_{t}^{P}$	β₃	0.07	[3.82]
$V_{t}^{G}$	β₄	0.05	[7.90]
Adj. R²		0.97
ARCH-LM		4.62
Variable descriptions:
y_t-1 = personal income tax revenue as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.
$V_{t}^{G}$ = the “volatility” of the output gap.

^{Table 7.}

ARCH-Specification Personal income tax Volatility: Estimation Results

^{Table 7.}

ARCH-Specification Personal income tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.89	[-21.71]
w_t	α₂	0.73	[3.43]
Part II, Conditional variance of the ARCH-specification, Equation (7):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}$
Constant	β₁	0.01	[9.16]
$ɛ_{t - 1}^{2}$	β₂	0.22	[4.53]
$V_{t}^{P}$	β₃	0.06	[10.66]
$V_{t}^{G}$	β₄	0.04	[1.88]
Adj. R²		0.16
ARCH-LM		4.37
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.
$V_{t}^{G}$ = the “volatility” of the output gap.

^{Table 7.}

ARCH-Specification Personal income tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.89	[-21.71]
w_t	α₂	0.73	[3.43]
Part II, Conditional variance of the ARCH-specification, Equation (7):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}$
Constant	β₁	0.01	[9.16]
$ɛ_{t - 1}^{2}$	β₂	0.22	[4.53]
$V_{t}^{P}$	β₃	0.06	[10.66]
$V_{t}^{G}$	β₄	0.04	[1.88]
Adj. R²		0.16
ARCH-LM		4.37
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.
$V_{t}^{G}$ = the “volatility” of the output gap.

88. The second part of our estimation result for both specifications show that pension premiums, business cycle, and a combination of these two play a significant role in determining the fluctuations in personal income tax revenues around its conditional mean. The parametric conditions for the ARCH-specification are satisfied, as all the estimated coefficients have the correct signs. In addition, a positive sign also supports our hypothesis that an increase in pension premiums or business cycle increases the volatility of personal income tax revenues. The estimate for the constant term is significantly different from zero. The parameter estimate for the lagged residual term, the pension premium volatility, and the volatility of the output gap are significantly different from zero. The semi-elasticities for the pension premium volatility imply that a one percent in crease in premium volatility leads to about 6 to 8 percent increase in personal income tax revenue volatility. The semi-elasticities for the output gap have a wider range of effect: the effect varies between 5 percent and 8 percent, depending on the functional form of the conditional mean and the conditional variance. To confirm our result, we regressed the squared residuals from both models for the conditional expectation of personal income tax revenue directly on the conditional variance of the pension contribution rates and the output gap volatility. With a t-value of 2.2 and 3.2, respectively, the results indicates that pension contributions and business cycle play a significant role in explaining the increased personal income tax revenue volatility.

89. The Akaike criterion discriminates in favor of the autoregressive model for the conditional mean and that the conditional variance is better explained by a combination of pension contribution and output gap volatility.

