Does Stock Market Listing Boost or Impede Corporate Investment?

Transcription

1. Does Stock Market Listing Boost or Impede Corporate Investment ? İbrahim Yarba Ahmet Duhan Yassa July 2021 Working Paper No: 21/12
2. © Central Bank of the Republic of Turkey 2021 Address: Central Bank of the Republic of Turkey Head Office Structural Economic Research Department Hacı Bayram Mah. İstiklal Caddesi No: 10 Ulus, 06050 Ankara, Turkey Phone: +90 312 507 80 04 Facsimile: +90 312 507 78 96 The views expressed in this working paper are those of the author(s) and do not necessarily represent the official views of the Central Bank of the Republic of Turkey.
3. Does Stock Market Listing Boost or Impede Corporate Investment ? İbrahim YARBA* Ahmet Duhan YASSA** Abstract This paper investigates investment behavior across public and privately held firms using a novel firm-level dataset. We use coarsened exact matching to construct a control group of firms with which we compare listed firms before and after listing in a difference-in-differences framework. Results reveal that stock market listing spurs growth significantly in terms of sales, employment and assets for manufacturing firms. Furthermore, results indicate that manufacturing listed firms invest more than their non-listed counterparts. In addition, their investment decisions are significantly more sensitive to changes in investment opportunities, and they respond more aggressively. These results constitute a rejecting evidence against existence of short-termism for manufacturing listed firms in Turkey. Moreover, these findings provide significant support for the arguments regarding the advantages of public firms in terms of better access to external finance and enhanced corporate structure, which enables them to fulfill growth potential much easily, and highlight the importance of policies that should be implemented to deepen the Turkish capital markets. Keywords: Stock market listing; corporate investment; firm growth; coarsened exact matching; difference-in-differences JEL codes: C23, D22, G31, G32, L25 The views expressed in this study are those of the authors and do not necessarily represent the official views of the Central Bank of the Republic of Turkey. * Structural Economic Research Department, Central Bank of the Republic of Turkey, Istiklal Cad. No:10, 06050 Ulus, Ankara, Turkey. E-mail: ibrahim.yarba@tcmb.gov.tr ** Structural Economic Research Department, Central Bank of the Republic of Turkey, Istiklal Cad. No:10, 06050 Ulus, Ankara, Turkey. E-mail: ahmetduhan.yassa@tcmb.gov.tr 1
4. Non-technical Summary Despite growing empirical research contrasting investment behavior of stock market-listed and privately held firms , evidence provided is mixed, which seems to be much more severe for emerging countries. In order to shed some light on the issue, we analyze how being public affects firm investment decisions in Turkey, one of the most important emerging economies. The ambiguity in existing empirical findings can be attributed to data availability and the main difficulty is finding a comparable counterfactual for how listed firms would behave absent their listed status. We address this difficulty by finding comparable privately held firms and analyze the differentiation in their investment behaviors, utilizing a novel dataset of privately held Turkish non-financial firms covering more than 1,000,000 firms over the period 2006-2018. Using the combination of coarsened and exact matching procedure and difference in difference methodology, first, we document that stock market listing has a significant positive impact on firm growth for manufacturing firms. Stock market-listed manufacturing firms become significantly larger in post listing period compared to their private counterparts as their number of employees, total assets and sales increase significantly. Results of this study also show that stock market-listed manufacturing firms invest more for expansion than their non-listed counterparts. In addition, their investment decisions are significantly more sensitive to changes in investment opportunities and respond more aggressively in post-listing period. This provides support for the findings of previous research regarding the advantages of public firms in terms of better access to external finance and enhanced corporate structure, which enables them to fulfil growth potential much easily. Moreover, results of this study provide an indirect evidence on the important role of financial market development in mitigating frictions regarding information asymmetry, and easing the access of firms to