40 3 Reliability of the data The data-set on tax legislation, which we want to employ here, is based on ex-ante projections. How reliable are these data? Generally two questions are especially important. First, we cannot know whether the expected fiscal effects reflect “planning or propaganda”51. It does not seem implausible that political attempts to delude voters play an important role in calculating the expected fiscal effects of tax reforms. Studies on fiscal planning (Heinemann 2006) or on tax revenue forecasts (Bischoff/Gohout 2006) indicate that such a bias in German government projections cannot per se be excluded. Second, even if the calculations of expected fiscal effects reflect unbiased predictions, it remains to be seen how accurate the fit in between expected and real effects is. -12% -10% -8% -6% -4% -2% 0% 2% 4% 6% 8% 10% 19 65 19 67 19 69 19 71 19 73 19 75 19 77 19 79 19 81 19 83 19 85 19 87 19 89 19 91 19 93 19 95 19 97 19 99 20 01 20 03 20 05 TAX REFORMS (IMPLEMENTED)/TAX REVENUES TAX REFORMS (ADOPTED)/TAX REVENUES ADOPTION AND IMPLEMENTATION OF TAX REFORMS Fiscal effects tax reforms/ Total tax revenues Own calculations based on: Federal Ministry of Finance (2004)/Federal Statistical Office/tax laws. Figure 14: Fiscal effects of tax reforms (adopted/implemented) How should we test the reliability of the data? If the projections of fiscal effects of tax reforms are largely correct, we should be able to relate the reform effects to revenue developments. Together with macroeconomic developments, which directly 51 To use the title of an article which analyzed government budget planning and finds evidence for strong biases in fiscal projections (Heinemann 2006). 41 affect tax revenues, the fiscal effects of tax reforms should increase our ability to explain tax revenue developments. However, one major problem arises. Although our data display the fiscal effects of reforms in the first twelve months from being effective, the data-base includes for most reforms only the dates when the regulations were adopted and not when they were implemented and became effective. To test (based on revenue developments) whether the data are reliable makes therefore a transformation of the data-set necessary. We therefore compiled and added the dates when the new regulations became effective.52 We display the timing of fiscal effects of tax reforms by the date of adoption and by the date of implementation in Figure 14.53 We find that there is no general pattern which would allow us to transform adopted reform effects in effective reform effects just by using a time lag of e.g. one period. Instead we see that there were times when adopted and implemented effects moved closely together and times when the development diverged strongly. Now we dispose of the relevant data to test the reliability of our data-set. Additionally to tax reforms the business cycle plays an important role for tax revenue developments. Most important is the influence of nominal GDP growth. In the longrun the average tax revenue over nominal GDP elasticity is close to one.54 After transforming our data-set, we can start with the empirical analysis. If the fiscal effects are reliable and – additionally to GDP growth – a second important variable to explain tax revenue developments, we should see a correlation of high tax revenue over GDP elasticities with tax burden increases via tax reforms and of low tax revenue over GDP elasticities with tax burden reductions via tax reforms.55 52 Based on the “Bundessteuerblatt” and tax laws. 53 We divided the effects of reforms which were not implemented at the beginning of a year exactly in between different years depending on the date of implementation (e.g. the fiscal effect of a regulation, which becomes effective on April 1st, is assigned with 75% of the fiscal effect to the same and 25% to the following year). 54 A statistical analysis shows that nominal GDP growth (without a time lag) is highly significant to explain tax revenue growth. Variations in nominal growth alone can explain 56% of the annual change in the growth of overall tax revenues from 1964 to 2004 (see first column of Table 3 on page 46). Nominal GDP growth is as well the most important macroeconomic variable used for the official annual tax revenue projections. We have tested several other variables as inflation and real GDP growth and changes in total employment as well as combinations of these and included several time-lags. But we derived the best results with nominal GDP growth without a time-lag. For a discussion of theory and empirics of tax revenues projections in Germany see Gebhardt (2001), Boss (1997), Boss/Elender (2003) and von der Lippe (1998). 