Asymmetric Treatment of Tax Losses and Corporate Investment

Transcription

1 FAccT Center Working Paper Series Asymmetric Treatment of Tax Losses and Corporate Investment Inga Bethmann Martin Jacob Maximilian A. Müller FAccT Center Working Paper Nr. 20/2016 FAccT Center - WHU Financial Accounting & Tax Center WHU Otto Beisheim School of Management Campus Vallendar, Burgplatz 2, Vallendar, Germany Phone: FAccT@whu.edu Internet: whu-facct.de Electronic copy available at:

2 Asymmetric Treatment of Tax Losses and Corporate Investment * Inga Bethmann, Martin Jacob, and Maximilian A. Müller This version: January 2016 ABSTRACT Using a large sample of European private firms, we find that a less asymmetric treatment of tax losses via loss carrybacks increases loss firms investment. About half of the tax refund from loss carryback is reinvested in capital investments. The effect is strongest for financially constrained firms. We also document a trade-off in granting less restrictive tax refunds: While allowing loss carrybacks translates into higher economic output, lowproductivity firms have higher survival rates under carryback regimes. This suggests that loss carrybacks could prevent market share from being moved from less productive loss firms to more efficient competitors. Keywords: Corporate taxation, tax policy, tax loss carryback, tax asymmetry, corporate investments, misallocation JEL classification: G31, H21, H25 * We are grateful to Anna Alexander, Kathleen Andries, Yakov Amihud, Kay Blaufus, Matthias Breuer, Martin Glaum, Michelle Hanlon, Wojciech Kopczuk, Dominika Langenmayr, Richard Sansing, Joel Slemrod, Kjetil Telle, and seminar and conference participants at WHU Otto Beisheim School of Management and the IIPF Doctoral School on Tax Systems in Warsaw, Poland for helpful comments and suggestions. We appreciate the help from local experts from Ernst & Young and KPMG in obtaining information on institutional tax details. WHU Otto Beisheim School of Management. Corresponding author: Martin Jacob, WHU Otto Beisheim School of Management, Burgplatz 2, Vallendar, Germany; phone ; fax ; martin.jacob@whu.edu. Visiting at University of Chicago Booth School of Business under the DAAD Postdoc Program. Electronic copy available at:

3 1. Introduction While profits lead to immediate tax payments, losses are treated asymmetrically because they do not lead to immediate tax refunds in many cases. Most countries allow reducing tax payments on future profits by carrying losses forward, but only some resolve the asymmetry by allowing immediate refunds of tax payments on past profits through loss carrybacks, without any consistent view emerging about the appropriate treatment of tax losses (Altshuler and Auerbach 1990: 61). While providing loss firms less restrictively with tax refunds could alter their marginal tax rate, relax financing frictions, and encourage risk taking, it bears the risk of supporting losers (Auerbach 1986). Under a carryback regime, firms receive tax refunds irrespective of their future profitability and, empirically, every second loss firm remains unprofitable. Given this tension, we examine whether a less asymmetric treatment of tax losses translates into higher investment by relaxing financing frictions and, if so, whether this less restrictive allocation of government funds to loss firms is inefficient. It is a long-standing question whether the asymmetric treatment of tax losses affects investment. Prior theoretical literature shows that tax asymmetry may affect investment by changing the user cost of capital (e.g., Auerbach 1986, Mayer 1986). The empirical evidence testing this channel, however, has been mixed potentially because the direction of the tax asymmetry effect crucially depends on a precise measurement of firms tax status and loss balance (e.g., Devereux, Keen, and Schiantarelli 1994, Cummins, Hassett, and Hubbard 1995). 1 While this literature assumes perfect capital markets in which firms invest until the marginal return is equal to the cost of capital, ample evidence suggests that firms could be financially constrained making their investment sensitive to the amount of cash available internally (e.g., Fazzari, Hubbard and Petersen 1988, Faulkender and Petersen 2012). Further, recent papers emphasize that the effectiveness of tax investment incentives depends on the level of financial constraints (Edgerton 2010, Zwick and Mahon 2014). Hence, given that tax asymmetry alters loss firms cash flows, we examine whether the asymmetric treatment of tax losses affects investment though increased liquidity. Loss firms are (weakly) more liquid under carryback as opposed to carryforward regimes. If loss firms are financially constrained, we expect their investment to be higher under the less asymmetric carryback regime 1 While Devereux, Keen, and Schiantarelli (1994) find that tax asymmetries are not important when modeling firms investment behavior, Cummins, Hassett, and Hubbard (1995) find no evidence that firms with tax loss carryforwards respond to tax changes. Edgerton (2010) documents that tax asymmetry affects firms investment response to other tax incentives but is silent on the direct effect of tax asymmetry on investments. Dreßler and Overesch (2013) report that a more asymmetric treatment of losses via stricter carryforwards decreases investments only in industries with a high loss probability. Langenmayr and Lester (2015) find that stricter carryforwards decrease the riskiness of investments. 1