capital. As in many emerging countries, in Turkey, dominant source of external finance for privately held firms is bank lending, and external financing alternative to straight bank debt is quite limited. Besides, bank lending is highly cyclical and vulnerable to financial and economic conditions. By disseminating information across different classes of investors, a firm can increase its funding opportunities beyond not only bank loans, but also this can help broaden options through bond markets and other alternative funding instruments. However, despite the improvement in financial development in the last decade, number of public firms is still limited in Turkey. There are only around four hundred firms listed on Borsa Istanbul. In comparison to her peer countries, stock market capitalization as a percent of gross domestic product is also low. This suggests room for growth and results of this study highlight the importance of policies that should be implemented to deepen the Turkish capital markets. 2
5. 1 . Introduction Financial markets and capital allocation process has been one of the most prominent topics in finance literature. Although the economic distinction between stock market listed and nonlisted firms have been widely discussed, prior researches in this area have rarely paid attention to differentiation in their investment behavior. Considering inherent trade-offs between the costs (Berle and Means, 1932; Jensen and Meckling, 1976) and the benefits associated with being part of the public equity markets (Clementi, 2002), the impact of stock market listing on corporate investments is theoretically ambiguous and needs empirical examination. Despite the importance of the issue, the empirical literature contrasting public and private firms’ investment behavior is scarce and evidence provided is mixed. Although, the issue is clarified for neither the advanced nor the emerging countries, the ambiguity seems to be much more severe for the latter. In order to shed some light on the issue, we analyze how being listed affects firm investment decisions in Turkey, one of the most important emerging economies. The scarcity in existing empirical findings can be attributed to data availability because of the confidentiality of financial information of privately held firms. Thus, the main difficulty is finding a comparable counterfactual for how listed firms would behave absent their listed status. We address this difficulty by finding comparable privately held firms and analyze the differentiation in their investment behaviors, utilizing a novel dataset of privately held Turkish non-financial firms covering more than 1,000,000 firms over the period 2006-2018. The primary concern in the empirical analysis is that firms that are more likely to increase their investments will also be the firms that decides to be listed in stock market (Bernstein, 2015). In order to deal with this potential endogeneity we first, construct a control group of privately held firms using coarsened and exact matching (CEM) methodology. In contrast with previous studies, our novel dataset allows us to match firms based on a rich criteria set such as firm size, number of employment, fixed assets, cash holdings, both financial and total debt structures, revenues, sector, and inventories. This enables us to find reliable and comparable control firms. This is crucial for the quality of impact analysis, which is mostly one of the main drawbacks of the previous studies. Then, since we have data both before and after firms are being listed, we 3
6. employ a difference in differences (DD) setup in order to compare listed firms with the matched control firms (Caliendo and Kopeinig, 2008). In the first place, results reveal that that stock market listing has a significant positive impact on firm growth for manufacturing firms. Using the combination of coarsened and exact matching procedure and difference in differences methodology (CEM-DD), results show that stock market-listed manufacturing firms become larger in post listing period relative to their private counterparts as their total assets, sales and number of employees increase significantly. Results also reveal a significant increase in both gross and net tangible fixed assets. This suggests that listed manufacturing firms invest more for expansion than their non-listed counterparts in the post-listing period. In order to assess whether manufacturing public firms invest more due to better investment opportunities they face in post-listing period, we also incorporate investment opportunities in our empirical model. Results show