Restructuring Announcements and Rivals’ Investment Elena Simintzi∗ February 2017

Abstract I examine the effect of firms’ restructuring announcements on the investment decisions of their rivals. I show that restructuring news that signal an improvement in the competitive position of UK manufacturing sites are associated with a 6% increase in local competitors’ capital investment and I demonstrate that this effect is driven by the low debt competitors. Those findings are consistent with the hypothesis that firms react to the competitive threat from the announcing firm by adjusting their physical capital. Several robustness tests, including a placebo test, suggest our results are not driven by confounding factors.

Affiliation: ∗ Sauder School of Business, University of British Columbia. E-mail: [email protected]. Acknowledgments: I am very grateful to Francesca Cornelli, Vikrant Vig, Julian Franks, Mariassunta Giannetti, Francisco Gomes, Denis Gromb, Christopher Hennessy, Ron Masulis, Gordon Phillips, Per Str¨ omberg, and Paolo Volpin for invaluable discussions and comments. This work has also benefited greatly from the thoughts of all the participants of the finance seminars at London Business School, Boston University, University of Toronto, UBC, USC, University of Michigan, University of Illinois Urbana-Champaign, University of Texas-Austin, Cornell University, Federal Reserve Board and the conference participants at the Western Finance Association Annual Meetings.

I

Introduction

Corporate finance has traditionally examined the nature of corporate reorganization decisions and their impact on firms’ future performance (Kang and Shivdasani, 1997; Denis and Kruse 2000; Denis and McConnell, 2006). Since the early work by Kovenock and Phillips (1997), it has been established that firms’ financial restructuring decisions, such as leverage buyouts and recapitalizations, have real effects not only on the restructuring firms themselves, but also on their rivals. Less is known, however, about the way firms’ operational reorganizations that aim to improve their competitive position in the product market impact their rivals’ investment decisions and how the rivals’ response interacts with financial leverage. I use a novel hand-collected dataset on restructuring announcements of UK manufacturing sites; all announcements involve layoffs and span the period 2002-2008. I group the restructuring announcements in two categories: those intended to improve the competitive position of the announcing site (e.g. a site modernizing its technology and laying off workers) and those revealing bad news about the firm or the industry (e.g. liquidation). I focus on the first set of announcements and show that close competitors, defined as those active in the same industry and located in the same region, increase their capital investment following the news that their peer is taking actions which may improve its competitiveness. I further document that financial leverage is a key determinant of the rivals’ ability to respond to the competitive challenge revealed by the news: only rivals with low leverage respond to the restructuring announcement. These results are consistent with industrial organization models which predict that

–1–

firms respond to competitive threats by investing in capital, with the aim to deter their competitors from strengthening their position within their market. Fudenberg and Tirole (1983) argue that capital investment can credibly threaten to lower the profitability of the competitor due to its commitment value to compete aggressively in the future; as a result, it is the most appropriate variable to consider, instead of prices or quantities. Furthermore, the evidence that rivals with low leverage increase investment is consistent with the notion that leverage softens competition (Fudenberg and Tirole 1986; Bolton and Scharfstein 1990).1 More broadly, this evidence suggests that firm reorganizations can be perceived as a threat by a nearby rival, prompting a response. The key empirical challenge is to exclude the possibility that there are aggregate shocks that induce both the restructuring decisions and the rival firms’ subsequent increase in capital investment. To do so, I employ a difference-in-differences (DID) methodology which differences out any aggregate shocks (e.g. demand, technology) at the industry or local level. I start by identifying the announcing firms’ local competitors (treated), namely competitors in the same three-digit SIC industry and UK region. Then, I compare the treated firms to those firms (controls) operating in the same 3-digit SIC industry but located in a different region, or operating in a different three-digit SIC industry and located in the same region. Given that changes in the investment opportunities of an industry would affect all firms in the industry regardless of their location, such changes would be differenced out in the estimation. A similar argument holds for changes in local macroeconomic conditions. 1

A simple model as in Fudenberg and Tirole (1984) or Tirole (1988) predicts that, in order to deter entry or expansion, rivals will increase investment if the investment makes them look “tough” irrespective of whether the competition is in quantities (Cournot) or prices (Bertrand). Indeed, an increase in investment is likely to provide a “tough” investment signal when undertaken by competitors with low leverage, who can credibly commit to compete aggressively in the future.

–2–

I find that competitors in the same industry and region as the restructuring firms increase their capital investment following the announcement relative to control firms. An increase in layoffs from 100 to 200 workers, in the context of restructuring actions indicating better ability to compete, is associated with an increase in capital expenditures of rival firms by approximately 6%. I further document that this effect is driven by the low debt competitors. To the extent that the omitted variables are uncorrelated with leverage, that estimate can be interpreted as a triple-difference effect. These results rely on the implicit assumption that to some extent firms compete locally, which is true in the presence of frictions either in input or goods markets. Naturally, firms compete globally; the results in the paper do not require the complete absence of such global competition, only that frictions are large enough to preclude perfectly integrated input and output markets. Indeed, the firms in our sample are much smaller compared to the typical Compustat firm, with average employment of 147 workers versus 7,000 workers in Compustat.2 In the presence of even moderate frictions, it is reasonable to assume that small firms in the same industry and region share the same customer and supplier base to a non-trivial extent. In fact, this is the premise behind the agglomeration literature: firms tend to be located in the same region when they depend on the same set of suppliers and customers and when they share the same type of workers (Ellison, Glaeser, and Kerr 2010; Ellison and Glaeser 1997).3 This assumption is supported by differences in the treatment effect across subgroups. In particular, I document that the positive effect on investment for 2

The calculation is based on the universe of Compustat.

3

Note that although industry localization is common, industries are rarely extremely concentrated geographically. Duranton and Overman (2005) show that the typical manufacturing industry in the UK is adequately spatially dispersed – which makes my identification possible.

–3–

low debt rivals is weaker for rivals which export a large share of their sales, as these firms are less sensitive to local competition. I further show that the positive treatment effect is stronger for industries which are more concentrated at the regional level, as competing firms are more likely to benefit from responding to a local competitive threat, while overall industry concentration does not seem to be relevant. In all regressions, I control for time-invariant firm characteristics by including firm fixed effects, for time-varying industry characteristics by including (two-digit SIC) industry×year fixed effects, and for time-varying region characteristics by including region×year fixed effects. These controls address concerns that aggregate shocks at the industry or local level are driving the results. However, they do not address the possibility of a shock that is both industry and region-specific; an example of such a shock is a demand shock for a specific good in a specific region. The triple difference specifications, in the subgroup analysis, partially address this concern. I include (two-digit SIC) industry×region×year fixed effects and show that, within a given industry-region-year, low-leverage firms, less export-oriented firms, and firms that operate in more locally concentrated industries respond to a peer firm’s restructuring announcement by increasing their investment more than their counterparts, consistent with the hypothesized economic story of reaction to competitive threat. To the extent that industry-region shocks do not have a differential impact on the two groups of firms, the subsample analysis implicitly controls for the effect of such shocks that does not operate through the mechanism of competitive threat. Moreover, I ensure that there are no pre-treatment trends in the data by performing a dynamic analysis on the treated variable (capital investment): I consider leads and lags of capital investment and find no significant differences between treatment and control firms

–4–

prior to treatment, while differences are significant post-treatment. On the other hand, I find no significant effects on the profitability (ROA) and the cashflow-to-sales ratios before and after the announcement event, contradicting the notion that the observed increases in investment are due to an industry-region shock. To rule out alternative explanations, I create a placebo test using negative restructuring announcements, i.e. announcements which signal bad news for the firm or the industry, such as firm liquidations. If both types of announcements, positive and negative, are driven by industry, region or industry-region shocks and not by rivals’ reaction to competitive threat, one would expect to find investment effects for both announcement types. However, in contrast to the announcements used in the baseline analysis, negative announcements have no investment effect. Furthermore, the absence of investment effects for negative announcements is evidence against alternative explanations for the association between positive announcements and rivals’ investment increases, such as cheaper labor or more abundant desirable capital. This is because those alternative effects should also be present for negative announcements and this is not the case empirically. I also conduct further empirical tests to address the possibility that treated and control firms have systematic differences. I compare the observable characteristics of the treated and the control firms one year before the restructuring announcements and find no statistically significant differences. To account for non-observable differences, I repeat my analysis within the subset of firms which share common ownership (i.e. when treated and control firms both belong to the same group) and find that my empirical results are robust within this subset. Finally, it is worth commenting on the external validity of the results in this paper.

–5–

In the industrial organization literature, the assumption that capital investment can deter competitors from strengthening their position in the market is applicable to physical capital, rather than intangible capital. My results come from a wide cross-section of manufacturing industries, which are precisely the industries in which physical capital is important. Testing the same effects in R&D intensive industries or industries with low physical capital (e.g. services) would not necessarily give similar predictions on capital investment. This paper contributes to an important literature studying firm restructuring decisions. Several studies examine corporate restructuring and its impact on subsequent firm performance and real outcomes (Gibbons and Katz, 1991; Ofek, 1993; Kang and Shivdasani, 1997; Denis and Kruse 2000; Perry and Shivdasani, 2005; Denis and McConnell, 2006). Others focus specifically on firms’ financial restructuring decisions, such as LBOs and leveraged recapitalizations and study their real effects on firms and their competitors (Denis and Denis, 1993, 1995; Denis 1994; Kovenock and Phillips 1995, 1997). This paper adds to this literature by focusing instead on firms’ operational restructuring decisions and their real effects on competitors’ investment responses. The paper also contributes to the literature emphasizing the role of debt in product market competition. Several papers argue that high leverage softens competition (Fudenberg and Tirole, 1986; Bolton and Scharfstein, 1990 on the theoretical side, Phillips, 1995; Chevalier, 1995; Chevalier and Scharfstein, 1996; Zingales, 1998 empirically). Consistent with this literature, I find that leverage is a key determinant of firms’ ability to increase physical capital in order to deter competitors from strengthening their position within the market. More broadly, my findings support the notion, expressed in Hennessy and Whited (2005), that firms may choose to be seemingly under-levered in anticipation of future unan-

–6–

ticipated financing needs, and the argument in Admati, DeMarzo, Hellwig, and Pfleiderer (2015) that firms may hold lower debt initially due to shareholders’ resistance to leverage reduction even in cases when total value maximization predicts the opposite. The paper proceeds as follows. Section II presents the restructuring announcements. Section III describes the empirical strategy. Section IV reports the results. Section V concludes.