4 because they have more internal cash available to invest (e.g., Cohn 2011). The drawback of a less asymmetric treatment of tax losses, however, is that loss firms receive liquidity irrespective of their future prospects and productivity under a carryback regime. Hence, subsidizing loss firms investment could delay their exit and distort the competitive selection of operating firms (Syverson 2011) if loss firms are financially constrained because of their low productivity. Our paper makes two key contributions to prior literature. First, we explore a new channel liquidity through which tax asymmetry affects investment. Second, we examine whether encouraging loss firms investment through a less asymmetric treatment of tax losses leads to a misallocation of government funds. To identify the effect of the asymmetric tax treatment of profits and losses on corporate investment, our identification strategy exploits idiosyncratic shocks to profitability and cross-country variation in more asymmetric treatment (only carryforward allowed) versus less asymmetric treatment (carryback also permitted). Since we include firm fixed effects and country year fixed effects, our identification strategy is akin to a difference-in-differences design. The first difference compares investments of firms in countries with loss carryback provisions to countries without these provisions. The second difference compares the investments of firms with losses and preceding profits, that is, firms that could be eligible for a loss carryback and which face a tax asymmetry, to the investments of all other firms, such as profitable firms. We include country year fixed effects to absorb any unobservable omitted time-varying country characteristics as well as observable macroeconomic characteristics and related tax policy variables. Our approach is thus distinct from prior literature that relies on economic crises that simultaneously lead to financing constraints and variation in how tax losses are treated (e.g., Dobridge 2015). We use data of over 900,000 private firms from 21 European countries, five of which permit loss carrybacks. Using private firm data allows us to make inferences about the investment responses of a substantial part of the overall firm population. Private firms represent roughly two-thirds (half) of the overall employment in the European Union (OECD countries). 2 We show that a less asymmetric treatment of tax losses via loss carryback increases the investments of loss firms. Controlling for country year and firm fixed effects, we find 4.5% higher capital investments for loss firms under a carryback relative to a carryforward regime when carrybacks provide loss firms with liquidity (i.e., in the year after their loss). Thus, a less asymmetric treatment of losses via loss carryback relative to loss carryforwards leads to higher investments in the period after the loss. Supporting our causal interpretation, we document that the effect of loss carrybacks increases with the size of the tax refund. 2 Source: last accessed November 30,

5 Our results indicate that a one standard deviation increase in the potential tax refund leads to an investment increase of % of average investments. To put this into perspective, this result indicates that up to half of the tax refund is used to fund capital investments. These primary results are robust to a variety of additional tests. To corroborate that the mechanism underlying the investment response stems from liquidity, we show that part of the tax refund is set aside as a financial cushion to increase liquidity. To further validate the mechanism, we examine the exact timing of the effect on investments. We find the effect is insignificant in the year of the loss, consistent with firms anticipating but not yet receiving refunds from carrybacks when they learn about their loss over the fiscal year. Further, the effect is insignificant two years after the loss, inconsistent with a potential omitted variable bias driving our primary results. Our primary results are also robust to the inclusion of other tax policy tools that can have similar effects on liquidity, such as group taxation regimes. This confirms our expectation that the investment effect is driven by loss carryback provisions and no other tax policy. Moreover and in contrast to Dobridge (2015), we document that our findings also hold during severe recessions. Finally, we run a placebo test in which we show that firms with consecutive losses that are not eligible for a tax loss carryback because they were not profitable and, hence, did not pay taxes in prior years do not increase investments in the year following their loss. In the next step, we examine cross-sectional differences. Recent studies of how tax incentives affect investment highlight the important role of financial frictions during adverse economic conditions (Zwick and Mahon 2014). In addition, Erickson, Heitzman, and Zhang (2013) document that financial constraints drive US firms decision to increase cash flow through loss carrybacks during economic downturns. Finally, the sensitivity of investment to liquidity provided via carrybacks should be high (low) if firms face high (low) financing constraints (e.g., Fazzari, Hubbard, and Petersen 1988, Faulkender and Petersen 2012). To this end, we sort unprofitable firms into quintiles of cash holdings in the year of the loss and examine how the responsiveness of investment to loss carryback provisions differs across them. As expected, we show that the effect of tax loss asymmetry on investment is primarily a function of financial constraints. We find the largest investment responses for firms in the bottom quintile of cash holdings. In contrast, we find no significant response for firms in the upper quintile of cash holdings. There are two possible interpretations of our primary findings. First, loss carryback regimes may efficiently alleviate financing frictions in capital markets that could, for example, result from adverse credit supply shocks (Zwick and Mahon 2014) and provide firms with the necessary liquidity. Second and in contrast to this view, loss firms may be financially constrained because they have lower 3