References

Berben, R.P., K. Bernoth, and M. Mastrogiacomo, 2006, “Households’ Response to Wealth Changes; Do Gains or Losses Make a Difference?” CPB Discussion Paper No. 63.
- Search Google Scholar
- Export Citation
Bollerslev, T., 1986, “Generalized Autoregressive Conditional Heteroskedasticity,” Journal of Econometrics No. 31, pp. 307–27.
- Search Google Scholar
- Export Citation
Chan, K., G. Karolyi, and A. Sanders, 1992, “An Empirical Investigation of Alternative Models of the Short Term Interest Rate,” The Journal of Finance No. 1, pp 345–61.
- Search Google Scholar
- Export Citation
Engle, Robert F., 1982, “Autoregressive Conditional Heteroskedasticity with Estimates of the Variance of U.K. Inflation,” Econometrica No. 50, pp. 987–1008.
- Search Google Scholar
- Export Citation
Franses, P.H. and D. van Dijk, 1999, “Non-linear Time Series Models in Empirical Finance,” Econometric Institute, Erasmus University, Rotterdam, (Cambridge University Press).
- Search Google Scholar
- Export Citation
Hill, Elisabeth G., 2005, Revenue Volatility in California, LAO Report, Sacramento.
- Search Google Scholar
- Export Citation
Kingdom of the Netherlands—Netherlands: Report on the Observance of Standard and Codes—Fiscal Transparency Module, and the Aide-Mémoire Regarding the Fiscal Framework, IMF Country Report No. 06/124, 2006.
- Search Google Scholar
- Export Citation
Koedijk, K., F. Nissen, P. Schotman, and C. Wolff, 1996, “The Dynamics of Short-Term Interest Rate Volatility Reconsidered,” European Finance Review 1, pp. 105–30.
- Search Google Scholar
- Export Citation
Longstaff, F.A., and E.S. Schwartz, 1992, “Interest Rate Volatility and the Term Structure: A Two-Factor General Equilibrium Model,” The Journal of Finance No. 4, pp. 1259–82.
- Search Google Scholar
- Export Citation
Purfield, Catriona, 2005, “Managing Revenue Volatility in a Small Island Economy: The Case of Kiribati,” IMF Working Paper No. 05/154, (Washington, DC: International Monetary Fund).
- Search Google Scholar
- Export Citation
Talvi, Ernesto and Carlo A. Vegh, 2000, “Tax Base Variability and Procyclical Fiscal Policy,” NBER Working Paper 7499, Cambridge [4] Van Ewijk, Casper and Martijn van de Ven, 2003, Pension Funds at Risk, CPB Report 2003/1, The Hague.
- Search Google Scholar
- Export Citation
Zee, Howell H., 1996, “Empirics of Cross-country Tax Revenue Comparisons,” World Development Vol. 24, No. 10, pp. 1659–71.
- Search Google Scholar
- Export Citation

Prepared by Wendly Daal, Manmohan S. Kumar, and Michael Skaarup. We are grateful to Robert A. Feldman for comments and suggestions.

Kingdom of the Netherlands—Netherlands: Report on the Observance of Standard and Codes—Fiscal Transparency Module, and the Aide-Mémoire Regarding the Fiscal Framework, IMF Country Report No. 06/124, 2006.

Studies on tax revenue and tax base volatility have usually focused on developing countries, see for instance Purfield (2005); Talvi and Vegh (2000); and Zee (1998).

Note that since the mid-1990s, the ratio of total tax revenues to GDP has declined, reflecting a reduction in personal income tax and social contributions, even while indirect taxes increased relative to GDP.

Including social contributions.

The latter measure seemed to be of particular interest as a determinant of the volatility in tax revenues since it captures the fluctuations in output relative to the underlying potential, rather than simply changes in GDP per se.

OECD estimates the personal income tax elasticity and social contributions to 1.7 and 0.6, respectively. A weighted average of those equals 1.25.

The implications of the above for the overall balance are worth noting. The OECD estimates the overall budget elasticity at around 0.6 percent for the Netherlands, implying that a 1 percentage point widening of the output gap changes the fiscal balance (as a share of GDP) by 0.6 percentage point of GDP. Over the past 30 years, the output gap has varied within an interval of around ±4 percent of GDP and has changed year-onyear by 1¼ percent, on average, in absolute terms (see Figure 5). On this basis, the fiscal balance can be expected to change by ¾ percentage points of GDP from year to year and vary within an interval of around ±2½ percent of GDP due to business cycle fluctuations.

Personal income tax revenue, hence forth, is for the sake of simplicity defined as personal income tax revenue and social contributions.

Longstaff and Schwartz (1992), Ball and Torous (1999), Chan and others (1992), and Koedijk and others (1997).

Several methodologies have been used in literature to obtain estimates of volatility, such as the “historical volatility” approach. This approach uses historical data to calculate volatility data. The disadvantage of this method is that it does not provide current information on the volatility.

This specification can be extended in line with Bollerslev (1986) to a generalized linear autoregressive conditional heteroskedastic (GARCH) framework by including the lagged conditional variance, V^t..