II

Restructuring Announcements

I construct a unique dataset of announcements of firms’ decisions to reorganize their operations. All these reorganizations involve layoffs. I hand-collect these announcements for manufacturing firms in the UK over the 2002-2008 period. This information is initially collected by the Industrial Relations Research Unit in the UK and is then provided to the European Union for monitoring purposes.4 The primary sources of these announcements are daily newspapers, the business press, specialized economic press and other online sources, such as firms’ and unions’ websites. The information is formatted in separate standardized fact sheets, which allows the collection of information in a systematic manner. The factsheets cover all announcements of layoff-related restructuring activities in UK manufacturing sites, regardless of the firm listing status (private or public) or its country origin; thus, restructuring announcements regarding the UK manufacturing sites of non-UK firms are included in the sample. An announcement is part of these factsheets if it entails a reduction of at least 100 jobs at the firm level, or if it involves 4

European Union collects this information to monitor the extent of restructuring activities in European countries and their consequences for labor markets. UK is one of the countries with the most intensive restructuring activity among the European countries monitored.

–7–

firms employing more than 250 people and at least 10% of their workforce is affected by the layoffs. The fact sheets include the following information: the name of the announcing company, the date of the announcement, the location of the firm-units restructuring, a description of the industry sector of the restructuring unit and the number of job losses announced. A summary of additional information is also available, which is relevant for understanding the reason for the restructuring. I match the description of the business activity of the restructuring unit to three-digit SIC industry codes (the most granular industry codes available in Amadeus database used for the analysis) and I map the location of the restructuring unit to 12 regions in the UK.5 Thus, each announcement is mapped to a given industry-region. Overall, there are 362 announcements, which are sufficiently dispersed across industries and regions. Table A1 shows the distribution of the announcements by (three-digit SIC) industry. Table A2 shows the distribution of the announcements by region. I parse the structured fact sheets and group the announcements in two categories: those that signal improvement in the competitive position of the announcing unit and those that reveal bad news. I use the former for my baseline analysis and the latter to create placebo tests for the baseline results. I distinguish between the two types of announcements based on the reasons stated in the factsheets. About one third of the announcements are positive, whereas the rest two thirds are negative. For the positive ones, I search for words indicating strengthening of the firm’s market position. Such words include, for example: “improve performance, ensure the achievement 5 The 12 regions are the following: the 9 England Government Office Regions (GORs) established in 1994, as well as Wales, Scotland and Northern Ireland.

–8–

of increased profitability and success in the long term, reduce costs and improve productivity, strengthen its presence, provide our customers with a high quality product at a competitive price, improve competitiveness, improve the efficiency of our operations, sustaining business growth, improve the cost-effectiveness, a better commercial decision, ensure that it is competitive and fit for the future, to deliver value to customers, supply high-quality and competitive products, already invested £2m in the site with another 2.7m investment in production facilities and equipment planned, increase flexibility, a large-scale investment in upgrading, generate annual profits of 3.2 million GBP, increase quality standards”. Two examples of announcements coded as positive are the following: “The company says it plans to invest more than £4 million in the Lackenby site, which employs 517 people, but that modernization will require job losses”, and “it is to close its plant in Wrexham, North Wales with the loss of 109 jobs [...] It plans to transfer the work to a new partner company [...] This approach will enable the company to more quickly and efficiently deliver a broader and deeper range of batteries for hearing impaired consumers”. On the other hand, negative announcements include words suggesting a slump in demand or a decline in the prospects of the firm. Two examples of announcements grouped as negative are the following: “Fast-food restaurant suppliers EBG Bakeries has announced it is to close its Hemel Hempstead plant with the loss of 100 jobs. The decision follows the bakery being severely damaged in a blast at the Buncefield oil depot last December[...]”., and, “the company was in a weak financial position with a lot of debt and the company had also recently been fined £12000 for a fuel leak”. Figure A1 presents the distribution of the number of workers announced to be laid off in the affected industry-region pairs for positive and negative restructuring announcements and shows they are pretty similar. The median

–9–

number of layoffs is 200 workers for announcements that signal strengthening of the firm and 210 workers for those that reveal bad news. Our sample announcements confirm findings in the literature that an increasing number of restructuring actions involving layoffs in the recent decades is not driven by a reduction in demand relative to existing production capacity, but it is rather positive news to the firm initiated in response to an internal corporate governance system to improve cost efficiency and competitiveness of the firm (Farber and Hallock 2009; Hallock, 2009). Indeed, the average stock price reaction to layoff announcements is negative in the 1970s and (weakly) positive in the 2000s (Hallock, 2009). Hallock (2009) argues that the weakly positive effect is explained by the fact that some layoff announcements reveal bad news but others signal a proposal to improve the condition of a firm.6

III

Empirical Strategy

The source of firm-level data is Bureau van Dijk’s Amadeus database. Amadeus provides financial information about public and private firms in the UK and other European countries. I use unconsolidated financial data of manufacturing firms that include domestic and foreign firms. The sample used in the main analysis of the paper covers over 17,000 firmyears, although sample size may vary due to missing information for some of the variables used in the analysis. The baseline sample spans a wide range of manufacturing industries: for example, 7% of the sample covers ‘food and kindred products’ (SIC: 20), 10% of the sample covers ‘printing, 6

Our sample includes mostly private firms as well as foreign firms which precludes a similar analysis.

– 10 –

publishing, and allied industries’ (SIC: 27), 6% of the sample covers ‘chemicals and allied products’ (SIC: 28), 14% of the sample covers ‘fabricated metal products’ (SIC: 34) and 7% of the sample covers ‘electronic and other electrical equipment, exc. computers’ (SIC: 36). The 3-digit SIC industries in the baseline sample are dispersed across the 12 regions in the UK: around 90% of the industries operate in at least 8 regions. Furthermore, as seen in Table A3, two-digit SIC industry code firm-level observations are also adequately dispersed across regions. Such spatial dispersion of industries is necessary for my identification to be meaningful. I examine the effect of the positive announcements on competitors’ investment decisions by employing a difference-in-differences (DID) methodology. In particular, I compare the effect of the announcements on close rivals, i.e. firms in the same (three-digit SIC) industry and located in the same region as the announcing firm (treated firms), with otherwise similar firms in the same industry located in different regions, or firms in the same region active in different industries (control firms). The first category of control firms differences out any changes in the dependent variable that are due to fluctuations in industry conditions, whereas the second category of control firms differences out changes due to regional macroeconomic shocks. A similar empirical methodology has been used in several studies (see for instance Card and Krueger, 1994). The multiple pre-announcement and post-announcement time periods address concerns regarding validity. To illustrate how the empirical design works we can think of the following example. Let t = 0 be the initial period in the sample and consider industry j and two UK regions, A and B. Suppose that a positive announcement (as defined previously) is made at time t = 1 by a firm in industry j located in region A, and

– 11 –

that another positive announcement is made at time t = 2 by another firm in industry j located in region B. From t = 1 to t = 2, the firm in region B belongs to the control group and after that it belongs to the treated group.7 Thus, it is possible that firms in an industry and region serve as both treated and control groups at different points in time. This specification is also robust to some groups not being treated at all, or to other groups being treated since the start date of the sample. To evaluate the effect of the announcements on local competitors, I estimate the following specification: yit = λi + γt + δ · treatedjk,t−1 + β1 · Xit + β2 · Xkt + β3 · Xjk,t + it

(1)

where i denotes a firm, t denotes a year, j is an industry, and k is a region in the UK. In the expression above, λi and γt are firm and year fixed effects respectively, and treatedjk,t−1 is the logarithm of the cumulative number of workers announced to be laid off in a industry j and region k (lagged by one year); treated is zero when no layoffs are announced. Thus, the definition of treated accounts for the magnitude of the event, since it takes into account the number of workers announced to be laid off. The DID estimator coefficient, δ, measures the average within-firm changes in the dependent variable following an announcement, controlling for changes in firms not affected by the announcement. Firm controls Xit include firm age (log-transformed), whereas regional controls Xkt include regional unemployment rates, provided by Eurostat, and measures of regional economic growth, published by the Office for National Statistics in the UK (ONS). Finally, Xjk,t is the number of firms operating in each industry-region (log-transformed), included in order to control for the supply of the different manufacturing activities in each region. Year fixed effects control for aggregate 7

The same intuition generalizes when we consider multiple industries and regions.

– 12 –

fluctuations and firm fixed effects control for time invariant, firm-level differences. In my analysis, I further control for industry-specific shocks by including (two-digit SIC) industry×year fixed effects. Thus, I compare treated and control firms within the same industry. I also control for region-specific shocks by including region×year fixed effects to address concerns that there might be changes at the local level which affect firms’ investment decisions. Therefore, I compare firms in the treated and control group within the same region. Next, I examine the role of leverage as a determinant of competitors’ investment following restructuring announcements. To do so, I sort firms in two groups based on their pre-treatment ratio of total debt net of cash over the book value of assets, with high (low) leverage firms defined as those for which the ratio is above (below) their pre-treatment industry median. I estimate the following specification:

yit = λi + γt + δ1 · treatedjk,t−1 + δ2 · (treatedjk,t−1 × high debt i ) + +β · Xit + β2 · Xkt + β3 · Xjk,t + it ,

(2)

where high debt is a dummy variable which takes the value of 1 if a firm has high debt compared to its industry, and is 0 otherwise. Thus, the main leverage effect is absorbed by the firm fixed effects. All the other variables and subscripts are defined as in the previous specification. In this framework, δ1 and δ2 are the difference-in-difference-in-differences estimators: δ1 measures the average within-firm changes in the dependent variable following announcements for the low debt competitors and δ1 +δ2 measures the same effect for the high debt competitors. This is a triple difference specification which estimates how the effect on investment

– 13 –

depends on leverage using the high-versus-low difference. This specification helps alleviate concerns that local industry shocks (e.g. local demand shocks), contemporaneous to the announcements, are driving the results. To the extent that high-versus-low debt firms do not respond differently, such shocks will be differenced out in this estimation. Importantly, this specification allows me to control for industry/region-specific shocks by including (two-digit SIC) industry×region×year fixed effects, further ensuring that industry/regionspecific shocks are not the drivers of the results. The main identifying assumption is that no omitted variable which predicts assignment into the treated or control group would also predict the outcome variable. Consistent with this, Table 1 displays summary statistics for key financial variables and presents evidence that firms in industry-regions with announcements are very similar to those in different industry-region pairs the year before the announcement. Column 1 reports means and standard deviations of the variables indicating that there is substantial variation in all the important variables. Sample firms are quite small, having an average of £25.3 millions in assets, they are, on average, 26 years old, and they are quite profitable (Return on Assets is 6.7%). Columns 2, 3 and 5 in Table 1 report the firm-level variables one year before the announcements: Column 2 includes firms in industry-regions with announcements, column 3 includes firms in the same industry but in different regions, and column 5 includes firms in the same region operating in different industries. As indicated by the reported p-values, treated (column 2) and control (columns 3 and 5) groups are not statistically different from each other across several observables one year before the announcements take place.