6 productivity (e.g., Hopenhayn 2014). Accordingly, carrybacks would unduly subsidize these firms and potentially distort the competitive selection of operating firms. It is hard to empirically disentangle these explanations. First, under both interpretations, carrybacks would increase loss firms survival rate, a result that we document. Second, loss firms would tend to hold more cash, a result described earlier, to finance future investment opportunities (Almeida, Campello, and Weisbach 2004) or protect against the threat of product market predation (Haushalter, Klasa, and Maxwell 2007) that they would likely be subject to as low-productivity firms. To disentangle whether these interpretations are complementary or mutually exclusive, we examine other economic consequences that potentially follow from increased liquidity and investment. On the one hand, we find that loss firms output under a carryback regime is higher a year after the tax refund consistent with the existence of capital market imperfections that are partly resolved by carrybacks. That is, investment differences translate into greater economic output under a less asymmetric treatment of tax losses. On the other hand, we show that the exit of low-productivity firms is significantly delayed when loss firms receive tax refunds less restrictively via carrybacks. Overall, our results suggest that a less asymmetric treatment of tax losses can efficiently relax financing frictions resulting from imperfect capital markets. However, the evidence also indicates that this conclusion is likely associated with a high type I error: Less restrictive allocation of government funds via loss carrybacks reduces competition because it prevents market share from being moved from less productive loss firms to more efficient competitors (Syverson 2004, 2011). While our findings that loss firms survive longer albeit with low productivity are consistent with this notion, they yield no inferences on the aggregate productivity gains or losses of treating tax losses less asymmetrically by allowing losses to be carried back. These findings are relevant to policymakers since they help them evaluate and anticipate the effects of different approaches to resolve the asymmetric treatment of tax losses. Loss firms represent a substantial part of the firm population. In about 20% of the firm year observations in our sample, firms report losses and half of these are eligible for tax loss carrybacks because they were profitable in the year before their loss. About 48% of the firms in the sample realize a loss at least once during our sample period. Carryback rules resolve the asymmetric treatment of tax losses less restrictively. This difference translates into higher investments, especially when firms face financial frictions. However, providing unconditional cash could come at the cost of reduced competition because loss firms survive longer albeit with low productivity. Consistent with the trade-off between providing liquidity to 4

7 financially constrained firms versus reduced competition, only few countries allow firms to carry back losses. Our paper also contributes to the literature on the investment effects of tax policy (e.g., Summers 1981, Cummins, Hassett, and Hubbard 1994, 1996, Chirinko, Fazzari, and Meyer 1999) by showing that differences in the asymmetric treatment of losses affect the investments of financially constrained firms. Prior literature focuses on statutory or effective corporate tax rates and their investment effects (e.g., Djankov et al. 2010) without explicitly relating the effects to specific tax policy tools such as loss carryback or carryforward provisions. Our paper also contributes to the literature on temporary tax incentives such as bonus depreciation (e.g., House and Shapiro 2008, Edgerton 2010). We complement prior papers that document that these incentives work primarily when firms face financing frictions and have a strong preference for immediate liquidity (Zwick and Mahon 2014). The remainder of the paper is organized as follows: Section 2 provides an overview of tax loss offset provisions in the countries and years of our sample and develops the hypothesis. Section 3 presents the estimation strategy and data. The main results and robustness tests are discussed in Section 4. Section 5 discusses whether loss carryback provisions are an efficient allocation of capital. Section 6 concludes the paper. 2. Asymmetric treatment of tax losses across Europe and investment responses 2.1 Expected investment responses to loss carryback regimes Typically, countries treat profits and losses asymmetrically: While profits are taxed when incurred, there is no immediate refund for losses. The two regimes present in most economies around the world differ in the way they resolve the asymmetric treatment of tax losses. While tax loss carryforward rules allow firms to offset current losses with future profits, tax loss carryback rules allow firms to offset current losses with the prior year s profits and to receive a refund of previously paid taxes. Typically, tax losses will result in earlier and less uncertain tax refunds under carryback regimes as opposed to carryforward regimes. Hence, carryback regimes result in lower asymmetry. We use this asymmetry differential between the two regimes as well as the transitory nature of losses to identify the effect of the asymmetric treatment of tax losses on the investment of loss firms. In countries with loss carryback provisions, firms experiencing a loss in the current period (t) receive a reimbursement of taxes paid on prior profits (e.g., t - 1) in the period after the loss (t + 1) 5

8 because the tax statement for fiscal year t is filed in t Hence, the tax refund in a carryback regime is unconditional on future profitability (t + 1 and later). In contrast, in countries with only loss carryforward provisions, a loss in the current period (t) can be used to reduce taxable income and, hence, tax payments in t + 1 or later if the firm reports taxable income (profits). That is, benefits arising from a loss in t are an increasing function of future profitability. To compare the effect of the two regimes, we focus on firms that report a profit in period t - 1 and a loss in t (identified in our empirical analyses by the variable Ploss, as explained below). These firms would be eligible for a carryback if their country allowed losses to be carried back. For this firm, the liquidity benefits arising from carrybacks and carryforwards are only equivalent if profit in t + 1 is at least equal to the firm s loss in t in absolute terms. In all other cases, that is, when the firm does not fully recoup its loss, it is weakly more liquid under a carryback regime in t + 1. In our sample, only 50% of loss firms return to profitability in t + 1, indicating that tax benefits are more uncertain and lower under tax loss carryforward regimes relative to a carryback regime for the average loss firm. 4 That is, if the loss firm does not fully recover its loss in t + 1 with profits, the after-tax cash available is higher under a loss carryback regime than under a loss carryforward regime. We examine whether this liquidity advantage under carrybacks resulting from a less asymmetric treatment of tax losses benefits loss firms investment. In perfect capital markets, liquidity differences across the two tax loss regimes would not result in investment differences in t + 1 because firms would already exploit all projects with a positive net present value. However, if the cost of external finance is greater than the cost of internal finance (imperfect capital markets), investment depends on available cash (Fazzari, Hubbard, Petersen 1988). Put differently, frictions that increase the wedge between external and internal finance prevent firms from investing in all projects with positive net present value. For example, capital markets may ration capital to the firm because of asymmetric information and agency conflicts (e.g., Blanchard, Lopez-de- Silanes, and Shleifer 1994) or adverse macroeconomic conditions (e.g., Lemmon and Roberts 2010, 3 4 We gain this insight from several conversations with tax professionals in different countries and note that it differs slightly from the way carrybacks have been modeled by prior literature (e.g., Graham 1996, Mahon and Zwick 2015). Assuming carryback refunds arise in period t would predict a larger investment response in that period. Instead and in line with our anecdotal evidence on the timing of the tax filing and processing by the tax authorities, we find the investment response is clustered in period t + 1, when the tax refund is paid. In addition, for some firms, changes in ownership rules that lead to the forfeiture of loss carryforwards in case of a substantial change in ownership could limit the value of loss carryforwards. Moreover, minimum taxation rules that limit the use of loss carryforwards to a certain amount in each year lead to lower values of loss carryforwards for some firms. Furthermore, loss carrybacks are more beneficial in the case of decreasing corporate tax rates in t + 1 because of higher refunds for the same amount of losses. While tax rate changes occur only in one out of every six country years in our sample, around 80% of these tax changes are tax rate decreases. 6