Further study is required to determine what is driving these turnarounds.

We also carried out a sign-and-size-biased test to check whether positive and negative shocks have an asymmetric impact on the conditional variance. The test showed no asymmetric effect.

Based on a formal test for outliers (OL) OL’s were detected in 1995 and 2001 reflecting tax-regime changes. An outlier robust estimation technique, the Generalized M (GM) estimation technique, in the context of a simple AR(1) model was used, because, as noted by Franses and van Dijk (1999), if the OL’s are neglected, the LM test rejects the null hypothesis of condition homoskedasticy too often when it is in fact true.

We used the Berndt-Hall-Hall-Hausman (BHHH, 1974) numerical algorithm to find the quasi maximum likelihood parameter estimates for the ARCH-specification. To ensure that the estimates of the volatility are robust and that the algorithm converges to the global maximum we used different starting values. In addition, we used Bollerslev-Wooldrige’s procedure to obtain robust standard errors and covariance.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/002/2006/284/article-A003-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Kingdom of the Netherlands—Netherlands: Selected Issues

Author: International Monetary Fund

Volume/Issue:: Volume 2006: Issue 284

Publisher:: International Monetary Fund

ISBN:: 9781451829556

ISSN:: 1934-7685

Pages:: 81

DOI:: https://doi.org/10.5089/9781451829556.002

Keywords
III. Volatility of Tax Revenues in the Netherlands40
APPENDIX I: Factors Determining Tax Revenue Volatility: Pension Contributions and Business Cycle
References

View raw image

Netherlands: Conditional Volatility of Personal Income Tax (PIT) Revenues, 1970-2005 ^1/ ^2/

View raw image

Coefficient of Variation of Personal Income Tax and Social Contributions

(In percent of GDP)

View raw image

Coefficient of Variation of Corporate Income Tax

(In percent of GDP)

View raw image

Coefficient of Variation of Indirect Taxes

(In percent of GDP)

View raw image

Output Gap in the Netherlands

View raw image

Trade Openness of EU Economies

(Exports and imports relative to GDP, In percent)

View raw image

Netherlands: Time Series of Selected Variables

^{Table 7.}

ARCH-Specification Personal income tax Volatility: Estimation Results

Variables		Parameter values	z-Statistics
Part I, Conditional mean of the ARCH-specification, Equation (4):
y_t = α₁ + α₂w_t + ε_t
Constant	α₁	-0.89	[-21.71]
w_t	α₂	0.73	[3.43]
Part II, Conditional variance of the ARCH-specification, Equation (7):
$V_{t}^{T} = β_{1} + β_{2} ɛ_{t - 1}^{2} + β_{3} V_{t - 1}^{P} + β_{4} V_{t - 1}^{G}$
Constant	β₁	0.01	[9.16]
$ɛ_{t - 1}^{2}$	β₂	0.22	[4.53]
$V_{t}^{P}$	β₃	0.06	[10.66]
$V_{t}^{G}$	β₄	0.04	[1.88]
Adj. R²		0.16
ARCH-LM		4.37
Variable descriptions:
w_t-1 = wage income as percent of GDP.
$ɛ_{t - 1}^{2}$ = the lagged squared residuals of Equation (5).
$V_{t}^{P}$ = the conditional variance of the pension premiums.
$V_{t}^{G}$ = the “volatility” of the output gap.

Topics

Countries

Series

About

Resources

Abstract

III. Volatility of Tax Revenues in the Netherlands40

A. Introduction

B. Tax Revenue Volatility in the Netherlands and in other EU-15 Countries

Experience of the Netherlands

Tax revenue volatility in a European context

C. Determinants of Tax Revenue Volatility in the Netherlands

The role of the business cycle

The role of pension funds

Other factors

D. Concluding Remarks

APPENDIX I: Factors Determining Tax Revenue Volatility: Pension Contributions and Business Cycle

Empirical framework

Time series properties of the data

Estimation results55

References

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics

III. Volatility of Tax Revenues in the Netherlands^⁴⁰

Estimation results^⁵⁵