Online Appendix to Do Tax Changes Affect Credit Markets and Financial Frictions? Evidence from Credit Spreads Beatrice Kraus and Christoph Winter August 17, 2016 Contents D: Alternative Definition of Unanticipated Tax Changes 1 E: Impact of Tax Changes on Other Credit Spreads 4 F: Controlling for Personal and Corporate Income Tax Changes at the Same Time 4 References 9 D: Alternative Definition of Unanticipated Tax Changes In this Section, we argue that our results are robust to using an alternative definition of unanticipated tax changes, which is due to Mertens and Ravn (2012). Mertens and Ravn (2012) classify those tax acts as unanticipated for which the gap between announcement and implementation date is at most 3 months. Relative to our classification, according to which the implementation gap is at most 90 days (see Section 2 of the main paper), the approach by Mertens and Ravn (2012) results in a higher number of unanticipated tax changes. In the following, we display the impulse response functions for our baseline VAR (in Figure 24) and for our baseline VAR augmented with EBP/FBP (in Figure 25) for a one percent increase in unanticipated taxes relative to nominal GDP, where unanticipated tax changes are constructed using the Mertens and Ravn (2012) definition. This Online Appendix is available at http://www.econ.uzh.ch/faculty/winter.html 1
Figure 24: Impulse responses of endogenous variables in baseline VAR Notes: Impulse responses of endogenous variables in baseline VAR (see Specification 1 in Section 3 of the main paper), consisting of tax revenues/gdp, GDP, non-durables consumption, durables purchases, private investment, and hours worked, to a one percent increase unanticipated taxes relative to nominal GDP with 95 percent confidence intervals. Unanticipated tax changes are constructed using the Mertens and Ravn (2012) definition. For detailed variable definitions, please refer to Table 3 of the main paper. Narrative shocks are corrected for measurement error, as described in Appendix C. 2
Figure 25: Impulse responses of endogenous variables in baseline VAR Notes: Impulse responses of the excess bond premium of non-financial firms (EBP, left panel) and of the excess bond premium of financial firms (FBP, right panel) to a one percent increase unanticipated taxes relative to nominal GDP with 95 percent confidence intervals. Unanticipated tax changes are constructed using the Mertens and Ravn (2012) definition. Results are obtained by adding either EBP or FBP to the baseline VAR, which is Specification 1 in Section 3 of the main paper. For detailed variable definitions, please refer to Table 3 of the main paper. Narrative shocks are corrected for measurement error, as described in Appendix C. 3
E: Impact of Tax Changes on Other Credit Spreads The results shown in this Section complement Section 4 of the paper. Figure 26: Impulse responses of other credit spreads Notes: Impulse responses of various credit spreads (please refer to Table 3 of the main paper for a detailed description of their definition) to a one percent increase in unanticipated taxes relative to nominal GDP with 95 percent confidence intervals. The responses are computed by adding one of the credit spread measures to the baseline VAR, consisting of tax revenues/gdp, GDP, non-durables consumption, durables purchases, private investment and hours worked, see Specification 1 in Section 3 of the main paper. Narrative shocks are corrected for measurement error, as described in Appendix C. F: Controlling for Personal and Corporate Income Tax Changes at the Same Time We now check whether our previous results are robust to controlling for both personal and corporate income tax changes. We therefore re-estimate Specification 2, but this time, we use both tax P I and tax CI as exogenous variables: y t = p A i y t i + i=1 s j=0 Bj P I tax P t j I + s j=0 Bj CI tax CI t j + D + Et + u t, (7) We incorporate the same endogenous variables as in Specification 3. In addition, we also include one of our variables of interest, i.e. either EBP, FBP, the return on assets of commercial banks, the loan loss provisions/loan ratio, or the fraction of non-performing C&I loans. We consider p = 3 lags for the endogenous variables, and we set the number of lags for the exogenous variables to s = 2, in order to allow for a lagged impact of tax liability changes on the endogenous variables. D and E again control for a constant and a linear time trend. It should be noted that in Specification 7, identification hinges on those tax acts for which personal and corporate income tax changes have a different sign. Since data on non-performing C&I loans are only available from 1984Q1 onwards, there are four tax acts that affect corporate income taxes, namely the TRA of 1986, the OBRA of 1987, the OBRA of 1990, and finally the JGTRRA of 2003. As shown 4
in Figure 5, the 1986 TRA is the only tax act for which personal income taxes and corporate income taxes are adjusted in opposite directions. Figure 27: Impulse responses of EBP and FBP Notes: Impulse responses of excess bond premia for non-financial (left panel) and financial (right panel) firms to a one percent increase in corporate income taxes with 95 percent confidence intervals. given by Specification 2, see Section 5. We include both tax P t I measurement error, as described in Appendix C. The underlying VAR is and tax CI t. Narrative shocks are corrected for Figure 28: Impulse responses of the return on assets of commercial banks Notes: Impulse responses of the pre-tax return on assets of commercial banks (left panel) and the after-tax return on assets of commercial banks (right panel) to a one percent increase in corporate income taxes with 95 percent confidence intervals. The underlying VAR is given by Specification 2, see Section 5. We include both tax P t I tax CI t. Narrative shocks are corrected for measurement error, as described in Appendix C. and Strikingly, we find that our previous results reported in Figures 12-16 regarding a one percent increase in ACITR remain unchanged, even if we control for personal income tax changes, see Figures 27-31. The only exception is the response of EBP, which becomes insignificant after incorporating tax P I. The reason is that the response of EBP is entirely driven by the 2003 JGTRRA. Recall from our previous results that the response of EBP becomes insignificant as soon as we drop the personal income 5