9 Chava and Purnanandam 2011). Hence, to the extent the average loss firm in our sample is weakly restricted in its access to external capital, we expect that the liquidity advantage of a less asymmetric treatment of tax losses via carrybacks will translate into more investment. 5 Accordingly, Zwick and Mahon (2014) find that firms only respond to investment tax incentives when they relax financial frictions and Erickson, Heitzman, and Zhang (2013) document that financial frictions are among the primary incentives to take up carrybacks. However, it is important to note that financial constraints also create incentives to save cash 6 (Almeida, Campello, and Weisbach 2004) or pay down debt (Acharya, Almeida, and Campello 2007) to finance future investment opportunities. This weakens the sensitivity of investment to cash flows resulting from loss carrybacks. Further, prior literature documents that the sensitivity depends on economic conditions (e.g., Dobridge 2015). We argue these effects will attenuate but not eliminate the sensitivity of loss firms investment to liquidity resulting from a less asymmetric treatment of tax losses. Hence, we formulate our hypothesis as follows. Hypothesis: A less asymmetric treatment of tax losses via carryback provisions leads to a higher investment of loss firms relative to a more asymmetric treatment of tax losses via carryforwards. 2.2 Tax loss treatment in Europe Table 1 provides an overview of tax loss carryback provisions in our sample countries over Only four countries (Germany, Ireland, the Netherlands, and the United Kingdom) allowed tax loss carrybacks in all sample years. 7 Countries appear to temporarily adopt loss carryback regimes to provide loss firms with liquidity during adverse economic and idiosyncratic conditions. For example, the United States extended the scope of its loss carryback provisions during the economic downturns in 2002 and 2009 to give cash infusions to unprofitable firms that will help them weather the current economic storm. 8 In 2012, Germany increased the loss amount that can be carried back to help small and medium-sized firms ride out difficult times with the additional short-term liquidity. 9 Norway introduced loss carryback provisions during the financial crisis Prior literature shows that the incidence of losses is negatively associated with the costs of capital (e.g., Dechow and Dichev 2002, Francis et al. 2004, Francis, Nanda, and Olsson 2008). Increasing cash holdings would also be consistent with the precautionary savings motive of high-risk firms (e.g., Bates, Kahle, and Stulz 2009) and to protect against the product market predation of competitors (Haushalter, Klasa, and Maxwell 2007), as further explored in Section 5.1. Jacob, Michaely, and Müller (2015) document a similar distribution across 68 countries during U.S. Committee on Ways and Means, Summary of Camp Cantor Substitute to H.R. 1, January 28, 2009, last accessed November 30, German parliament, September 5, 2012, p. 18, last accessed November 30,