that holding investment opportunities constant does not alter our findings. Moreover, results reveal that manufacturing listed firms not only invest more that their non-listed counterparts in terms of investment level, but also their investment decisions are significantly more sensitive to changes in investment opportunities and they respond more aggressively in post-listing period. Our results contribute to several strands of the literature. First, our paper adds to the growing empirical literature on the investment behavior distinction between stock market listed and nonlisted firms. Conventional literature argues that listing in stock market paves the way for future growth of firms by allowing access to cheap finance, enhanced corporate structure and firm reputation. Jain and Kini (1994) and Mikkelson et al. (1997) provide some pioneer empirical evidences in support for the claim that public firms grow in terms of sales and capital expenditure in US. In the same vein, Kim and Weisbach (2008), Brav (2009) and Mortal and Reisel (2013) focus on capital expenditure behavior of public firms in their studies and they show that public firms invest more and their responsiveness to investment opportunities is higher than privately held firms in both US and Europe. Maksimovic et al. (2019) confirms same results and contributes that especially public firms which supported by venture capital are more responsive compared to similar counterpart private firms. 4
7. On the other hand , some counter arguments compatible with agency theory put forwards that listing in stock market can create agency cost and distort investment tendency because of dispersion on ownership structure (Jensen and Meckling, 1976). Such a distortion may occur through different channels. First, managers tends to give up from long term objectives in order to increase firms’ value in the eye of investors (Stein, 1989). Second, when a firm become public, monitoring of managers by shareholders become crucial. “Empire building” hypothesis predicts that if monitoring is weak, managers can decide self-ordainedly and this may damage optimal investment policy of firms (Jensen, 1986). Concordantly, managers can behave shorttermist and prioritize short-term profit instead of long term goals to obtain reputation (Narayanan, 1985). Third, similar to empire building, according to “quiet life” hypothesis, managers can avoid from the risky investments and lower their investment level (Bertrand and Mullainathan, 2003). Common consequence of all these channels is distortion on investment policies in public firms. This is evidenced by Graham et al. (2005) reporting that most of the managers smooth short-term earnings by giving up long term investment. Asker et al. (2015) also provide significant evidence in support for this short-termism. They show that US listed firms invest substantially less, and they are less sensitive to investment opportunities. In contrast to these concerns mentioned in the literature, we find no evidence in favor of shorttermism in Turkey. On the other hand, our findings corroborates traditional theory, which argue that public firms catch higher growth and investment rate than similar privately held firms in post-listing period. Our result can also be seen as rejection of empire building and quiet life hypothesis in terms of investment behavior. Furthermore, our study adds to the growing literature on financial market development and firm growth. Specifically, our findings shed new light on the debate as to whether the economic advantage of listed firms varies with institutional setting. Mortal and Reisel (2013), for instance, argue that investment sensitivity to growth opportunities is higher for listed firms only in countries with well-developed stock markets by utilizing a large cross-country data set. Contrary to these findings, our results suggest that even in a country with a relatively lessdeveloped stock market, stock market listing leads firms to take advantage of investment opportunities that might not be undertaken if the firms were private. 5