– 14 –

IV IV.1

Results Baseline Investment Results

Figure 1 provides a snapshot of the results. I plot the average demeaned investment (measured as logarithm of one plus capital expenditures) one year before and one year after the restructuring news.8 The graph compares rivals in industry-regions with announcements (treated) with those in the same industry as the treated but located in a different region and those located in the same region but active in a different industry (controls). Treated firms sharply increase their investment following the announcement, and that increase is statistically significant. However, there is no statistically significant change in investment for control firms. For a more formal analysis, I next turn to a multivariate regression setup. Table 2 estimates Equation (1) and presents the baseline results. In columns 1-4, investment is measured as the logarithm of one plus capital expenditures. Column 1 includes firm and year fixed effects but does not include any other controls. Column 2 additionally includes a number of time-varying controls: firm age, the number of firms in the industryregion, and measures of local economic conditions (the regional unemployment rate and the regional GDP growth rate). Column 3 adds industry×year fixed effects to the previous specification to control for changes in investment opportunities at the industry level, and column 4 adds region×year fixed effects to control for regional shocks. Note that region×year fixed effects absorb the macroeconomic controls included in columns 2 and 3. The coefficients on the announcement variable are of similar magnitudes across all four specifications. The fact that the additional controls have little impact on the results indi8

The demeaning is done relative to each 3-digit SIC industry.

– 15 –

cates that our estimates are not driven by differential industry or regional trends between the two groups. The coefficients are statistically significant at the 1% level across specifications. They are also economically significant: an increase in layoffs from 100 to 200 workers is associated with an increase in capital expenditures of approximately 6% (column 4). Columns 5-8 repeat the specifications in columns 1-4 of Table 2 defining rivals’ investment as the ratio of capital expenditures divided by the book value of assets. Again, the announcement variable coefficients are positive and statistically significant across specifications. The results are also economically meaningful: an increase in layoffs from 100 to 200 workers is associated with an increase in capital expenditures over assets by approximately 4% relative to the mean (column 8). My findings suggest that firms respond to an expected increase in competition intensity by increasing their physical capital. Those investments credibly threaten to lower the profitability of the announcing firm in the future, as investment in capital is irreversible and, therefore, signals the commitment of rival firms to aggressively compete in the future. As a result, investing in capital may deter competitors from further strengthening their position within their market. Fudenberg and Tirole (1983) further argue that, unlike capital, prices or quantities do not entail commitment value, as firms cannot credibly commit not to change prices or quantities in the future.

IV.2

The role of leverage

Next, I examine the role of leverage in rivals’ investment response to the news that the restructuring firm may be strengthening its competitive position. Caves and Porter (1977) note that investment can be risky and can dampen short-run profits. The same view is

– 16 –

shared by Gilbert and Lieberman (1987) and Porter and Spence (1982), who claim that firm characteristics such as financial constraints can lead to asymmetric responses to competitive threats by firms. Tirole (1998) argues that rivals increase investment as a response to a competitive threat if they can send a “tough” signal to compete aggressively in the future and this holds irrespective of whether the competition is in quantities (Cournot) or prices (Bertrand). I proxy for this notion of “tough” competition by focusing on low debt firms, as these are the firms which can credibly commit to compete aggressively. I sort firms in high and low debt based on their pre-treatment level of leverage, measured as debt net of cash over the book value of assets, as compared to the industry median. In Table 3, I show that high debt firms share several characteristics with financially constrained firms. As expected, low debt firms have much higher interest coverage ratios compared to high debt firms and can thus more easily service their interest payments. Also, low debt firms are more productive as compared to high debt firms – they have higher return on assets, profit margins, and lower production costs. On the contrary, low debt firms do not seem to be significantly different from high debt firms in terms of their age, their number of employees, or how fast they grow (as measured by assets, sales, or employees). Figure 2 provides a graphical snapshot of the results by plotting the Epanechnikov kernel densities for rival firms’ capital expenditure before and after the announcement dates; both panels consider treated firms, i.e. rivals in the same industry and region as the announcing firm, with Panel A referring to low debt treated firms and Panel B to high debt ones. The plots suggest that there is a rightward shift in the kernel density following the restructuring news for the low debt (Panel A) but not the high debt (Panel B) treated firms. The observed shift in the density of the low debt firms is statistically significant

– 17 –

and the Kolmogorov-Smirnov test of the equality of distribution functions is rejected at the 1% level. On the contrary, the Kolmogorov-Smirnov test cannot reject the equality of distributions for the high debt firms in Panel B (p-value: 0.72), suggesting that high debt rivals do not respond to the competitive threat posed by the announcing firm. Thus, the rival firm investment response appears to be concentrated among the low debt rivals. In order to contrast low debt treated firms to low debt control firms, Figure 3 presents the capital expenditure kernel densities for the two kinds of control firms, conditional on having low debt: Panel A refers to the density of low debt rivals in the same industry, but different region than the announcing firm, whereas Panel B refers to different industry, same region low debt rivals. Both graphs show that the distributions do not appear to be different before and after the announcements. The Kolmogorov-Smirnov test cannot reject the equality of distributions in both Panels (p-value: 0.60 in Panel A and p-value: 0.30 in Panel B), confirming the baseline result that control firms do not appear to react to restructuring announcements. To formally test the impact of leverage, we turn to regression analysis. Table 4 augments the baseline specification in Table 2 with interaction terms of the treated variable and a firm-level dummy (high debt i ), which sorts firms in two groups depending on their pre-treatment level of leverage. To the extent that the omitted variables are uncorrelated with levels of leverage, the estimate can be interpreted as a triple-difference effect. As in the baseline regressions, the effect of treatment is positive and significant. Furthermore, the interaction coefficient is negative and statistically significant at the 5% level, indicating that the rival response is driven by low debt, rather than high debt, rivals. These results are robust to different definitions of the dependent variable and to the inclusion of vari-

– 18 –

ous controls. Most importantly, the results are robust to including industry×region×year fixed effects which control for industry-region specific shocks.9 As it can be seen, low debt competitors increase their investment following the announcements (δ1 ), while high debt competitors do not (δ1 +δ2 ). In fact, the magnitudes of the DID coefficient roughly double when we focus on low debt competitors, compared to the full sample coefficients reported in Table 2. This finding is consistent with a large literature which establishes that high leverage softens competition (Fudenberg and Tirole (1986) and Bolton and Scharfstein (1990) as regards theoretical models, Phillips (1995), Chevalier (1995), Chevalier and Scharfstein (1996), Kovenock and Phillips (1997) and Zingales (1998) as regards empirical findings). According to Table 3, high leverage firms are different from low leverage firms across several characteristics in ways that resemble financial constraints. However, it could be that these differences are not independent from leverage and are instead confounding the results. I address such concerns in Table 5. First, in column 1, I repeat column 5, Table 4, adding differential leverage trends as a control. To this end, I interact the high debt i dummy with year fixed effects. This test allows to flexibly control for mean-reversion of capital expenditures to a firm-specific equilibrium level and to control for the fact that differential leverage levels may be endogenously correlated with firm investment decisions. The results are very similar, indicating that mean-reversion or differential trends based on pre-treatment leverage are not driving the findings. Second, in columns 2-5, I run horseraces between treatment and high debt i and treatment and other firm-level variables (also defined pre-treatment) that seem to be signif9

The fixed effects do not absorb the treatment indicator as industry in the fixed effects is defined based ont two-digit SIC industries, while the treatment indicator is defined based on three-digit SIC industries.

– 19 –

icantly different between high and low leverage firms in Table 3. Thus, in column 2, I interact treatment with profit margin i ; in column 3, I interact treatment with ROAi ; in column 4, I interact treatment with COGS/Sales i . These results address concerns that investment may be higher for low leverage treated firms not because they are less financially constraint but because they are more profitable or more cost effective. All three interactions are not statistically or economically significant. On the contrary, the interaction with leverage remains statistically significant and similar in magnitudes to column 1. Columns 5-8, Table 5, repeat the same specifications using capital expenditures normalized by assets as the dependent variable and yield similar results. Throughout this table, I control for firm, industry×region×year fixed effects, and leverage differential trends.

IV.3

Investment and Local Competition

The interpretation of our results relies on the implicit assumption that firms to some extent compete locally, which is true in the presence of frictions which preclude perfectly integrated input or goods markets. In the presence of such frictions, it is reasonable to assume that small firms in the same industry and region are more likely to share the same customer and supplier base. In fact, this has been well documented in the agglomeration literature: firms which share the same customer and supplier base and the same labor pool tend to co-locate (Ellison, Glaeser, and Kerr, 2010; Ellison and Glaeser, 1997). Naturally, firms also compete on the global markets; our results do not require absence of such competition. The assumption of local competition is very likely to be true for the sample used in this study: note that the firms in my sample are private and mostly small, with 147 employees on average, compared to the large firms in the Compustat sample (7,000 employees on

– 20 –

average). Next, I present evidence that further supports this assumption. In particular, I focus on low debt rivals, as they are the ones that respond to announcements, and repeat the baseline empirical analysis sorting firms into subgroups that are likely to be differentially sensitive to local competition. I show that the rivals’ investment response differs across subgroups in a way that is consistent with the assumption that higher sensitivity to local competition is positively associated with a sharper investment response. First, I show that the announcement treatment effect is weaker when firms export a large share of their sales. This is consistent with the implicit assumption that firms which export a large share of their sales are not very sensitive to local competition and, therefore, do not respond sharply to local competitive threats. In particular, I sort firms on the pre-treatment ratio of export to sales and construct a dummy variable which takes a value of 1 if the export share is higher than the industry median and 0 otherwise.10 The results are reported in columns 1 and 4 of Table 6. The interaction coefficient is negative and statistically significant and robust to including industry×region×year fixed effects, in line with the local competition hypothesis.11 Then, I sort firms based on the level of industry competition at the regional level; if the local competition hypothesis is true, then firms in more regionally concentrated industries will respond more aggressively to local restructuring announcements, as they are more sensitive to local competition. To test this hypothesis, I construct the Herfindahl index for a given industry-region using information on firms’ sales and I interact the pre10

When export data is missing, I consider that the ratio of export to sales is 0. Results are robust to dropping observations with missing export data. 11

Given that the local unit is the UK region, ideally I would like to observe information on firms’ sales within and outside of the UK region in which they are located. However, this information is not available.

– 21 –

treatment value of this indicator with treated (columns 2 and 5). I find that the effect on investment is more pronounced in more concentrated industries at the regional level, as in concentrated markets “local” competitors are more likely to recover some of the short-run costs of their investments in the long-run (Kovenock and Phillips, 1997; and Zingales, 1998). For comparison, I then repeat the analysis defining the Herfindahl Index at the industry level instead: now, the Index measures concentration in a given industry, not industryregion. The results are reported in columns 3 and 6: this time, the interaction coefficient is not significant, indicating that competition at the industry level has no differential effect on local competitors’ investment. Thus, it is sensitivity to local, not nation-wide, competition that matters, consistent with the presence of local competition effects.