10 While most of the countries allowed a carryback period of one year, Netherlands allowed a threeyear carryback window in 2006 and Norway allowed a two-year carryback window in 2008 and France grants a tax credit for losses carried back to the previous three years that can be used to reduce future tax liabilities during the following five years; however, because loss firms do not receive immediate refunds, we exclude France from our main analyses and discuss robustness to the inclusion of France in our sample in Section [Insert Table 1 about here] All our sample countries, as most economies around the world, allow firms to carry forward losses. Tax loss carryforward provisions primarily differ across countries in the length of the carryforward period. While some countries limit the loss carryforward period to five to 15 years, more than half the countries in our sample allow unlimited loss carryforwards (see Table A.1 of the Online Appendix for an overview). 12 Further, some countries limit the yearly carryforward amount. 3. Estimation strategy and data 3.1 Empirical identification To identify the effect of an asymmetric treatment of tax losses on the investment of loss firms, we compare the investment levels of loss firms across countries with and without loss carryback provisions in our research design. Since we include firm and year fixed effects in our regression model, our identification is akin to a difference-in-differences design. The first difference compares investments in countries with and without loss carryback provisions. The second difference compares investments of firms with a loss in year t and a profit in t - 1 (Ploss = 1) to firms with a profit in year t However, the length of the carryback period does not provide enough variation to measure loss carryback provisions as a continuous variable, which would allow inferences about the investment effect of extending the loss carryback period by one year. Therefore, our study focuses on differences in the existence of any loss carryback rules. Sweden allows a loss offset with a profit periodization reserve. A taxpayer is allowed to allocate up to 25% of net profits to this reserve, while allocations are not included in taxable income. The reserve must be added to taxable income six years after the allocation at the latest. We do not treat the loss offset with this profit reserve as a loss carryback provision since there will be no cash effect in the period after the loss event (only a reduction in the profit periodization reserve). Moreover, we were not able to observe the amount of this reserve at the firm level. While our analysis controls for the restrictiveness of loss carryforward provisions, we do not think that differences in the length of the loss carryforward period represents a binding constraint for the average loss firm. That is, the difference between a 15-year carryforward window and unlimited carryforward windows is most likely diminishing. Dreßler and Overesch (2013), who examine how tax loss offset provisions affect a firm s investment behavior irrespective of an actual loss event, find that loss carryforward periods decrease investments only if the carryforward is restricted to less than six years and for firms with high loss probabilities. 8

11 or firms with subsequent losses (Ploss = 0). 13 The former firms are potentially eligible for a tax refund, while the latter would not receive an immediate and unconditional refund, irrespective of the tax regime. Our identification stems from the interaction of these two differences. The first difference reflects cross-country differences and the second difference is based on idiosyncratic shocks to profitability. To investigate the effect of loss carryback provisions on investments, we test the following model: Investmenti, t 1 = α 0 + αi + αt + β1 LCB j, t + β2 Plossi, t + β3 Ploss LCB + i, t j, t k γ X + δ C + ε k kit,, l l, jt, + 1 it, l + (1) where Investment i,t+1 is defined as the change in fixed assets before depreciation relative to the prior year s total assets in period t + 1. The interaction between Ploss i,t and LCB j,t captures the main coefficient of interest. The variable LCB j,t is a dummy variable equal to one if the country allows tax loss carrybacks and Ploss i,t is an indicator variable equal to one if the firm is potentially eligible for a tax loss carryback, irrespective of whether its country allows a carryback, that is, if the firm reports a loss EBIT < 0 in the current year and a profit EBIT > 0 in the previous period. 14 Our hypothesis that unprofitable firms in loss carryback regimes have higher investments when they receive the tax refund predicts a positive sign for the interaction of Ploss and LCB (β 3 > 0). As an alternative measure for Ploss, we use the continuous variable Ploss Intensity, our refined measure of the expected tax refund. The variable Ploss Intensity is a proxy for the refund for a given loss in t and a given profit in t - 1. In case a firm realizes a profit in the current year or losses in two consecutive years, we set Ploss Intensity to zero. Since we predict that the effect of a less asymmetric treatment of tax losses on investment works through increasing internal cash flows, we expect that the refund size for firms in countries allowing carrybacks is positively related to investments. Hence, we expect to obtain a positive sign for the interaction of Ploss Intensity and LCB (β 3 > 0). We assume that the average loss firm eligible for a loss carryback claims the refund. For the United States, Edgerton (2010) and Mahon and Zwick (2015) report that only a third of eligible firms claim loss carryback refunds. In contrast to the United States, carrybacks are the default option in some of our sample This is a simplification. The Netherlands allowed a loss carryback for the previous three periods in 2006 and Norway allowed a two-year carryback window in 2008 and Thus, we potentially underestimate the effect, since we neglect some firms that are eligible for a tax loss carryback. We use EBIT and not profits before tax because several sample countries limit the amount of interest expenses that are tax deductible. We test the robustness of our results to using profits before tax as an alternative profitability proxy and find even stronger investment effects of loss carryback provisions. If EBIT are negative, a firm very likely also has negative taxable income. 9