Figure 29: Impulse response of the excess stock market return Notes: Impulse response of the excess stock market return to a one percent increase in corporate income taxes with 95 percent confidence intervals. The underlying VAR is given by Specification 2, see Section 5. We include both tax P I t and tax CI t. Narrative shocks are corrected for measurement error, as described in Appendix C. Figure 30: Impulse response of loan loss provisions/loans Notes: Impulse response loan loss provisions/loans to a one percent increase in corporate income taxes with 95 percent confidence intervals. The underlying VAR is given by Specification 2, see Section 5. We include both tax P I t and tax CI t. Narrative shocks are corrected for measurement error, as described in Appendix C. tax component associated with the 2003 JGTRRA from our sample of personal income tax changes. Since the 2003 JGTRRA changes personal income taxes and corporate income taxes in the same direction, we cannot separate the impact of the two tax categories using Specification 7. As a result, the response becomes insignificant as soon as we incorporate tax P I. Thus, the following concern arises. It could be that it is not the personal income tax component of the 2003 JGTRRA that is important for the impact of the 2003 JGTRRA on excess bond premia, financial intermediaries balance sheet conditions, and stock market returns, but rather the change in corporate income taxes associated with the 2003 JGTRRA. And similarly, it could be that the 1986 TRA affects banks balance sheet conditions, and therefore excess bond premia, through the adjustment of personal income taxes. We will analyze this point in greater detail in the next subsection. Specifically, 6
Figure 31: Impulse responses of the share of non-performing loans Notes: Impulse responses of the share of non-performing loans to a one percent increase in corporate income taxes with 95 percent confidence intervals. The underlying VAR is given by Specification 2, see Section 5 of the main paper. We include both tax P t I in Appendix C. and tax CI t. Narrative shocks are corrected for measurement error, as described we will show that this is not the case. A comparison of two tax acts and their impact on banks balance sheets: the 1986 TRA versus the 2003 JGTRRA. We now argue that, by combining our evidence for the 1986 TRA and for the 2003 JGTRRA, we can infer that the 1986 TRA operates through corporate income tax adjustments, whereas the 2003 JGTRRA affects financial intermediaries balance sheet conditions via its personal income tax component. Our main observation is that the two tax acts change intermediaries balance sheets through different channels. Key for our argumentation is the fact that determinants of intermediaries balance sheets that are more likely to respond after an adjustment in personal income taxes are affected by the 2003 JGTRRA, whereas the 1986 TRA appears to change determinants that are more likely to be affected by corporate income tax changes. Before we turn to the details, recall from Figure 5 that the 1986 TRA is by far the most important corporate income tax change in our sample, whereas the 2003 JGTRRA had a major impact on personal income taxation. Now, consider the 2003 JGTRRA in greater detail. We have documented in Section 5.1 of the paper that this tax act gives rise to changes in (excess) stock market returns. Stocks are the second most important component of financial intermediaries assets, after loans. The quality of loans does not appear to be affected by the 2003 JGTRRA, since we concluded above that the 1986 TRA is able to explain the response of non-performing loans and of the loan loss provisions/loan ratio that we observe using all tax changes. Given that the 2003 JGTRRA affects the return on assets for financial intermediaries, it seems likely that excess stock market returns matter for the transmission of the 2003 JGTRRA to intermediaries balance sheet conditions. Regarding the question whether the response of stock market returns is a consequence of adjustments in personal or in corporate income taxation, we note that the 2003 JGTRRA entails by far the most 7
significant dividend/capital gain tax change in our sample. There is evidence by Sialm (2009) and others that dividend/capital gain taxes influence equity valuations. In contrast, there is no such strong evidence for the impact of corporate income tax changes on aggregate stock prices, see Cutler (1988). Since dividend/capital gain taxes are part of the personal income tax component of the 2003 JGTRRA, we conclude that the 2003 JGTRRA affects stock market returns through adjustments in personal income taxation. We now turn to the 1986 TRA. As noted above, corporate income tax changes in general and especially the 1986 TRA do not have an impact on excess stock market returns. For the 1986 TRA, this finding is to be expected, given that it only had a minor impact on dividend/capital gain taxes (see e.g. Figure 1 in Sialm 2009). Instead, we argued in Section 5.2 of the paper that the 1986 TRA affects banks balance sheet conditions through changes in the fraction of non-performing C&I loans. Arguably, it is more likely that these changes in the fraction of non-performing C&I loans stem from a corporate income tax adjustment, and not from a personal income tax change. In conclusion, we find that our results are robust to controlling for both personal and corporate income tax changes. In particular, we argue that the 1986 TRA affects financial intermediaries balance sheets via its corporate income tax component, whereas the 2003 JGTRRA operates through its personal income tax component. 8
References Cutler, David M. 1988. Tax reform and the stock market: An asset price approach. American Economic Review 78 (5): 1107 1117. Mertens, Karel, and Morten O. Ravn. 2012. Empirical evidence on the aggregate effects of anticipated and unanticipated US tax policy shocks. American Economic Journal: Economic Policy 4 (2): 145 181. Sialm, Clemens. 2009. Tax changes and asset pricing. American Economic Review 99 (4): 1356 1383. 9