IV.4

Robustness

My identification strategy differences out industry-wide or region-wide shocks, as I compare firms in the same industry located in different regions, or firms in the same region operating in different industries. However, there is a concern that the findings may be driven by shocks specific to an industry and region; for example, a demand shock for a specific good in a given region may lead to both the restructuring announcements and the subsequent increase in rivals’ investment. To partially address this concern, I use triple-difference estimations and find that, within an industry-region, low-leverage firms, less export-oriented firms, or firms that operate in more regionally concentrated industries have sharper investment reactions to restructuring announcements, consistent with the hypothesis of reaction to competitive threat. Notably, the estimated coefficients remain virtually unchanged (both statistically and economically) after saturating the specification with numerous controls, including (two-digit SIC) industry×region×year fixed effects which absorb time-varying

– 22 –

industry-local shocks. In this section, I provide a series of robustness tests that further address the concern that the rival firms’ investment reaction to restructuring announcements is driven by local industry shocks. First, I examine firms’ capital investment and find no pre-trends in the data. I also show that variables likely to be affected by local demand shocks, such as firm profitability and cashflow-to-sales ratios, exhibit no changes around the announcement dates. Second, I consider a placebo test using negative restructuring announcements, i.e. announcements that reveal bad news for the firm or the industry. If both positive and negative announcements are driven by industry-region shocks, one would expect to find symmetric investment effects, but I show that this is not the case empirically, as negative announcements do not generate any investment reactions. Finally, I repeat the baseline analysis in a small subset of firms sharing common ownership. This test further mitigates concerns that unobservable systematic differences in treated and control samples are driving the results.

IV.4.1

Dynamic Analysis

Table 7 presents a dynamic analysis around the restructuring announcements for the baseline sample. I look at the effect on capital investment before and after the treatment events: if unobserved industry-region shocks cause both the restructuring announcements and rivals to increase investment, then one should expect to see increases in investment before the arrival of the restructuring news. Contrary to that hypothesis, Table 7 shows no significance pre-treatment, while capital investment is positive and significant following the announcement. The results are robust to a variety of controls, as in Table 2. In unreported

– 23 –

regressions, I find similar results when I examine only low debt rivals, while there is no significance for high debt rivals around the events. I repeat the same analysis for other measures in Table 8 to provide further evidence consistent with my channel and not consistent with the interpretation of unobserved regional industry shocks driving the results. Columns 1-3, Table 8, show that competitors’ assets grow following the restructuring events, while there is no significance pre-treatment. This finding supports the argument that competitors expand by investing and purchasing assets following the restructuring news responding to restructuring firms taking actions to improve their competitiveness. On the contrary, I capture no effect on measures of operating performance likely to be affected by economic shocks. In particular, I examine return on assets to measure firm operating performance (columns 4-6) and operating cashflows normalized by sales (columns 7-9). Across all six specifications, I do not find any significance in the data around treatment, contradicting the hypothesis that unobserved regional industry shocks can explain the results.

IV.4.2

Falsification Test

While the restructuring announcements in the baseline analysis signal a strengthening of the competitive position of the restructuring firm, there are cases in which layoff announcements reveal bad news for the announcing firm, such as a decline in demand. Here, I focus on the set of negative restructuring announcements and do a placebo test, finding no differential investment response between treatment and control firms. To the extent that positive and negative announcements are associated with local industry shocks with similar investment effects, this test adds to the evidence that region-industry shocks do not seem to be the

– 24 –

driver of our empirical findings. The findings of the placebo test are presented in Table 9, which replicates the results in Table 4 using the negative, rather that the positive, restructuring announcements. Notably, the distribution of the number of workers announced to be laid off in each type of restructuring announcement is similar across the two types (Figure A1), alleviating the concerns that the two types of announcements refer to different restructuring magnitudes. Across specifications, neither the level nor the interaction treatment effect on investment is statistically or economically significant, indicating that negative restructuring announcements do not affect local rivals’ investment levels. The placebo test also helps to address two alternative explanations for our results. The first alternative story is that the documented increase in investment may be due to a labor market effect, and not a direct response to a competitive threat: layoffs due to the restructuring event may cause a reduction of local wages and, to the extent that labor and capital are complements, rivals may increase capital along with labor.12 Since the two types of restructuring announcements should have the same impact on labor markets, they should also have the same effect on rivals’ investment if the investment effect is solely through the labor market channel. However, this is not the case empirically. The second alternative story is that the increase in rivals’ investment may be attributed 12

Indeed, in Appendix B, I provide evidence that layoff announcements have a negative effect on average wage per employee (Table B1, Panel A), which is true for both low debt and high debt competitors (Table B1, Panel B). Interestingly, I find that the mechanism that explains the lower wages per employee differs for the two groups of firms. Low debt firms hire more workers following the layoff announcements, suggesting lower entry wages, while high debt firms do not increase employment, and thus seem to be renegotiating down current employee wages (Tables B1, Panel C). Layoff announcements lower the outside option of workers in the affected region, negatively impacting wages. Table B2 in the Appendix further supports this explanation, showing that the negative effect on wages is weaker for industries characterized by higher unionization rates, as the bargaining power of workers is higher.

– 25 –

to desirable capital being available because of the restructuring: local rivals may be exploiting the opportunity to grow relatively cheaply or get high-synergy plants and equipment by acquiring assets from the restructuring firms. However, this effect should be present in the case of both positive and negative announcements, so the placebo test results provide evidence against this explanation.

IV.4.3

Firms with Common Ownership

I collect data on firms’ ownership from the Amadeus database. Using that information, I focus on groups that share the same owners and have operations in industry-regions with restructuring announcements (treated) as well as in industry-regions with no announcement (control). Only a small subset of firms in my sample satisfy those matching criteria; nevertheless, despite the low statistical power, there is value in redoing the analysis using this small subset, as it controls for unobservable group-specific (or ownership-specific) differences between treated and control firms. Figure 4 presents the Epanechnikov kernel densities for capital expenditures for firms with common ownership before and after the restructuring announcements. Panel A refers to treated firms, while Panel B to control firms. There is a rightward (statistically significant) shift in the kernel density following the announcements in Panel A, but there is no such shift in Panel B. Therefore, among firms with common ownership, I find an increase in investment only for firms in the same industry-region as the announcing firms. Table 10 repeats the baseline regressions for the common-ownership sample. The treatment effects are bigger than in Table 2 and, mostly statistically significant, albeit the lower statistical power. These results provide additional support for my hypothesis.

– 26 –

IV.4.4

Alternative Definitions of Treatment

Table 11 repeats the baseline analysis in Table 2 defining the treatment indicator treated in two alternative ways. In Panel A, the treatment indicator takes a value of 1 following an announcement in an industry-region. Capital expenditures for treated firms increase by 25% following the news (column 4), and the ratio of capital expenditure over assets increases by 20% relative to the mean (column 8). In Panel B, the treatment indicator is the ratio of the cumulative number of workers announced to be laid off in an industry-region divided by the total number of workers in the industry-region (computed in-sample). This measure proxies for the magnitude of the event relative to the size of the local market. A restructuring announcement involving 30% of workers laid off in an industry-region is associated with a 9% increase in capital expenditures by local competitors (column 4) or with a 11% increase in the ratio of capital expenditures over assets relative to the mean (column 8).

V

Conclusion

This paper contributes to the strand of the financial economics literature that explores the drivers and outcomes of firms’ restructuring decisions. While the extant literature has focused on the effect of the restructuring firms themselves (Kang and Shivdasani 1997; Denis and Kruse 2000; Denis and McConnell, 2006), little is known about the effect on their competitors. This paper focuses on restructuring announcements that involve layoffs and enhance the announcing firms’ competitive position and shows that such announcements prompt investment by local competitors, consistent with the notion that competitors commit to compete aggressively in the future (Fudenberg and Tirole, 1983). This paper also

– 27 –

highlights the role of debt in determining competitors’ responses, as it is only the low debt firms which respond by investing (Bolton and Scharfstein, 1990). My empirical analysis excludes the possibility that aggregate industry or regional shocks are driving the results. A variety of additional robustness checks mitigate concerns that industry-region shocks can explain the empirical findings. My findings may have implications for regional economic policy, as they suggest that firms’ restructuring decisions have first-order effects on the investment decisions of local rival firms. They also imply that firms’ restructuring decisions matter for industry dynamics, an aspect that has yet to be fully explored in the literature.

– 28 –

References [1] Admati, A. R., P. M. DeMarzo, M. F. Hellwig, and P. Pleiderer, 2015,“The Leverage Ratchet Effect,” Working Paper. [2] Bolton, P., and D. E. Scharfstein, 1990, “A Theory of Predation Based on Agency Problems in Financial Contracting,” Americal Economic Review, 80, 93-106. [3] Card, D., and A. B. Krueger, 1994, “Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania,” American Economic Review, 84, 772-793. [4] Caves, R.E., and M.E. Porter, 1977, “From Entry Barriers to Mobility Barriers: Conjectural Decisions and Contrived Deterrence to New Competition,” Quarterly Journal of Economics, 91, 241-262. [5] Chevalier, J., 1995, “Capital Structure and Product Market Competition: Empirical Evidence from the Supermarket Industry,” American Economic Review, 85, 206-256. [6] Chevalier, J., and D. Scharfstein, 1996,“Capital-Market Imperfections and Countercyclical Markups: Theory and Evidence,” American Economic Review, 86, 703-725. [7] Denis, D. J., 1994, “Organizational Form and the Consequences of Highly Leveraged Transactions: Kroger’s Recapitalization and Safeway’s LBO,” Journal of Financial Economics, 36, 193-224. [8] Denis, D. J., and D. Denis, 1993, “Managerial Discretion, Organizational Structure, and Corporate Performance: A Study of Leveraged Recapitalizations,” Journal of Accounting and Economics, 37, 209-236. [9] Denis, D. J., and D. Denis, 1995, “Causes of Financial Distress Following Leveraged Recapitalizations,” Journal Financial Economics, 16, 129-158. [10] Denis, D. J., and T. Kruse, 2000, “Managerial Discipline and Corporate Restructuring Following Performance Declines,” Journal Financial Economics, 55, 391-424. [11] Denis, D. J., and J. J. McConnell, 2006, “Introduction to Corporate Restructuring,” in Corporate Restructuring, Edward Elgar Publishing. [12] Duranton G., and H. G. Overman, 2005, “Testing for Localization Using MicroGeographic Data,” Review of Economic Studies, 72, 1077-1106. [13] Ellison, G., and E. L. Glaeser, 1997, “Geographic Concentration in U.S. Manufacturing Industries: A Dartboard Approach,” Journal of Political Economy, 105, 889-927. [14] Ellison, G., E. L. Glaeser, and W. R. Kerr, 2010, “What Causes Industry Agglomeration? Evidence from Coagglomeration Patterns,” American Economic Review, 100, 1195-1213.

– 29 –

[15] Fudenberg, D., J. Tirole, 1983, “Capital as a Commitment: Strategic Investment to Deter Mobility,” Journal of Economic Theory, 31, 227-250. [16] Fudenberg, D., and J. Tirole, 1984, “The Fat-Cat Effect, the Puppy-Dog Ploy, and the Lean and Hungry Look,” American Economic Review: Papers and Proceedings, 74, 361-366. [17] Fudenberg, D., and J. Tirole, 1986, “A ‘Signal-Jamming’ Theory of Predation,” Rand Journal of Economics, 17, 366-376. [18] Gibbons, R., and L. Katz, 1991, “Layoffs and Lemons,,” Journal of Labor Economics, 9, 351-380. [19] Gilbert, R. J., and M. B. Lieberman, 1987, “Investment and Coordination in Oligopolistic Industries,” The Rand Journal of Economics, 18, 17-33. [20] Farber, H. S., and K. F. Hallock, 2009, “The Changing Relationship between Job Loss Announcements and Stock Prices: 1970-1999,” Labour Economics, 16, 1-11. [21] Hallock, K. F., 2009, “Job Loss and the Fraying of the Implicit Employment Contract,” Journal of Economic Perspectives, 23, 69-93. [22] Hennessy, C. A., and T. M. Whited, 2005, “Debt Dynamics,” Journal of Finance, 9, 1129-1165. [23] Kang, J., and A. Shivdasani, 1997, “Corporate Restructuring during Performance Declines in Japan”, Journal of Financial Economics, 46, 29-65. [24] Kovenock, D., and G. M. Phillips, 1995, “Capital Structure and Product Market Rivalry: How Do We Reconcile Theory and Evidence?,” The American Economic Review, 85, 403-408. [25] Kovenock, D., and G. M. Phillips, 1997, “Capital Structure and Product Market Behavior: An Examination of Plant Exit and Investment Decisions,” Review of Financial Studies, 10, 767-803. [26] Ofek, E., 1993, “Capital Structure and Firm Response to Poor Performance,” Journal of Financial Economics, 34, 3-30. [27] Perry, T., and A. Shivdasani, 2005, “Do Boards Affect Performance? Evidence from Corporate Restructuring,” Journal of Business, 78, 1403-1431. [28] Phillips, G. M., 1995, “Increased Debt and Industry Product Markets: An Empirical Analysis,” Journal of Financial Economics, 37, 189-238. [29] Porter, M. E., and M. A. Spence, 1982, The Economics of Information and Uncertainty, John McCall, ed. University of Chicago Press, Chapter 8, 259-316. [30] Tirole, J., 1988, Theory of industrial organization, MIT Press. [31] Zingales, L., 1998, “Survival of the Fittest or the Fattest? Exit and Financing in the Trucking Industry,” Journal of Finance, 53, 905-938.

– 30 –

Figure 1: Positive announcements and competitors’ investment The figure plots demeaned average values of firms’ capital expenditures (log-transformed) the year before and after the announcements of restructuring news which signal the improvement of the competitive position of the announcing firm. The figure plots three different sets of firms: (i) local competitors, i.e., firms in the same (3-digit SIC) industry and region as the announcing firms, (ii) firms in the same (3-digit SIC) industry located in different regions, and (iii) firms in the same region but operating in different (3-digit SIC) industries. The demeaning is done relative to each 3-digit SIC industry. The figure reports t-statistics for the difference between firms’ investment before and after the announcements. Standard errors are corrected to take into account the lack of independence of observations.

0.4

Demeaned investment

0.3

Before

After

0.2 0.1 0

-0.1

t-stat: 2.33

t-stat: 1.39

t-stat: -0.91

-0.2 -0.3

same industry, same region

same industry, different region

– 31 –

same region, different industry

Figure 2: Positive announcements, competitors’ investment and leverage The figure plots the distribution functions of capital expenditures (log-transformed) for treated firms before and after the announcements. Panel A presents investment distributions for low debt firms and Panel B for high debt firms. I sort firms in low and high debt based on their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to their industry median. The Kolmogorov-Smirnov test of equality of distributions functions is rejected at the 1% level in Panel A, but cannot be rejected in Panel B (p-value: 0.72).

.15 .1

after before

0

.05

Epanechnikov Kernel Density

.2

Panel A: Low-debt treated

5

10

15

20

Competitors' investment kernel = epanechnikov, bandwidth = 0.8000

.15 .05

.1

after before

0

Epanechnikov Kernel Density

.2

.25

Panel B: High-debt treated

5

10

15

Competitors' investment kernel = epanechnikov, bandwidth = 0.8000

– 32 –

20

Figure 3: Positive announcements, competitors’ investment and leverage The figure plots the distribution functions of capital expenditures (log-transformed) for low debt control firms before and after the announcement. Panel A includes firms operating in the same industries as treated but located in different regions. Panel B includes firms located in the same regions as treated but operating in different industries. I sort firms in low and high debt based on their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to their industry median. The Kolmogorov-Smirnov test of equality of distributions functions cannot be rejected (p-value: 0.60, Panel A and p-value:0.30, Panel B).

Panel A: Low-debt controls

.2 .15 .05

.1

after before

0

Epanechnikov Kernel Density

(same industry, different region )

5

10

15

20

Firms' investment kernel = epanechnikov, bandwidth = 0.8000

Panel B: Low-debt controls

.2 .15 .05

.1

after before

0

Epanechnikov Kernel Density

(same region, different industry)

5

10

15

Firms' investment kernel = epanechnikov, bandwidth = 0.8000

– 33 –

20

Figure 4: Positive announcements and competitors’ investment for firms with common ownership The figure plots the distribution functions of capital expenditures (log-transformed) for firms before and after the announcement. This analysis includes only treated firms (Panel A) which share common ownership with control firms (Panel B). The Kolmogorov-Smirnov test of equality of distributions functions is rejected at the 10% level in Panel A, but cannot be rejected in Panel B (p-value: 0.24).

.15 .05

.1

after before

0

Epanechnikov Kernel Density

Panel A: Treated

5

10

15

20

Competitors' investment kernel = epanechnikov, bandwidth = 1.0000

.15 .05

.1

after before

0

Epanechnikov Kernel Density

.2

Panel B: Control

5

10

15

Firms' investment kernel = epanechnikov, bandwidth = 1.0000

– 34 –

20

Table 1: Summary statistics This table reports summary statistics for key firm-level variables. Summary statistics are computed for the sample used in the baseline analysis which covers over 17,000 firm-years. Column 1 reports the mean and standard deviation (in parentheses) for the full sample. Columns 2-6 report summary statistics one year before the restructuring announcements. Column 2 includes firms in treated industry-regions and columns 3 and 5 include firms in control industry-regions. Column 3 includes firms in the same industries as the announcing firms but located in different regions and column 5 includes firms located in the same regions as the announcing firms but operating in different industries. Means and standard deviations (in parentheses) are reported. Column 4 reports the p-value of a t-test for the difference between firms in column 2, and 3. Column 6 reports the p-values of a t-test for the difference between firms in columns 2 and 5. p-values are adjusted to take into account the lack of independence of the observations. Age is defined as the difference between each year and the year of establishment. I match firms into 3-digit SIC industries and 12 regions in the UK. The sample covers years 2001-2008.

Log(1+Capex)

Capex/Assets

Net Debt/Assets

ROA

Asset Growth

Sales Growth

Cashflows/Sales

Log(Age)

Mean

Same industry-region

Same industry

p-value of difference (2-3)

Same region

p-value of difference (2-5)

(1)

(2)

(3)

(4)

(5)

(6)

11.704

11.755

11.571

0.37

11.634

0.53

(1.938)

(1.992)

(1.988)

0.054

0.046

0.051

(0.071)

(0.060)

(0.070)

0.135

0.118

0.138

(0.280)

(0.285)

(0.300)

0.067

0.046

0.038

(0.138)

(0.162)

(0.159)

0.101

0.102

0.089

(0.290)

(0.330)

(0.277)

0.071

0.077

0.072

(0.236)

(0.269)

(0.236)

0.059

0.063

0.057

(0.099)

(0.010)

(0.099)

3.018

2.954

2.937

(0.831)

(0.100)

(0.953)

– 35 –

(1.935)

0.39

0.054

0.13

(0.069)

0.14

0.125

0.67

(0.297)

0.43

0.046

0.99

(0.154)

0.44

0.094

0.37

(0.278)

0.77

0.070

0.70

(0.227)

0.34

0.059

0.52

(0.095)

0.75

2.886 (0.941)

0.27

Table 2: Positive announcements and competitors’ investment This table reports results of regressions of competitors’ investment decisions following the restructuring announcements in a given industryregion as compared to a set of controls in different industry-regions with no announcements. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. All columns include firm and year fixed effects. Columns 3 and 7 also include (two-digit SIC) industry×year fixed effects and columns 4 and 8 add region×year fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

(5)

Log(1+Capex)

– 36 –

treatedt−1

(6)

(7)

(8)

Capex/Assets

0.0428

0.0461

0.0527

0.0572

0.0020

0.0021

0.0025

0.0022

(0.0083)***

(0.0080)***

(0.0064)***

(0.0125)***

(0.0008)**

(0.0008)**

(0.0007)***

(0.0010)**

0.226

0.204

0.189

-0.0148

-0.0147

-0.0150

Age

(0.191)

(0.176)

(0.172)

(0.0055)**

(0.0055)**

(0.0053)**

No of firms

-0.0749

-0.131

-0.121

-0.0063

-0.0084

-0.0068

(0.155)

(0.159)

(0.160)

(0.0090)

(0.0095)

(0.0096)

Unemployment rate

-0.0918

-0.1040

-0.0023

-0.0029

(0.0387)**

(0.0415)**

(0.0018)

(0.0018)

-3.438

-5.766

-0.123

-0.209

(3.571)

(3.350)

(0.155)

(0.155)

% GDP growth

Adj. R2 Obs.

0.64

0.64

0.64

0.67

0.45

0.45

0.46

0.46

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Firm FE

Yes

Yes

Year FE

Yes

Yes

Industry×Year FE Region×Year FE

Yes

Yes Yes

Yes

Yes Yes

Table 3: Leverage and other firm-level variables This table sorts firms in high and low debt firms and presents summary statistics across several observable characteristics computed pretreatment. I sort firms in high and low debt based on their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to the pre-treatment industry median. The last column reports the p-value of the difference-in-means test comparing both groups. Profit margin is defined as the ratio of ebitda over sales. COGS/Sales is cost of goods sold normalized by sales. Age is defined as the difference between each year and the year of establishment. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. All variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

Low Debt

High Debt

– 37 –

Mean

Median

St. Dev.

Mean

Median

St. Dev.

p-value

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Interest Coverage Ratio

51.24

13.81

95.24

15.97

3.18

54.50

0.00***

Log(Age)

3.02

3.04

0.819

3.00

3.00

0.864

0.59

No. Employees

168

73

406

181

91

364

0.36

Sales Growth

0.099

0.069

0.222

0.104

0.070

0.246

0.65

Asset Growth

0.157

0.097

0.285

0.159

0.106

0.284

0.80

Employee Growth

0.032

0.016

0.135

0.034

0.018

0.149

0.75

Profit Margin

9.76

8.93

12.68

6.37

6.33

12.19

0.00***

ROA

0.102

0.094

0.126

0.047

0.051

0.128

0.00***

GOGS/Sales

0.654

0.676

0.168

0.702

0.716

0.147

0.00***

Table 4: Positive announcements, competitors’ investment and leverage This table reports results of regressions of competitors’ investment decisions and how these decisions interact with leverage following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. high debt is a dummy which takes the value of 1 if a firm has high debt and the value of 0 if a firm has low debt. I sort firms in high and low debt based on their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to the pre-treatment industry median. The level effect of debt is absorbed by firm fixed effects and, thus, not reported. All columns include firm and year fixed effects. Columns 3 and 8 also include (two-digit SIC) industry×year fixed effects and Columns 4 and 9 add region×year fixed effects. Columns 5 and 10 include (two-digit SIC) industry×region×year fixed effects in addition to the firm fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Log(1+Capex)

– 38 –

treatedt−1

treatedt−1 ×high debt

(8)

(9)

(10)

Capex/Assets

0.116

0.120

0.123

0.124

0.124

0.0049

0.0050

0.0053

0.0051

0.0041

(0.0303)***

(0.0293)***

(0.0269)***

(0.0303)***

(0.0358)***

(0.0014)***

(0.0014)***

(0.0014)***

(0.0017)***

(0.0020)*

-0.131

-0.131

-0.125

-0.121

-0.124

-0.0056

-0.0054

-0.0054

-0.0054

-0.0057

(0.0509)**

(0.0507)**

(0.0512)**

(0.0517)**

(0.0526)**

(0.0021)**

(0.0021)**

(0.0022)**

(0.0023)**

(0.0025)**

X

X

X

X

X

X

X

X

0.64

0.64

0.64

0.65

0.68

0.44

0.45

0.45

0.46

0.51

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Controls Adj. R2 Obs.

Industry×Year FE Region×Year FE Industry×Region×Year FE

Yes

Yes Yes

Yes

Table 5: Leverage and robustness to alternative channels This table reports results of regressions of competitors’ investment decisions and how these decisions interact with leverage following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. high debt is a dummy which takes the value of 1 if a firm has high debt and the value of 0 if a firm has low debt. I sort firms in high and low debt based on their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to the industry median. profit margin, ROA, COGS/Sales are defined in a similar way. The level effect of debt, profit margin, ROA, COGS/Sales is absorbed by firm fixed effects and, thus, not reported. All columns include firm and year fixed effects and (two-digit SIC) industry×region×year fixed effects. They also include leverage differential trends by interacting high debt and year fixed effects. Controls include firm age (log-transformed) and the number of firms in an industry-region (log-transformed).*, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008. (1)

(2)

(3)

(4)

(5)

Log(1+Capex)

– 39 –

treatedt−1

treatedt−1 ×high debt

(8)

Capex/Assets

0.115

0.122

0.114

0.148

0.0037

0.0036

0.0031

0.0045

(0.0395)***

(0.0268)***

(0.0758)*

(0.0021)*

(0.0019)*

(0.0012)**

(0.0034)

-0.108

-0.142

-0.129

-0.169

-0.0050

-0.0059

-0.0045

-0.0063

(0.0505)**

(0.0739)*

(0.0408)***

(0.0799)**

(0.0026)*

(0.0028)**

(0.0018)**

(0.0028)**

0.0238

-0.0006

(0.0172)

(0.0013)

treatedt−1 ×ROA

-0.0032

-0.0011

(0.0307)

(0.0022)

treatedt−1 ×COGS/Sales

Adj. R2

(7)

(0.0343)***

treatedt−1 ×profit margin

Controls

(6)

X

X

X

0.0058

0.00001

(0.0388)

(0.0021)

X

X

X

X

X

0.68

0.71

0.69

0.71

0.51

0.53

0.51

0.53

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Industry×Region×Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

high debt×Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Obs.

Table 6: Positive announcements, competitors’ investment and local competition This table reports results of regressions of competitors’ investment decisions and how these decisions interact with a proxy for firms’ exports, a proxy for competition at the industry-region level, and a proxy for competition at the industry level following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The regressions are similar to those reported in Table 4 but the sample includes low debt firms. export, in columns 1, 4, is a dummy which takes the value of 1 if the pre-treatment percentage of sales that firms export is higher compared to the pre-treatment industry median, and 0 otherwise. industry-region HHI in columns 2, 5 is a Herfindahl Index at the industry-region level based on pre-treatment firms’ sales. industry HHI in columns 3, 6 is measured at the industry level instead of the industry-region level. The level effects of the variables interacted with treated are absorbed by firm fixed effects and, thus, not reported. All columns include firm fixed effects and (two-digit SIC) industry×region×year fixed effects. Controls include firm age (log transformed) and the number of firms in an industry-region (log-transformed). *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

Log(1+Capex)

(5)

(6)

Capex/Asset Low debt firms

treatedt−1

treatedt−1 ×export

0.175

0.007

0.218

0.005

-0.005

0.004

(0.041)***

(0.073)

(0.094)**

(0.004)

(0.004)

(0.007)

-0.300

-0008

(0.031)***

(0.002)***

treatedt−1 × industry-region HHI

0.610

0.036

(0.166)***

(0.007)***

treatedt−1 ×industry HHI

Controls

X

X

-1.428

-0.021

(1.689)

(0.092)

X

X

X

X

Adj. R2

0.70

0.73

0.73

0.57

0.59

0.58

Obs.

8,162

8,162

8,162

8,162

8,162

8,162

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Industry×Region×Year FE

Yes

Yes

Yes

Yes

Yes

Yes

– 40 –

Table 7: Dynamic analysis: Competitors’ investment This table reports results of regressions of competitors’ investment decisions on the treated indicator defined two years after the announcement (treatedt−2 ), one-year after the announcement (treatedt−1 ), contemporaneously with the announcement (treatedt ), one-year before the announcement (treatedt+1 ), and a set of controls. The sample is the baseline sample used in Table 2. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. All columns include firm and year fixed effects. Columns 3 and 7 also include (two-digit SIC) industry×year fixed effects and columns 4, and 8 add region×year fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

(5)

(6)

Log(1+Capex)

treatedt−2

treatedt−1

treatedt

(7)

(8)

Capex/Assets

0.0911

0.0915

0.0916

0.101

0.0045

0.0046

0.0042

0.0042

(0.0356)**

(0.0357)**

(0.0438)*

(0.0510)*

(0.0023)*

(0.0046)*

(0.0042)*

(0.0042)*

0.0601

0.0643

0.0556

0.0469

0.0025

0.0024

0.0021

0.0013

(0.0488)

(0.0472)

(0.0534)

(0.0568)

(0.0013)*

(0.0012)**

(0.0014)

(0.0017)

0.0451

0.0431

0.0490

0.0727

0.0011

0.0011

0.0013

0.0020

(0.0477)

(0.0471)

(0.0453)

(0.0432)

(0.0010)

(0.0010)

(0.0011)

(0.0015)

0.0229

0.0241

0.0226

0.0261

0.0004

0.0004

0.0002

0.0002

(0.0433)

(0.0435)

(0.0417)

(0.0455)

(0.0018)

(0.0019)

(0.0019)

(0.0017)

0.68

0.68

0.68

0.69

0.50

0.50

0.51

0.52

11,374

11,374

11,374

11,374

11,374

11,374

11,374

11,374

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

treatedt+1

Adj. R2 Obs.

Industry×Year FE

Yes

Region×Year FE

Yes Yes

– 41 –

Yes

Yes Yes

Table 8: Dynamic analysis: Other Measures This table reports results of regressions of competitors’ investment decisions (Columns 1-3), Return on Assets (Columns 4-6) and Cashflows to Sales (Columns 7-9) on the treated indicator defined two years after the announcement (treatedt−2 ), one-year after the announcement (treatedt−1 ), contemporaneously with the announcement (treatedt ), one-year before the announcement (treatedt+1 ), and a set of controls.The sample is the baseline sample used in Table 2. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. All columns include firm and year fixed effects. Columns 2, 5, and 8 also include (two-digit SIC) industry×year fixed effects and columns 3, 6, and 9 add region×year fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

Asset Growth

treatedt−2

– 42 –

treatedt−1

treatedt

treatedt+1

Adj. R2

(5)

(6)

(7)

ROA

(8)

(9)

Cash Flows/Sales

0.0086

0.0080

0.0138

0.0032

0.0032

0.0041

0.0880

0.0671

0.142

(0.0035)**

(0.0026)***

(0.0052)**

(0.0024)

(0.0024)

(0.0025)

(0.188)

(0.160)

(0.210)

0.0017

0.0009

0.0010

-0.0008

-0.0016

-0.0012

-0.0400

-0.0790

-0.0639

(0.0019)

(0.0019)

(0.0027)

(0.0012)

(0.0014)

(0.0013)

(0.133)

(0.126)

(0.125)

-0.0046

-0.0054

-0.0039

0.0004

-0.0002

-0.0004

0.166

0.120

0.155

(0.0032)

(0.0033)

(0.0030)

(0.0008)

(0.0003)

(0.0004)

(0.126)

(0.133)

(0.099)

0.0051

0.0036

0.0050

0.0005

0.0009

0.0013

0.0725

0.105

0.0779

(0.0032)

(0.0027)

(0.0034)

(0.0006)

(0.0008)

(0.0010)

(0.0753)

(0.0844)

(0.100)

0.33

0.33

0.33

0.66

0.66

0.66

0.62

0.62

0.62

11,281

11,281

11,281

11,374

11,374

11,374

11,374

11,374

11,374

Controls

X

X

X

X

X

X

X

X

X

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes Yes

Yes

Yes

Yes

Yes

Yes

Obs.

Industry×Year FE Region×Year FE

Yes

Yes

Yes

Yes

Yes

Table 9: Falsification test: Negative announcements This table reports results of regressions of competitors’ investment decisions and how these decisions interact with leverage following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The table repeats the estimations in Table 4, but in a sample of restructuring announcements which reveal negative news. All columns include firm and year fixed effects. Columns 3 and 8 also include (two-digit SIC) industry×year fixed effects and Columns 4 and 9 add region×year fixed effects. Columns 5 and 10 include (two-digit SIC) industry×region×year fixed effects in addition to the firm fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Log(1+Capex)

– 43 –

treatedt−1

treatedt−1 ×high debt

Adj. R2 Obs.

Capex/Assets

-0.0098

-0.0168

-0.0115

0.0034

-0.0001

-0.0002

-0.0005

-0.0003

0.0005

(0.0226)

(0.0198)

(0.0234)

(0.0026)

(0.0018)

(0.0017)

(0.0016)

(0.0016)

(0.0018)

-0.0367

-0.0395

-0.0366

-0.0320

-0.0351

-0.0009

-0.0009

-0.0008

-0.0007

-0.0013

(0.0320)

(0.0314)

(0.0318)

(0.0319)

(0.0303)

(0.0024)

(0.0024)

(0.0024)

(0.0024)

(0.0022)

0.46

0.46

0.46

0.47

0.47

0.19

0.19

0.19

0.19

0.19

18,857

18,857

18,857

18,857

18,857

18,857

18,857

18,857

18,857

18,857

X

X

X

X

X

X

X

X

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Industry×Region×Year FE

(10)

-0.0086

Firm FE

Region×Year FE

(9)

(0.0223)

Controls

Industry×Year FE

(8)

Yes

Yes Yes

Yes Yes

Yes

Table 10: Positive announcements and competitors’ investment for firms with common ownership This table reports results of regressions of competitors’ investment decisions following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The sample includes firms which share common ownership. treated is defined as the logarithm of the total number of workers announced to be laid off in a given three-digit SIC industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. All columns include firm and year fixed effects. Columns 3 and 6 also include (two-digit SIC) industry×year fixed effects and columns 4 and 8 add region×year fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008.

(1)

(2)

(3)

(4)

(5)

Log(1+Capex)

treatedt−1

(6)

(7)

(8)

Capex/Assets

0.086

0.094

0.120

0.129

0.0036

0.0037

0.0048

0.0027

(0.044)*

(0.048)*

(0.068)*

(0.081)

(0.0015)***

(0.0013)**

(0.0012)***

(0.0020)

X

X

X

X

X

X

Controls Adj. R2

0.69

0.70

0.73

0.75

0.49

0.49

0.57

0.62

Obs.

1,092

1,092

1,092

1,092

1,092

1,092

1,092

1,092

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Industry×Year FE Region×Year FE

Yes

– 44 –

Yes

Table 11: Robustness: Alternative definitions of treatment indicator This table reports results of regressions of competitors’ investment decisions following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The table repeats the estimations in Table 2. In Panel A, treated takes the value of 1 following restructuring announcements in a given industry-region and is 0 otherwise. In Panel B, treated is the ratio of the cumulative number of workers announced to be laid off in a specific industry-region scaled by the total number of workers in the industry-region and is 0 otherwise. treated is lagged by one year. All columns include firm and year fixed effects. Columns 3 and 7 also include (two-digit SIC) industry×year fixed effects and columns 4 and 8 add region×year fixed effects. Controls include firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008. (1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Panel A Log(1+Capex)

– 45 –

treatedt−1

Adj. R2

Capex/Assets

0.198

0.215

0.229

0.249

0.0111

0.0115

0.0129

0.0113

(0.0566)***

(0.0536)***

(0.0480)***

(0.0807)***

(0.0042)**

(0.0042)**

(0.0039)***

(0.0057)*

0.64

0.64

0.64

0.65

0.45

0.45

0.46

0.46

Panel B Log(1+Capex)

treatedt−1

Adj. R2 Obs.

0.289

0.273

0.282

0.288

0.0187

0.0188

0.0191

0.0203

(0.0304)***

(0.0303)***

(0.0276)***

(0.0409)***

(0.0038)***

(0.0038)***

(0.0039)***

(0.0048)***

0.64

0.64

0.64

0.65

0.45

0.45

0.46

0.46

17,298

17,298

17,298

17,298

17,298

17,298

17,298

17,298

X

X

X

X

X

X

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Yes

Controls

Firm FE

Yes

Yes

Year FE

Yes

Yes

Industry×Year FE Region×Year FE

Capex/Assets

Yes

Yes Yes

Yes

Yes Yes

Appendix A Figure A1: Distribution of number of announced layoffs The figure plots the distribution of the total number of workers announced to be laid-off for restructuring announcements which signal that competition intensifies and for those which reveal negative news.

784

95%

724 573

90%

660

351

Percentiles

75%

500 210

50%

200 132

25%

Negative News

145

Positive News

90

10%

100 39

5%

60

0

200

400

600

No of workers announced to be laid-off

– 46 –

800

1000

Table A1: Distribution of restructuring announcements by industry This table lists the 3-digit SIC industries with restructuring announcements. Column 1 reports the description of the industry and column 2 reports the 3-digit SIC code. Column 3 reports the distribution of announcements by industry. These statistics are based on both positive and negative announcements.

Industry

SIC code

Distribution of announcements

Food and Kindred Products Meat Products

201

2.75%

Dairy Products

202

3.58%

Canned, Frozen, and Preserved Fruits, Vegetables, and Food Specialties

203

0.83%

Bakery Products

205

3.03%

Sugar and Confectionery Products

206

2.75%

Beverages

208

1.10%

Miscellaneous Food Preparations and Kindred

209

5.23%

Cigarettes

211

0.83%

Cigars

212

0.55%

Tobacco Products

Textile Mill Products Broadwoven Fabric Mills, Cotton

221

0.55%

Knitting Mills

225

0.50%

Carpets and Rugs

227

0.28%

Miscellaneous Textile Goods

229

0.55%

Men’s and Boys’ Furnishings, Work Clothing, and Allied Garments

232

0.55%

Women’s, Misses’, and Juniors’ Outerwear

233

0.28%

Miscellaneous Apparel and Accessories

238

0.55%

249

0.28%

251

0.83%

Paper Mills

262

1.65%

Paperboard Containers and Boxes

265

0.83%

Converted Paper and Paperboard Products, Except

267

0.55%

271

2.20%

Apparel and Other Finished Products

Lumber and Wood Products, Except Furniture Miscellaneous Wood Products Furniture and Fixtures Household Furniture Paper and Allied Products

Printing, Publishing, and Allied Industries Newspapers: Publishing, or Publishing and Printing Periodicals: Publishing, or Publishing and Printing

272

0.55%

Books

273

0.55%

Commercial Printing

275

2.75%

Chemicals and Allied Products

– 47 –

Table A1 (continued)

Industry

SIC code

Distribution of announcements

Industrial Inorganic Chemicals

281

2.75%

Drugs

283

2.20%

Soap, Detergents, Cleaning Preparations; Perfumes, Cosmetics, and Others

284

1.10%

Agricultural Chemicals

287

0.55%

Miscellaneous Chemical Products

289

0.83%

Tires and Inner Tubes

301

0.83%

Miscellaneous Plastics Products

308

1.93%

314

0.55%

Glass Products, Made of Purchased Glass

323

0.28%

Structural Clay Products

325

0.28%

Pottery and Related Products

326

1.10%

331

3.58%

Rubber and Miscellaneous Plastics Products

Leather and Leather Products Footwear, Except Rubber Stone, Clay, Glass, and Concrete Products

Primary Metal Industries Steel Works, Blast Furnaces, and Rolling and Finishing Mills Iron and Steel Foundries

332

0.55%

Primary Smelting and Refining of Nonferrous

333

1.38%

Miscellaneous Primary Metal Products

339

0.55%

Fabricated Metal Products Cutlery, Handtools, and General Hardware

342

1.10%

Fabricated Structural Metal Products

344

0.55%

Metal Forgings and Stampings

346

0.28%

Miscellaneous Fabricated Metal Products

349

4.13%

Engines and Turbines

351

1.65%

Construction, Mining, and Materials Handling

353

0.83%

Industrial and Commercial Machinery and Computer Equipment

Special Industry Machinery, Except Metalworking

355

0.83%

Computer and Office Equipment

357

3.86%

Electric Transmission and Distribution Equipment

361

1.65%

Household Appliances

363

1.93%

Electric Lighting and Wiring Equipment

364

0.28%

Electronic and Other Electrical Equipment, Exc. Computers

Communications Equipment

366

3.58%

Electronic Components and Accessories

367

1.93%

Miscellaneous Electrical Machinery, Equipment, and Supplies

369

3.31%

371

11.27%

Transportation Equipment Motor Vehicles and Motor Vehicle Equipment Aircraft and Parts

372

2.75%

Ship and Boat Building and Repairing

373

1.38%

– 48 –

Table A1 (continued)

Industry

SIC code

Distribution of announcements

Railroad Equipment

374

1.93%

Miscellaneous Transportation Equipment

379

0.28%

Measuring, Analyzing, and Controlling Instruments etc. Search, Detection, Navigation, Guidance, Aeronautical, and Nautical Systems etc.

381

0.28%

Surgical, Medical, and Dental Instruments and Supplies

384

0.83%

399

2.20%

Miscellaneous Manufacturing Industries Miscellaneous Manufacturing Industries

Total

100%

– 49 –

Table A2: Distribution of restructuring announcements by region This table reports the distribution of announcements by region. These statistics are based on both positive and negative announcements.

Region in the UK

Distribution of announcements

East of England

6.89%

East Midlands

8.82%

Greater London

1.93%

Northern Ireland

5.23%

North East

5.51%

North West

10.47%

Scotland

9.09%

South East

4.68%

South West

9.92%

Wales

14.33%

West Midlands

14.60%

Yorkshire and the Humber

8.54%

Total

100%

– 50 –

Table A3: Distribution of firms across regions by industry This table reports the distribution of firm-year observations across regions for each two-digit SIC industry. The 12 regions correspond to the government official regions (GORs) in the UK. Shares (in percentages) in each row sum up to one. These statistics are computed for the baseline sample of positive announcements used in the analysis.