12 countries with carryback provisions. That is, eligible firms would have to actively opt out to carry forward their loss (Germany) or do not have a choice (the Netherlands). While the system in the United Kingdom and Ireland is comparable to that in the United States, our anecdotal evidence from conversations with tax professionals in these countries indicates that most firms eligible for a carryback actually claim this refund, because, for example, in the United Kingdom, the tax refund from the carryback is higher than a refund from loss carryforward in most of our sample years because of decreasing corporate tax rates. While we do not have access to proprietary data that would allow us to further examine this issue, our evidence largely confirms that the effects we document are a function of the liquidity differential created by carrybacks. Firm fixed effects further control for all unobservable and observable time-invariant firm, industry, and country characteristics, such as time-invariant differences in risk taking that may relate to the level of tax asymmetry in the respective country (e.g., Domar and Musgrave 1944, Stiglitz 1969, Auerbach 1986). The main effect of LCB (β 1 ) is difficult to interpret, since its identification stems only from the change in loss carryback provisions in Norway in 2008 and Further, we predict a negative sign for the coefficient of Ploss (β 2 ) since we expect firms with a loss in t to invest less, on average, in loss years than in profitable years. 15 X k,i,t is a vector of firm-level variables that comprises controls for firm-level investment determinants. The variable Start-Up is an indicator variable equal to one if the firm was registered for less than five years and Sales growth is measured as the average percentage change in sales over the past two years and captures investment opportunities. We expect young, growing firms to invest more than large and mature firms. Hence, we expect positive signs for the coefficients of Start-Up and Sales growth. We also include Size, which is measured as the natural logarithm of lagged total assets, and expect a negative sign since mature firms invest less than young firms. We control for Cash (EBIT), defined as cash holdings (earnings before interest and tax) scaled by lagged total assets to control for the availability of internal funds. We expect Cash and EBIT to have positive signs. We also control for a firm s leverage and wage expenses. We further control for the importance of fixed assets, defined as the prior year s fixed assets over the prior year s total assets, in a company and expect a negative coefficient, since high levels of fixed assets are an indicator that firms have fewer unused investment 15 One concern could be that firms manage their losses. Studies show that firms engage in earnings management to optimize tax refunds from tax loss carryback provisions (e.g., Maydew 1997, Erickson, Heitzman, and Zhang (2013). However, conceptually, potential earnings management incentives cannot influence our results because we examine the effect of investment in the year following the loss. Assuming constant corporate tax rates, shifting expenses from t + 1 (the subsequent profit year and the year of investment) to t (the loss year) does not change the after-tax cash flow or the tax payment in t

13 opportunities. To isolate the cash effect of a loss carryback in year t + 1 from the effect of control variables, we control for firm variables in period t, that is, in the year in which we measure Ploss and Ploss Intensity, respectively. C l,j,t+1 is a vector of country-level controls, where ln(gdp) represents the natural logarithm of the gross domestic product (GDP) of country j, and GDP growth is the annual percentage change in the GDP in percent. The variable RQ is the World Governance Indicator proxy for regulatory quality. This variable captures the perception of a government s ability to formulate and implement sound policies and promote private sector development. We expect all three variables to have a positive effect on investments. In line with prior literature, we expect a negative sign for the coefficient of STR, which represents the statutory corporate income tax rate of country j. We include macroeconomic variables in the year of investments (t + 1). The variable LCF is an index for the restrictiveness of loss carryforward provisions that ranges between zero (most restrictive) to one (least restrictive). 16 Consistent with prior literature and the risk-incentive argument (Dreßler and Overesch 2013, Haufler, Norbäck, and Persson 2014, Langenmayr and Lester 2015), we expect less restrictive loss carryforward provision to incentivize investment. 17 Further, we include year fixed effects (α t ) that control for yearspecific factors, such as general economic downturns. In all tests, the standard errors are robust and clustered at the firm level to account for the within-firm correlation of our observations (Petersen 2009). We also address concerns that unprofitable firms generally recover better in loss carryback countries, irrespective of the tax legislation, since these countries have, on average, a higher GDP and high regulatory quality. In addition, higher GDP growth can also facilitate the recovery of unprofitable firms. Therefore, we include the interactions of Ploss and all macroeconomic variables to ensure that we separate confounding effects. Moreover, we include an interaction of Ploss and LCF because loss carryback countries are generally less restrictive with respect to loss carryforward provisions (see Table 1 and Table A.1 of the Online Appendix). In addition, we address concerns that specific firm We define three categories to evaluate the restrictiveness of loss carryforward periods: (i) the length of the loss carryforward period, (ii) minimum taxation rules, and (iii) other criteria, such as longer carryforward periods for start-up firms or special industries. We assign zero points if the carryforward period is less than five years, 0.5 points for periods between five and 15 years, and one point if the carryforward period is more than 15 years. We subtract 0.25 points if there are minimum taxation rules in place. We add 0.25 if there are any other benefits such as longer carryforward periods for start-up firms if the carryforward period is no longer than 15 years. With this ranking, we assign an index that ranges between zero and one for every jurisdiction in every year of our sample. Our main results are robust to controlling for other loss offset provisions such as group taxation regimes that may correlate with loss carryback provisions and that are likely to have a positive effect on the investment behavior of unprofitable firms that are part of a corporate group. 11

14 characteristics (e.g., size, age, and leverage) facilitate recovery from a negative earnings shock. To this end, we include the interactions of Ploss with all firm-level control variables. To address potential concerns that our results may be explained by unobserved variation in economic conditions or broader tax policy changes (e.g., Jacob, Michaely, and Müller 2015), we also run tests including country year fixed effects that control for any unobservable and unobservable timevarying country-level characteristics affecting investment. Controlling for country year fixed effects, for example, addresses the concern that some countries with loss carryback provisions implemented other non-tax regulations to stabilize the economy during crises. 18 We also interact the control variables with either country or year fixed effects to demonstrate that our results are not driven by differences in the investment sensitivity of our control variables over time or across countries. 3.2 Data sample and descriptive statistics We use firm-level panel data of European private firms from Bureau van Dijk s Amadeus database. The database contains information on the unconsolidated financial statements of European companies. Using private firm data allows us to make inferences about the investment responses of a substantial part of the overall firm population that is responsible for roughly two-thirds (half) of the overall employment in the European Union (OECD countries). 19 In addition, private firms are more heterogeneous than public firms are with respect to characteristics such as financial constraints that are important for our analysis. 20 Our final dataset covers Austria, Belgium, Croatia, the Czech Republic, Denmark, Finland, Germany, Hungary, Ireland, Italy, Luxembourg, the Netherlands, Norway, Poland, Portugal, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom. We analyze 916,099 firms over the period with no missing data for any of our main and control variables. Using the sample selection criteria discussed in the following, we obtain a total of 3,712,216 firm year observations. We complement the firm-level data with country-level information on loss offset provisions and statutory tax rates that we collected from IBFD s tax handbooks and Ernst and Young s Worldwide Corporate Tax Guide. We merge additional country-level variables on GDP, GDP growth (World Bank), and regulatory quality (World Governance Indicator). As explained earlier, we exclude France from our main analyses because it cannot be clearly classified as a loss carryback country. We document the For example, Germany introduced scrappage schemes in Source: last accessed November 30, Compared to private firms, public firms have easy access to debt and equity markets. Small (i.e., private) firms are more likely to be affected by financial constraints and are more bank dependent than public firms are (e.g., Beck, Demirgüç- Kunt, and Maksimovic 2005, Behr, Norden, and Noth 2013). 12

15 robustness of our findings to the inclusion of France as a loss carryback country or as a country without loss carryback provisions in Section 4.2. We eliminate financial and utility firms since they are subject to different regulation that most likely affects their investment behavior. We exclude small companies with fixed assets and sales lower than 50,000 because regular investments of such small, fast-growing firms inflate our investment measure that is defined relative to the prior year s assets. We also exclude firms with negative values for total assets and cash. The data are adjusted for inflation using the Consumer Price Index with 2010 as the base year for each country. To compute our size measure (the natural logarithm of lagged total assets), we convert total assets into euro amounts using the exchange rates provided by Amadeus. All firm-level variables except for Size are winsorized at the 1% and 99% levels in each year. Table 2 provides an overview of the sample composition by country. While loss carryback countries represent 6.6% of the overall sample in terms of observations, they represent 34.4% of the overall sales volume, with Germany and the United Kingdom constituting the majority. Observations under a carryforward only regime mainly stem from Spain and Italy, which represent 27.5% and 31.4%, respectively, of the sample in terms of observations. [Insert Table 2 about here] Descriptive statistics for all the variables are reported in Table 3. In our sample, 10.6% of the observations report a loss following a profit. Figure 1 shows the distribution of Ploss over the sample period for loss carryback countries and countries without loss carryback provisions. The numbers of Ploss events increase for both groups during 2008 and 2009 to 12.1% for loss carryback countries and 13.5% for countries without loss carryback provisions. The mean of Ploss Intensity amounts to 0.31% of total assets. For all firms of the sample with Ploss = 1, average Ploss Intensity amounts to 3.13% of total assets. About 11% of our sample firms are start-ups, that is, firms that have been registered for less than five years. The average firm has a size of 1.51 million. The largest 1% of firms have total assets above 90 million. The average firm has a ratio of fixed assets to total assets of 42%, a wage-toassets ratio of 33.5%, a cash ratio of 11.6%, a debt-to-assets ratio of 72.4%, and a sales-to-assets ratio of 154.4%. EBIT average 6% of total assets and Sales growth over the past two years averages 11.1%. [Insert Figure 1 about here] [Insert Table 3 about here] 13

16 4. Empirical results 4.1 Baseline results Table 4 presents the coefficient estimates for our main variables of the investment regression from equation (1). For our baseline test, we regress Investments in t + 1 on Ploss, LCB, the interaction term Ploss LCB, and all firm- and country-level control variables. 21 We then add interactions and different fixed effects. Consistent with our hypothesis, the coefficient for the interaction term Ploss LCB is positive and significant. That is, a less asymmetric treatment via loss carrybacks leads to higher investments of loss firms that are eligible for a tax refund. The corresponding investment increase amounts to 0.51% of total assets, which is equivalent to 7% of average investments. In columns (2) to (5) of Table 4, we include the interactions of Ploss with the macroeconomic variables ln(gdp), GDP growth, and RQ to address the concern that tax loss carryback provisions are correlated with ln(gdp), GDP growth, and RQ. If these variables are correlated, then our interaction Ploss LCB could, for example, simply pick up quicker investment recovery in countries with higher economic growth (Ploss GDP growth). Moreover, we include the interaction of Ploss and LCF because loss carryback countries are generally less restrictive with respect to loss offset possibilities. With this test, we ensure that the interaction Ploss LCB does not pick up beneficial carryforward provisions. To control for the possibility that loss carrybacks may be correlated with tax rates and that tax rates determine the size of the tax refund facilitating recovery, we additionally control for the interaction of Ploss with STR. Finally, we interact Ploss with all the firm control variables to address the concern that certain firm characteristics, such as size or maturity, could foster the recovery. Importantly and consistent with our hypothesis, we still find a positive and significant coefficient for our main variable of interest, Ploss LCB. The economic magnitude is somewhat smaller but still large: The corresponding relative investment difference amounts to 0.46% of total assets, or 6.3% of average investments. In columns (3) and (4) of Table 4, we interact year and country fixed effects, respectively, with all the firm-level control variables to demonstrate that our main result cannot be explained by differences in investment sensitivity to our control variables over time or across countries. That is, we allow 21 The coefficient estimates of these control variables are broadly consistent with prior literature (see Table A.2 of the Online Appendix). Large and mature firms show, on average, lower investment levels than growing firms. High levels of fixed assets as well as high leverage lead to lower investments, while cash holdings are positively associated with investments. High profitability and high sales are positively related to investments. We observe higher investment levels in large, high-growth economies. 14