55 See for the general approach of combining tax elasticities and reform effects Gebhardt (2001, pp. 131 ff.). 42 TAX REFORMS AND TAX REVENUE DEVELOPMENTS 1965-2004 -12% -7% -2% 3% 8% 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 F IS C A L E FF E C T S T A X R E FO R M S /T A X R E V E N U E S -2 -1,5 -1 -0,5 0 0,5 1 1,5 2 2,5 3 3,5 G D P E LA S T IC IT Y O F T A X R E V E N U E S FISCAL EFFECTS TAX REFORMS (IMPLEMENTED)/TAX REVENUES GDP ELASTICITY OF TAX REVENUES Own calculations based on: Federal Ministry of Finance (2004)/Federal Statistical Office/tax laws. Figure 15: GDP elasticity of tax revenues and fiscal effects of tax reforms The data are displayed in Figure 15. The scale on the left hand side and the data columns display the net fiscal effects (increases minus reductions) of tax reforms over total tax revenues for the years 1964 till 2004 by the date of implementation. The right hand scale and the line show the GDP elasticity of tax revenues. The xaxis displays the different years and intersects at 0 on the left hand scale (= no net fiscal effect of tax reforms) and at 1 on the right hand scale (= the long-term average GDP elasticity of tax revenues). If we take a closer look at the development of the two series, we find the expected relationship in between the fiscal effects of tax reforms and the development of the GDP elasticity of tax revenues in a majority of years (although the relationship is far from perfect). This can be seen as evidence speaking in favor of the reliability of our data. A second finding is that the strongest differences are observable in the early 1970s. Especially in 1973 and 1975 the expected fiscal effects of tax reforms were far larger than the GDP elasticity of tax revenues pointing at unrealistic expectations of the Federal Ministry of Finance especially in those years. If we take a closer look at tax policy in these years, we find that attempts of the federal governments to influence the business cycle via tax policy based on Keynesian macroeconomics were decisive for these differences. 43 Excursus: Attempts to stabilize the business cycle via temporary changes in taxation Inspired by Keynesian macroeconomics, the German government employed tax policy measures to smoothen the business cycle in the 1960s and the early 1970s. In 1961 temporary special depreciation schemes were introduced that could be implemented if the macro economy is in (or is expected to get into) disequilibrium with negative effects on growth and employment.56 Furthermore, the so-called stabilization law of June 8th, 1967 created the possibility to introduce surcharges and reductions of 10% on the income and corporate tax liabilities.57 Business investments could be stimulated by an investment bonus of up to 7.5% of investment costs58 or depressed by temporary abolishment of special and favorable depreciation schemes.59 These instruments to smoothen the business cycle by tax policy were applied especially from 1967 to 1975. In 1970/71 and 1973/74 the government tried to cool down the economy (and to contain inflation) by raising the tax burden while in 1967 and 1974/75 the economy should be stimulated by an investment bonus. In 1967 the stimulation took place via a special depreciation scheme of 10% for mobile and 5% for immobile investments in machinery and equipment.60 From August 1st, 1970 to June 30th, 1971 a 10% surcharge on income and corporate tax payments was raised to cool down the economy. This surcharge was reimbursed (without interest) after June 15th, 1972.61 Furthermore, the demand for investment goods should be reduced by a temporary abolishment of favorable depreciation schemes for investments in between July 5th, 1970 and February 1st, 1971.62 The last application of such a surcharge on income and corporate taxes liabilities took place from July 1st, 1973 to June 30th, 197463 and was not refunded this time. The most important measure aimed at economic stimulation was the introduction of an investment bonus of 7.5% for all investment in machinery and equipment in between December 1st, 1974 and June 30th, 1975.64 Our review of the development of tax reforms and tax revenues (see Figure 15) showed that the effects of these temporary measures were – especially in 1973 and 1975 – largely overestimated. Additionally, the application of these measures was extraordinary and was not repeated in later years. 56 See BGBl I (1961), pp. 981 ff. 57 See BGBl I (1967), pp. 582 ff. 58 Substractable from income and corporate tax payments. 59 For a more detailed discussion see Koch (1976), pp. 42 ff. or Muscheid (1986), pp. 128 ff. 60 Applicable from January until October 1967. 