Industry

– 51 –

East of England

East Midlands

Greater London

Northern Ireland

North East

North West

Scotland

South East

South West

Wales

West Midlands

Yorkshire & Humber

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

20

8.2

4.2

24.2

6.5

1.5

9.5

7.9

12.3

4.1

3.9

8.7

9.1

21

–

–

46.2

21.5

–

21.5

–

10.8

–

–

–

–

22

2.7

18.2

8.9

5.0

0.3

27.3

7.8

3.8

4.7

0.8

6.1

14.5

23

7.8

12.2

32.2

1.9

1.1

19.5

3.4

4.4

3.2

1.0

4.4

8.9

24

10.0

6.5

6.1

4.0

3.7

13.5

14.1

10.9

6.8

4.5

8.3

11.8

25

11.2

4.1

14.9

10.8

–

16.4

6.1

17.0

11.0

0.4

1.8

6.4

26

10.3

15.3

11.3

4.5

5.3

6.5

5.4

15.5

6.2

2.3

7.4

10.1

27

6.1

5.3

38.8

2.0

2.5

8.0

5.0

17.3

3.4

0.3

5.6

5.8

28

16.7

9.9

15.6

2.5

1.6

9.6

3.4

14.3

5.7

5.0

9.2

6.6

29

–

7.4

24.0

–

4.2

16.3

6.5

14.5

–

5.0

5.3

16.6

30

3.6

15.6

11.3

5.0

3.8

13.0

1.2

15.4

–

6.8

11.7

12.6

31

4.0

7.5

18.9

–

–

6.5

15.9

8.5

11.4

–

12.9

14.4

32

7.8

17.2

6.2

3.3

3.2

12.8

7.7

11.1

5.9

4.8

7.2

12.7

33

5.5

8.4

9.8

3.1

4.0

8.1

6.3

9.4

8.2

2.1

22.4

12.6

34

5.5

4.3

9.9

0.1

2.0

15.4

3.2

18.7

2.7

6.4

12.2

19.5

35

5.8

8.9

8.3

6.4

1.3

10.6

4.8

23.5

8.4

2.1

12.5

7.3

36

15.8

3.8

20.8

0.2

4.8

4.6

2.7

22.6

7.8

1.0

9.1

6.8

37

7.3

5.6

9.1

1.8

6.8

5.2

8.2

4.4

7.5

9.2

16.0

19.0

38

17.2

6.3

11.7

1.7

1.2

9.3

4.3

26.0

5.4

4.1

6.7

6.1

39

12.5

9.3

11.9

2.2

1.7

13.7

4.3

17.4

0.3

0.5

14.5

11.8

Appendix B: Wages and Hiring In this section, I examine the effect of layoff announcements on local labor markets. In particular, I study the effect of layoffs on wages per employee for firms operating in industryregions with restructuring announcements, as compared to firms in industry-regions with no announcements. For this analysis, I use all announcements, both those that signal an improvement in the competitive position of the restructuring firm and those that reveal bad news, as they all involve layoffs. Table B1 reports the results. In Columns 1-3, Panel A, the dependent variable is wages per employee (log-transformed). Wages are deflated using the consumer price index (CPI) provided by the UK Office for National Statistics (ONS). Column 1 includes firm and year fixed effects and firm-level controls; column 2 additionally controls for (twodigit SIC) industry×year fixed effects; column 3 adds region×year fixed effects to the previous specification. Panel A shows that average wages per employee are lower for firms in industry-regions with layoff announcements compared to those with no announcements. The result is statistically significant across specifications at the 1% or 5% level. These results suggest that a 10% increase in treatment will lower the £25,633 median real wage per worker by £533. Columns 4-6 repeat the analysis for number of employees (log-transformed) as the dependent variable. The coefficient is positive, weakly statistically significant in columns 1, and 2, and just outside regular levels of significance in column 3. There is a fairly recent literature which argues that leverage interacts with wage setting and employment policies.13 Table B1 reports results on wages per employee (Panel B) and 13

Chodorow-Reich (2014) finds that firms which borrowed from lenders deeply affected by the Lehman bankruptcy reduced employment relatively more during the crisis. Benmelech, Bergman,

– 52 –

employment (Panel C) for low and high debt firms.14 Panel B shows a negative effect on wages per employee for both low and high debt firms. Panel C shows that the channel explaining the lower wages differs for the two groups of firms: employment increases for the low debt firms suggesting lower entry wages, while employment does not change for high debt firms suggesting wage renegotiation with current employees.15 My findings suggest that competitors in treated industry-regions take advantage of workers’ lower outside option when bargaining with them on wages. The news of layoffs have an impact on the reservation utilities of workers employed in competitor firms worsening their bargaining position and ability to extract a higher surplus. Table B2 provides further support for that explanation exploiting heterogeneity across industries with different unionization rates. Since unionization rates are not available for the UK, I use US data, available at the 3-digit SIC level for 2002 by the Barry T. Hirsch and David A. Macpherson (2003) database. The use of US data to measure industries’ dependence on unions relies on the assumption that some industries are more unionized than others due to structural (e.g., technological) reasons and these reasons are similar across countries (Rajan and Zingales, 1998).16 Consistent with my hypothesis, Table B2 shows a lower reduction in the average and Seru (2011) show that the availability of credit plays an important role in firm-level employment decisions. 14

As in the previous analysis, I sort firms in low and high debt using their pre-treatment ratio of total debt net of cash over the book value of assets as it compares to their industry median. 15

I cannot rule out, however, the possibility that low debt firms also renegotiate down wages of their current employees. 16 This is not an extreme assumption since UK industries are likely to be sharing common characteristics with their US counterparts, while the institutional environment governing industrial relations and the organization of labor bargaining are similar in the two countries. Several papers in the labor literature which study the determinants of union membership (Blanchflower, 2006), or the effect of unions on wages and wage structure (Blanchflower and Bryson, 2003, Card, Lemieux, and Riddell, 2003) uncover similar patterns in the UK and the US.

– 53 –

wage per employee when industry unionization is higher. These results make the implicit assumption that local labor markets are quite segmented in the UK. This assumption is plausible if we consider that it is relatively costly for workers to move. To support this assumption, I use data on workers’ commutes provided by the Annual Population Survey (APS, 2008) for the UK. The APS provides information on commuting flows between different districts and counties in the UK. I find that the percent of local jobs taken by local residents is very high - on average higher than 90%.17

17 I also looked at a predecessor survey of the APS, the Local Labour Force Survey 2001, and found similar results.

– 54 –

References [1] Benmelech, E., N. Bergman, and A. Seru, 2011, “Financing Labor”, NBER Working Paper. [2] Blanchflower, D. G., 2006, “A Cross-Country Study of Union Membership,” IZA Discussion Papers, Chapter 7. [3] Blanchflower, D. G., and A. Bryson, 2003, “Changes over Time in Union Relative Wage Effects in the UK and the US Revisited,” International Handbook of Trade Unions, Chapter 7. [4] Card, D., T. Lemieux, and C. W. Riddell, 2003, “Unionization and Wage Inequality: A Comparative Study of the U.S., the UK., and Canada,” NBER Working Paper Series. [5] Chodorow-Reich, G., 2014, “The Employment Effects of Credit Market Disruptions: Firm-Level Evidence from the 2008-9 Financial Crisis”, The Quarterly Journal of Economics, 129, 1-59. [6] Hirsch, B. T., and D. A. Macpherson, 2003, “Union Membership and Coverage Database from the Current Population Survey: Note,” Industrial and Labor Relations Review, 56, 349-354. [7] Rajan, R. G., and L. Zingales, 1998, “Financial Dependence and Growth,” American Economic Review, 88, 559-586.

– 55 –

Table B1: Wages and hiring This table reports results of regressions of competitors’ wages per employee (Panels A, B) and employment (Panels A, C) following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The sample includes all restructuring announcements involving layoffs. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. All columns include firm and year fixed effects. Columns 2 and 5 also include (two-digit SIC) industry×year fixed effects and columns 3 and 6 add region×year fixed effects. Controls include firm profitability (ROA), firm size in columns 1-3 of Panels A and B (measured by the logarithm of assets), firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008. (1)

(2)

(3)

(4)

(5)

(6)

Panel A Log(Wages/Employees)

treatedt−1

Adj. R2 Obs.

Log(Employees)

-0.0019

-0.0023

-0.0021

0.0036

0.0026

0.0022

(0.0008)**

(0.0007)***

(0.0007)**

(0.0018)*

(0.0015)*

(0.0013)

0.90

0.91

0.91

0.96

0.96

0.96

40,389

40,389

40,389

40,389

40,389

40,389

Panel B Log(Wages/Employees) Low debt firms

treatedt−1

Adj. R2

High debt firms

-0.0024

-0.0020

-0.0019

-0.0017

-0.0024

-0.0021

(0.0007)***

(0.0008)**

(0.0009)*

(0.0009)*

(0.0010)**

(0.0011)*

0.90

0.90

0.90

0.91

0.91

0.91

Panel C Log(Employees) Low debt firms

treatedt−1

Adj. R2

High debt firms

0.0055

0.0041

0.0040

0.0024

0.0016

0.0011

(0.0016)***

(0.0012)***

(0.0012)***

(0.0024)

(0.0025)

(0.0024)

0.96

0.96

0.96

0.95

0.95

0.95

Controls

X

X

X

X

X

X

Firm FE

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes Yes

Yes

Yes

Yes

Ind.*Year FE Region*Year FE

Yes

Yes

– 56 –

Yes

Table B2: Wages and workers’ outside option This table reports results of regressions of competitors’ wages per employee and how wages per employee interact with unionization following the restructuring announcements in a given industry-region as compared to a set of controls in different industry-regions with no announcements. The sample includes all restructuring announcements involving layoffs. treated is defined as the logarithm of the total number of workers announced to be laid off in a given (three-digit SIC) industry-region and is 0 in industry-regions with no restructuring announcements. treated is lagged by one year. union measures unionization rates at the (three-digit SIC) industry level for the US in 2002 (Barry T. Hirsch and David A. Macpherson database, 2003). Union main effects are absorbed by the firm fixed effects and, thus, not reported. All columns include firm and year fixed effects. Column 2 includes (two-digit SIC) industry×year fixed effects, column 3 adds region×year fixed effects. Column 4 controls for (two-digit SIC) industry×region×year fixed effects. Controls include firm profitability (ROA), firm size (measured by the logarithm of assets), firm age (log-transformed), the number of firms in an industry-region (log-transformed), and macro controls (regional unemployment rate, regional GDP growth rate). Macro controls are absorbed by region×year fixed effects. *, **, ***, indicates significance at the 10%, 5% and 1% level respectively. Standard errors, reported in parentheses, are clustered at the industry and region level. Firm-level variables are winsorized at the 1% tails. The sample includes manufacturing firms in the UK and covers years 2001-2008. Log(Wages/Employees)

treatedt−1

treatedt−1 ×union

(1)

(2)

(3)

(4)

-0.0054

-0.0053

-0.0054

-0.0074

(0.0026)*

(0.0018)***

(0.0018)***

(0.0029)**

0.0167

0.0148

0.0160

0.0206

(0.0094)*

(0.0058)**

(0.0054)***

(0.0088)**

X

X

X

X

Firm FE

Yes

Yes

Yes

Yes

Year FE

Yes Yes

Yes

Controls

Industry×Year FE Region×Year FE

Yes

Industry×Region×Year FE Adj. R2 Obs.

Yes

0.91

0.91

0.91

0.91

35,784

35,784

35,784

35,784

– 57 –