8. The remaining of the paper is as follows . Section 2 depicts Turkish stock market structure and its development over time, Section 3 explains data and methodology, Section 4 presents results and robustness tests, and finally Section 5 concludes. 2. Stock Market in Turkey Founding in 1986, Borsa Istanbul (BIST) has grown rapidly with financial liberalization reforms in 1990s. The development of Turkish stock market is depicted in Figure 1. 1000 900 800 700 600 500 400 300 200 100 0 450 430 410 390 370 350 330 310 290 270 250 Number of Listed Firms (Right Axis) GDP level Figure 1. Gross Domestic Product and number of listed firms in Turkey This figure shows GDP,and number of listed firms in Turkey over 1999-2019. GDP is in billion US dollars and obtained from the World Bank dataset while number of listed firms including financial firms is compiled from Capital Market Board dataset. Following 2001 national crisis, Turkey entered a fast growth period. Between the years of 20022013, GDP grew up steadily and market capitalization accompanied this increasing as well. Then, both GDP and stock market capitalization have decreased at some extent, which can be attributable to the negative effect of FED’s tapering on emerging market and decreasing business dynamism in Turkey (Akcigit et al. 2020). As of January 2020, there exists 402 firms listed in BIST including financial firms and market capitalization has reached to 184 billion USD. Firm size distributions of listed and unlisted firms are given in Figure 2. Listed firms in BIST are mostly large firms in contrast to unlisted firms. Of the firms listed in BIST, 70.85% are large firms, 21% are medium-sized firms, 8.15% are small firms, and none is micro-sized firms, while the same ratios for all incorporated unlisted firms are 1.19%, 5.05%, 24.49% and 69.27%, respectively. 6
9. 80 % 70% 60% 50% 40% 30% 20% 10% 0% Listed firms Unlisted firms Large Medium Small Micro Figure 2. Size distributions: Listed vs unlisted firms The graph shows size distribution of listed and unlisted Turkish firms in 2018 based on authors’ own calculations using data from the Revenue Administration dataset. Listed firms account for 8.96% of sales and 3.91% of employment of all Turkish firms covered in the database of Revenue Administration on average, which includes the universe of all incorporated firms’ financial tables in Turkey (Figure 3). Besides, their total assets and tangible fixed assets share are around 9.13% and 12.47%, respectively, which indicates that investment reactions of listed firms are of capital importance for Turkish economy. 20% 0.08% 18% 0.07% 16% 0.06% 14% 12% 0.05% 10% 0.04% 8% 0.03% 6% 0.02% 4% 0.01% 2% 0% 0.00% 2006 2007 2008 2009 2010 Number of firms (right axis) Total assets 2011 2012 2013 2014 Employment Tangible fixed assets 2015 2016 2017 2018 Sales Figure 3: Share of listed firms in terms of employment, total assets, sales and tangible fixed assets (% of all incorporated firms) Source: Authors’ calculations using data from the Revenue Administration dataset. 7
10. However , Figure 3 shows that the shares of listed firms in terms of total assets, employment, sales, and tangible assets decrease over time in Turkey, which is compatible with Figure 1. Moreover, only around 0.05% of all incorporated firms are listed in Turkey on average over the sample period (Figure 3). In comparison to both advanced and peer emerging countries, BIST has also relatively low stock market capitalization as a percent of gross domestic product Market capitalization/GDP 400 350 300 250 200 150 100 50 0 Argentina Russian… Mexico Indonesia TURKEY Brazil Germany Saudi Arabia India France Australia Canada Korea, Rep. Japan China United States South Africa (Figure 4), which suggests room for growth for the Turkish capital markets. Stock Traded/GDP Figure 4: Fundamental Stock Market Indicators of Selected Countries The graph shows five-year average of stock market indicators of selected countries. Data is obtained from World Bank World Development Indicators dataset, which is compiled from World Federation of Exchange and Bloomberg. 3. Data and Methodology We construct our unique dataset using various sources. Our main data source is Revenue Administration dataset, which includes the universe of all incorporated firms’ financial tables in Turkey. Data consists of 1,095,330 non-financial firms’ annual balance sheet and income statements over the period 2006-2018. Additionally, we use Credit Registry database provided by Banks Association of Turkey to Central Bank of Republic of Turkey (CBRT), which provides all firms’ monthly credit details. Besides, we use firm-level annual employment data from Turkstat, which is originally collected by Social Security Institution of Turkey. We merge all these datasets using unique tax identity number of firms. We detect listing firms over our sample period, 2006-2018 from FINNET. It is a private database including all listed firms in Turkey with their exact listing date since BIST has found. 8