61 Low incomes (below DM100 (€51) per month) were excluded. See BGBl I (1970), pp. 1125 ff. 62 See BGBl I (1970), pp. 1128 ff. 63 Applicable only for annual incomes above DM24,000 (€ 12,300). See BGBl I (1973), pp. 530 ff. 64 See BGBl I (1974), pp. 3676 ff. 44 Therefore, we separate the effects of these temporary measures in our data-set. This gives us the possibility to analyze tax reforms either including or excluding these temporary measures. In our following analyses we discuss on a case by case basis whether to analyze the data including or excluding these temporary measures. END OF EXCURSUS TAX REFORMS AND TAX REVENUE DEVELOPMENTS 1965-2004 (ADJUSTED) -12% -7% -2% 3% 8% 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 FI S C A L E F FE C T S T A X R E F O R M S /T A X R E V E N U E S -2 -1,5 -1 -0,5 0 0,5 1 1,5 2 2,5 3 3,5 G D P E LA S TI C IT IY O F T A X R E V E N U E S FISCAL EFFECTS TAX REFORMS (IMPLEMENTED)/TAX REVENUES GDP ELASTICITY OF TAX REVENUES ADJUSTED FOR TEMPORARY MEASURES (1967 to 1975) Own calculations based on: Federal Ministry of Finance (2004)/Federal Statistical Office/tax laws. Figure 16: GDP elasticity of tax revenues and fiscal effects of tax reforms (adjusted for temporary measures) Figure 16 shows the fiscal effects of tax reforms adjusted for all temporary measures directly aimed at a manipulation of the business cycle (taking place in between 1967 and 1975). The adjusted data shows a better fit of expected fiscal effects of tax reforms and the GDP elasticity of taxes in between 1967 and 1975. We see that especially the strong increase of fiscal effects of tax reforms in 1973 and the decrease in 1975 are strongly moderated by the adjustment. The remaining decrease in 1975 was largely due to a tax relief by the first important income tax reform (with a tarif change) since 1965.65 The surcharge on the income and wage tax as well as an in- 65 Compare the discussion of the reform pattern in the wage and income tax in part IV.2.1. 45 vestment tax in 1973 and the investment subsidy from December 1974 on (effecting tax revenues especially in 1975) were responsible for the strong divergence of expected and realized growth of tax revenues in the unadjusted data. Obviously the real effects were far smaller than the expected effects, which contributed to the decision to abstain from a smoothing of the business cycles via temporary tax surcharges in later years. How should we test the reliability of the expected fiscal effects of tax reforms based on tax revenue developments? Based on the discussed long-run GDP elasticity of tax revenues of one, we expect a linear relationship in between nominal GDP growth and the growth rate of tax revenues (our dependent variable) with a coefficient of one. To integrate the expected fiscal effects of tax reforms (increases minus reductions) as another independent variable in this relationship we divide the expected fiscal effects of tax reforms by total tax revenues. For this second independent variable we expect a coefficient of one in case of a perfect fit of the expected fiscal effects with observed revenue developments. We restrict our analysis here to the time period from 1965 to 2003 as only these years are completely covered in our data-set. We test the effect of different specifications of the expected fiscal effects of tax reforms (adopted as well as implemented; adjusted and unadjusted for temporary measures) in the following equations: GDPln ln tt10 ??? +?+=? tREV GDPln ln tt2t10 ???? +?+?+=? TREVt With: 0? , x? Constant, coefficients ?ln REVt Tax revenue growth rate ?ln GDPt Nominal GDP growth rate ?Tt Fiscal effects of tax reforms (adopted or implemented; adjusted and unadjusted)/Total tax revenues t? Stochastic error term In the regression analyses (see Table 3), nominal GDP growth turns out to be an important determinant of tax revenue growth. As expected based on the long-run GDP elasticity, the coefficient is in all specifications close to one and highly significant. GDP growth alone can explain 56% of the variation in tax revenue growth from 1965 to 2003. Additionally we find, that the fiscal effects of reforms by the date of adoption (column B and C) can contribute only slightly to the explanation of tax revenue developments. The coefficients are only significant on a 5% level (10% level for fiscal effects adjusted for temporary measures) and the explanatory impact of the 46 inclusion of adopted reform effects is low.66 Fiscal effects of reforms by the date of implementation (column D and E) perform far better. Our analysis shows that implemented fiscal effects of tax reforms are highly significant (at the 1% level) for explaining the tax revenue developments and the coefficients show the expected signs (net tax burden increases lead to increases in tax revenues and net tax burden reductions lead to reductions in tax revenues). In case of a perfect fit we would expect the coefficients of the fiscal effects of reforms over total tax revenues to equal one, but we find a size of 0.62 or 0.7267 (see column D and E) to be already very satisfying. Furthermore, we find that fiscal effects of implemented tax reforms adjusted for temporary measures perform slightly better than unadjusted fiscal effects. Together with nominal GDP growth the fiscal effects of implemented tax reforms can explain 65 to 66% of the variation in tax revenue growth from 1965 to 2003. Constant Nominal GDP growth ratet R2 (adjusted) Expl. Var: DW Notes: t-statistics of the estimated parameters in parentheses. * significant at the 10% level; ** significant at the 5% level; *** significant at the 1% level Observations Tax revenue developments and fiscal effects of tax reforms in Germany 1965-2003 Fiscal effects tax reforms/revenues (implemented) t Fiscal effects tax reforms/revenues (implemented) t Adjusted for temporary measures Fiscal effects tax reforms/revenues (adopted) t Fiscal effects tax reforms/revenues (adopted) t Adjusted for temporary measures Total Tax revenues (Growth rate) t (OLS) A (OLS) B Total Tax revenues (Growth rate) t (OLS) C Total Tax revenues (Growth rate) t (OLS) D Total Tax revenues (Growth rate) t (OLS) E Total Tax revenues (Growth rate) t 0.60 2.48 0.005 (0.484) 0.952*** (7.012) 0.538** (2.289) 39 0.58 2.47 0.004 (0.388) 0.962*** (6.890) 0.490* (1.735) 39 0.66 2.70 0.621*** (3.506) 0.005 (0.546) 0.956*** (7.672) 39 0.65 2.65 0.725*** (3.206) 0.005 (0.512) 0.968*** (7.631) 39 1.005*** (7.135) -0.001 (-0.099) 0.56 2.50 39 Table 3: Testing the reliability of the expected fiscal effects of tax reforms68 66 This explains as well why studies based on adopted reforms find only a very limited reliability of the data. See e.g. Koester (2006). 67 Meaning that the data shows a revenue increase of 0.72% when the expected fiscal effects of reforms equals a 1% increase. 68 The Durbin-Watson test statistics for our regressions (reported in table 3) do not support the hypothesis of serial autocorrelation in the error terms. We found no indication for perfect multicollinearity of the independent variables and the residuals show no signs for heteroske- 47 Based on these findings we conclude that the expected fiscal effects of tax reforms included in our data-set are a very reliable indicator of tax policy and well-suited for a quantitative analysis of tax policy in the following chapters. 4 Summary The empirical literature on tax policy and tax reforms draws on a large variety of data. However, there are barely any quantitative studies based on first-hand data on tax reforms. The data published by the German Federal Ministry of Finance on tax reforms from 1964-2004 (including expected fiscal effects of reforms) are highly promising to enable a detailed quantitative study of tax policy in Germany. The data-set includes 1,045 new tax policy regulations within 223 tax reforms, their expected fiscal effects, and the date of adoption of the new regulations. We transform the data-set especially by integrating the implementation dates of all new regulations. This does not only increase our ability to test economic theories of tax policy based on the data-set (see part V) but allows us as well to test the reliability of the expected fiscal effects of tax reforms. Based on our statistical analysis we find that the expected fiscal effects are a reliable indicator for a quantitative tax policy analysis. While some first analyses of the basic data published by the Federal Ministry of Finance exist (see e.g. Corneo 2005), our data-set is so far unique69 and offers a variety of possibilities for empirical testing. To exploit the possibilities of our dataset in an empirical analysis, we start by establishing new insights on tax policy in Germany in part IV.1. Then we discuss the most important tax reforms in the main German taxes and show how they are reflected in our data-set (part IV.2). In part V we derive and test a variety of normative and positive theories of tax policy based on our data-set. dasticity. Unit-root tests show that the variables employed here (growth rate of nominal GDP, growth rate of tax revenues, fiscal effects of reforms over total tax revenues) can be treated as stationary. 69 Based on the inclusion of the implementation dates for all new regulations.