17 different coefficients for the firm-level investment determinants in every year and country. The coefficient of Ploss LCB is still positive and significant. In column (5) of Table 4, we document that our results are also robust to the inclusion of country year fixed effects. This provides evidence that our results are not driven by any country year-specific factors such as unobserved variations in economic conditions likely correlated with tax policy setting and investment. Accordingly, we exclude all time-varying country-level variables in this regression. With this conservative estimation, we continue to find that the investments of unprofitable firms increase by 0.33% of total assets in the presence of loss carryback provisions, which equals 4.5% of average investments in our sample. Taken together, we find that a less asymmetric tax treatment of profits and losses via carrybacks increases the average investment of loss firms between 4.5% and 8.1% relative to a more asymmetric treatment via carryforwards. [Insert Table 4 about here] In Panel B of Table 4, we rerun equation (1) and replace the indicator variable Ploss with the continuous loss variable Ploss Intensity. We use this alternative variable as a refined measure of the expected tax refund. We expect that the effect of loss carryback provisions on investment is a function of the size of the tax refund. Recall that our model includes EBIT to control for the level of the actual negative income shock. Hence, Ploss Intensity is a proxy for the refund for a given loss in t and a given profit in t - 1. To the extent that refund differences resulting from a lower tax asymmetry under a loss carryback regime drive our primary result, we expect that higher tax refunds relative to a firm s total assets lead to larger investment responses for loss firms. The results are in line with our expectations. A one standard deviation increase in Ploss Intensity an increase by 1.6% of total assets translates into higher investments of loss firms in loss carryback countries equal to 0.16% of total assets or 2.2% of average investments. To put this effect into perspective and assuming a corporate tax rate of 24% (U.K. tax rate of 2012), a one standard deviation increase in Ploss Intensity an increase by 1.6% of total assets is equivalent to a refund of 0.38% of total assets (= 1.6% 24%). Hence, about half of the tax refund is directly diverted to capital investments. The size of the investment effect is similar to that of Dobridge (2015), who finds that $0.40 of every tax refund dollar is allocated to investments after the extension of the loss carryback period from two to five years in the United States in In 15

18 contrast to Dobridge, who finds no investment effect during the 2009 crisis, we find that this effect is robust during severe economic downturns with negative GDP growth (see Section 4.2). In columns (2) to (5) of Table 4, we repeat the other four specifications with additional interactions and more extensive fixed effects using Ploss Intensity. Importantly, the results hold when we interact Ploss Intensity with the other control variables and when we interact country and year fixed effects with all the firm-level control variables. The results are also robust to including country year fixed effects (column 5). In terms of economic magnitude, the effect of loss carryback provisions for a one standard deviation increase in Ploss Intensity ranges between 1.1% and 2.2% of average investments. To validate the mechanism through which tax loss carrybacks affect firms investment behavior liquidity through tax refunds we now examine whether tax loss carrybacks lead to higher cash holdings in the period after the loss. Since about half of the tax refund is used for investments, part of it may be set aside as cash due to precautionary savings motives during uncertain financing conditions (e.g., Bolton, Cheng, and Wang 2013), the general savings of financially constrained firms (e.g., Almeida, Campello, and Weisbach 2004), or because low-productivity firms maintain a financial cushion to protect against product market predation (Haushalter, Klasa, and Maxwell 2007). We use Cash in t + 1, which is defined as cash holdings relative to the prior year s total assets, as the dependent variable and rerun specifications (2) and (5) of our main regression analysis. Consistent with our expectations, we observe higher cash holdings for loss firms in loss carryback countries in the period after the loss relative to loss firms in a country without loss carryback rules (see Table 5). The less asymmetric loss treatment via loss carrybacks leads to cash holdings that are higher by 0.32% of total assets, which is equal to 2.8% of average cash holdings in our sample. In column (2) of Table 5, we show that this result is robust to the inclusion of country year fixed effects. With this conservative measure that controls for any country year-specific unobservables, the effect is equal to 0.27% of total assets, or 2.3% of average cash holdings. We run the same regression with the continuous loss variable Ploss Intensity, our refined measure of the expected tax refund, in Panel B of Table 5. We again expect that the cash increase is a function of the size of the expected cash inflow resulting from the tax refund. Consistent with our expectations, the results indicate that a one standard deviation increase in Ploss Intensity an increase by 1.6% of total assets translates into an increase in cash holdings by 0.08% of total assets (0.7% of average cash holdings in our sample). To put this result into perspective, assuming a corporate tax rate of 24% (the U.K. tax rate in 2012), a one standard deviation increase in Ploss Intensity an increase by 1.6% of 16