11. In order to compare listed firms with the matched control firms , our CEM-DD setup is built in a seven-year time window, three years before and four years after the listing year. For robustness, we also construct alternative time windows around listing years. The details are discussed in Section 4.4. Thus, we focus the period between 2009-2015 in which 103 nonfinancial firms become listed. Also, as can be seen in Figure 1, the listings in Turkey are predominantly in this period. In addition to financial firms, we also exclude firms with inconsistent data from the analysis, such as observations with negative total assets, negative liabilities and negative fixed assets. Besides, in order to increase our matching quality, in addition to using firms’ covariates a year prior to listing, we also use two-year lags of covariates in our matching procedure. Thus, we condition firms to exist in both one and two years before the listing year. Moreover, we exclude firms that change their legal status over the sample period. These firms changed their tax identity number, which makes it impossible to be traced. As a result, we identify 81 nonfinancial listing firms over the sample period. Descriptive statistics for listed and unlisted firms are presented in Table 1. In the first place, listed firms appear to be larger than unlisted firms in terms of sales and employment. This is not surprising due to size criterion to become listed in Turkey. Besides, substantial differences can also be seen across listed and non-listed firms for the rest of variables, such as tangible assets, total financial debt and liabilities. Thus, in order to control for observable differences between listed and privately held firms, we employ a matching procedure following the prior literature (e.g., Saunders and Steffen, 2011; Michaely and Roberts, 2012; and Gao et al., 2013). A reliable control group can reduce the bias in estimating the listing effect and allows to reduce the likelihood of confounding when analyzing the nonrandomized and observational data (Haukoos and Lewis, 2015). 9
12. Table 1 : Descriptive statistics and balancing tests for listed and matched control firms Mean T test Median Pearson χ2 Wilcoxon RankSum Test Covariates Listed Unlisted t p-value Listed Unlisted χ2 p-value z p-value Unmatched 17,16 10,49 -11,04 0,00 17,13 12,43 75,11 0,00 -13,98 0,00 Matched 16,80 16,78 -0,04 0,97 16,97 16,79 0,03 0,87 -0,09 0,93 Unmatched 4,55 1,47 -22,98 0,00 4,76 1,10 81,57 0,00 -13,90 0,00 Matched 4,38 4,35 -0,11 0,91 4,52 4,51 0,03 0,87 -0,03 0,98 Unmatched 13,27 4,29 -13,88 0,00 15,52 0,00 74,33 0,00 -12,79 0,00 Matched 13,02 12,85 -0,17 0,86 15,36 15,20 0,03 0,87 -0,27 0,79 Tangible fixed assets (Gross) (Pre_1) Unmatched 16,23 9,55 -13,06 0,00 16,47 10,91 75,11 0,00 -13,96 0,00 Matched 15,95 15,87 -0,24 0,81 16,25 16,07 0,03 0,87 -0,23 0,82 Tangible fixed assets (Net) (Pre_1) Unmatched 15,49 8,93 -12,71 0,00 16,11 10,34 71,31 0,00 -13,38 0,00 Matched 15,19 15,01 -0,40 0,69 15,95 15,65 0,71 0,40 -0,51 0,61 Total Liability (Pre_1) Unmatched 16,82 11,65 -14,39 0,00 16,67 12,21 75,11 0,00 -14,03 0,00 Sales (Pre_1) Employment (Pre_1) Financial debt (Pre_1) Matched 16,49 16,27 -0,71 0,48 16,27 16,29 0,03 0,87 -0,39 0,70 Cash & cash equivalents (Pre_1) Unmatched 13,78 9,22 -14,49 0,00 13,70 9,37 60,50 0,00 -12,26 0,00 Matched 13,46 13,22 -0,62 0,54 13,35 13,44 0,03 0,87 -0,36 0,72 Operating income (Pre_1) Unmatched 11,26 0,16 -14,45 0,00 2,16 0,004 30,86 0,00 -9,18 0,00 Matched 5,60 4,78 -0,34 0,73 1,53 0,91 2,31 0,13 -0,88 0,38 Inventories (Pre_1) Unmatched 14,24 8,11 -9,65 0,00 15,47 10,62 44,45 0,00 -11,55 0,00 Matched 13,77 13,31 -0,56 0,58 15,22 14,97 0,03 0,87 -0,42 0,67 Unmatched 16,36 10,75 -9,75 0,00 16,89 12,47 60,49 0,00 -12,61 0,00 Matched 15,91 16,42 0,90 0,37 16,64 16,60 0,03 0,87 0,28 0,78 Unmatched 4,42 1,50 -21,88 0,00 4,61 1,10 72,86 0,00 -13,44 0,00 Matched 4,23 4,25 0,07 0,94 4,31 4,43 0,00 1,00 0,08 0,94 Unmatched 13,65 4,32 -14,51 0,00 15,04 0,00 84,93 0,00 -13,65 0,00 Matched 13,52 12,73 -0,84 0,40 14,99 14,87 0,03 0,87 -0,57 0,57 Tangible fixed assets (Gross) (Pre_2) Unmatched 16,00 9,73 -12,87 0,00 16,14 10,94 75,11 0,00 -13,78 0,00 Matched 15,70 15,69 -0,05 0,96 15,88 15,94 0,03 0,87 -0,10 0,92 Tangible fixed assets (Net) (Pre_2) Unmatched 15,22 9,08 -12,38 0,00 15,58 10,37 71,31 0,00 -13,18 0,00 Matched 14,88 14,96 0,17 0,86 15,46 15,43 0,03 0,87 -0,16 0,87 Total Liability (Pre_2) Unmatched 16,69 11,65 -14,78 0,00 16,48 12,14 75,11 0,00 -14,10 0,00 Matched 16,36 16,07 -0,97 0,33 16,33 16,04 0,26 0,61 -0,76 0,45 Cash & cash equivalents (Pre_2) Unmatched 13,74 9,23 -15,01 0,00 13,69 9,33 67,61 0,00 -12,67 0,00 Matched 13,42 13,10 -0,84 0,40 13,23 13,54 0,26 0,61 -0,60 0,55 Operating income (Pre_2) Unmatched 3,59 0,15 -4,62 0,00 1,56 0,01 28,44 0,00 -8,73 0,00 Matched 3,77 3,88 0,07 0,94 1,29 0,44 1,40 0,24 -0,84 0,40 Inventories (Pre_2) Unmatched 14,10 8,25 -9,41 0,00 15,16 10,68 47,46 0,00 -11,43 0,00 Matched 13,67 13,02 -0,80 0,42 15,02 14,67 0,26 0,61 -0,53 0,60 Sales (Pre_2) Employment (Pre_2) Financial debt (Pre_2) This table presents descriptive statistics for the full samples of listed and non-listed firms and matched sample over the period 2009-2015. Pre_1 and Pre_2 denote for one and two years before the listing, respectively. See Section 3 for the details of how we construct the full sample and details of our matching procedure. The table reports means, medians and balance tests of the key variables used in our empirical analysis. Operating income is in millions TL while rest of the variables are in natural logarithms. 10
13. We conduct one to one matching via Coarsened Exact Matching (CEM) algorithm of Iacus et al. (2012). This methodology has become widely used in recent years in different fields of economics such as labor (Mattson, 2019) and corporate finance (Butler et al., 2019). The advantage of CEM is that it provides flexibility and simplicity to users to arrange imbalance level manually. As a member of monotonic imbalance bounding method family, imbalance level in CEM is chosen ex-ante. Covariates are split into some intervals and then coarsens to different stratas. It also offers superior computational efficiency for large datasets.1 Following the related literature (e.g., Gao et al., 2013; Asker et al., 2015), we use firm size and industry in our matching specification. It is important to control them in our analysis since Gala and Julio (2011) report significant evidence that corporate investment increases in firm size while Jorgenson (1971) shows that it varies across sectors. We use sales and employment as proxies of firm size and Nace-Rev 2 two digits for the industry classification. Constructing a reliable control group is essential for impact analysis (Haukoos and Lewis, 2015). Thus, in addition to firm size and industry, we match firms on rich criteria set such as tangible fixed assets, cash holdings, both financial and total debt structures, revenues, and inventories. Besides, in addition to one year before the listing, we match the variables in two years before the listing as well. We ended up with an 87% matching rate with 980 firm-year observations including treatment and control firms for our main analysis. Our novel dataset consisting of entire population of incorporated firms allows us to select control firms from a large sample, which enables us to find comparable control firms. The significant improvement in the gap between listed firms and the control firms via the matching procedure is presented in Table 1. The mean of listed (treatment) firms and unlisted (control) firms are reported in columns 1 and 2, respectively. The mean of sample groups converges considerably after matching process. This is also the case for the medians (columns 5 and 6). In order to test statistically whether control firms share similar distributions with the listed firms, first standard t-test for means is adopted. None of the covariates have significant differences between listed (treatment) firms 1 See Stuart (2010) for detailed information for Coarsened Exact Matching methodology. 11
14. and matched control firms (columns 3 and 4). This is also the case for Pearson χ2 test for medians (columns 7 and 8). Distributions can still be different even their mean and medians are equal. For this reason, we also conduct Wilcoxon test to show whether the distributions are practically same (Gao et al. 2013). Results reported in the last column of Table 1 reveals that for all covariates, null hypothesis that distributions across treatment and matched control firms are same cannot be rejected as p-values after matching is substantially high. After the matching, none of the covariates appears to be unbalanced, and all these balancing tests of the matched listed and control unlisted samples confirm the matching quality. After constructing the matched listed and control groups, we employ a difference in differences (DD) setup in order to compare these groups and to estimate the impact of listing on firm outcomes (Caliendo and Kopeinig, 2008). The DD methodology is a quasi-experiment design for measuring average treatment effect (ATE) over time between balanced control and treatment samples, which is used in previous research extensively.2 The econometric specification used in this study is given below: