1
The effect of working capital on fixed investment: a comparison of listed
and unlisted firms in the United Kingdom
Kevin Braamhaar - s2352540
University of Groningen, Faculty of Economics and Business MSc Finance
Supervisor: dr. J.H. von Eije
2
1.
Introduction
This study analyses the influence of the effect of working capital on firm investments in the United Kingdom (UK) for listed and unlisted firms. Working capital is very important to firms. It is defined as the difference between a firms’ current assets (accounts receivable, inventories and cash & cash equivalents) and current liabilities (accounts payable and short term debt) at the end of each year (Fazzari and Petersen, 1993). It represents the source and use of short term capital and it is used to measure a firm’s liquidity, which is an indicator that a firm is able to meet its short-term obligations. Insufficient liquidity can lead to bankruptcy. Yet, too much liquidity can be detrimental for firms’ profitability (Ding et al., 2013). Good management of working capital will find the optimum point in-between these two extremes.
In the extensive literature on working capital, various aspects and their effects on firm
performance are well described. Studies on working capital management fall into two competing views of working capital investment. Under one view, high levels of working capital, if caused by large inventory and generous trade credit may lead to higher sales. Large inventories reduce the risk of a stock-out (Deloof, 2003) and protect against price fluctuations (García-Teruel & Martinez-Solano, 2007). Trade credit may boost sales because customers can check product quality before payment (Deloof, 2003) and it may also function as an inexpensive source of credit for customers (Petersen and Rajan, 1997). Alternatively, higher working capital levels require financing and financing expenses, which will increase the chance of going bankrupt (Baños-Caballero et al., 2014). Also, firms that are less profitable will take more time to pay their suppliers, resulting in a negative relationship between performance and working capital. However, in this case it is profitability that has an impact on the level of working capital whereas the other examples show relationships in which the level of working capital can affect
profitability.
This study will focus on the role working capital has on fixed investment as a fund and storage for liquidity. To my knowledge, this role of working capital is often ignored by literature,
although described by Fazzari & Petersen (1993) and Ding et al. (2012). In this study, I will look at the impact of working capital on fixed investment. This is done in a two stage regression in which the estimated value of working capital investment is used because it may be a decision value for the firm.
3
This study finds that working capital is used to smooth fixed investment by both listed
(unconstrained) and unlisted (constrained) firms. However, the extent to which listed firms use working capital to smooth their fixed investment is significantly smaller than for unlisted firms. These findings indicate that listed firms are financially constrained as well to some extent, although they are less constrained than unlisted firms. Furthermore, when looking at the effect of different components of working capital, I find that the change in cash holdings and short term debt, as well as stocks, accounts receivables and accounts payables, are used by both listed and unlisted firms to smooth their fixed investments. Although, again, listed firms apply less smoothing using these two subdivisions of working capital compared to unlisted firms.
This study adds to the existing literature is that it looks at listed as well as unlisted firms. Recent studies have shown that financial constraints can be attributed to whether a firm is listed or not. This study shows that listed firms are much less financially constrained than unlisted firms. This study will therefore be less sensitive to this new few on financial constrainedness.
The remainder of this paper is structured as follows. Section 2 will discuss relevant findings in the existing literature on working capital, financial constraints and investment smoothing. In section 3, the empirical model will be described. In section 4, the data will be described. In section 5, results will be shown and a robustness check will be presented. Finally, section 6 will conclude this paper.
2.
Literature review
In perfect capital markets, the level of working capital will not influence firm performance or behaviour as it would be possible to acquire external finance without any costs (Modigliani and Miller, 1958). However, according to Fazzari & Petersen (1993) real investment does depend on financial factors. Attracting external finance, if available at all, may be costly due to transaction costs, agency problems or information asymmetry. Asymmetric information is the biggest of these problems. It describes the situation in which “insiders” have specific information about a firm’s quality and prospects. Whereas “outsiders” cannot determine the quality and prospects for a firm on an individual basis but only for a population of firms. Stiglitz and Weiss (1981) state that this leads outside lenders to ration credit when firms issue debt which then leads to
increasing interest rates. A higher interest rate can cause more safe firms to leave the debt market (adverse selection) or it may cause firms to undertake the more risky projects with higher returns due to limited liability (moral hazard).
4
shareholders. They find that firms can best raise capital by retaining earnings, an internal source of funds. Second best is to raise external capital by using the bonds market. The third option is issuing new shares. This is less favourable because Myers and Majluf (1984) find that if managers have superior information, and stock is issued to finance the investment, stock price will fall, other things equal. Although lower stock prices are harmful for existing shareholders, the action is still in their interest because of the positive NPV of the project. However, part of the NPV is lost to new shareholders because share prices do not yet reflect the value of a new
investment. If the firm issues safe debt to finance investments, which is default-risk free, the stock price will not fall. So this would also be better than issuing shares.
2.1. Financial constraints
In this study, financial constraints, describe factors that make it harder for firms to acquire external capital. Most studies determine whether a firm is financial constrained based on indicators such as firm size, information on whether they have a bond rating and/or access to commercial paper, dividend pay-out ratio (Almeida, 2004;Guariglia, 2009), the Kaplan-Zingales index (Almeida, 2004) or age (Guariglia, 2009). These criteria are proxies for the degree of external financial constraints faced by the firms due to information asymmetry. Smaller and younger firms face more information asymmetry effects since little public information is available on them, and it is more difficult for financial institutions to gather this information. Therefore, finding external finance is relatively more costly for these firms. Furthermore, firms with a low dividend pay-out ratio will be prone to moral hazard and adverse selection problems (Guariglia, 2009). Dividends signal the wellbeing of a firm in a world with asymmetric
information, so high dividend pay-outs signal good future prospects.
In a recent paper, Farre-Mensa & Ljungqvist (2016) criticize popular measures identifying financially constrained firms. They look at paying dividends, having a credit rating, the Kaplan-Zingales, Whited-Wu, and the Hadlock-Pierce indices. They find that none of these measures identify firms that behave as if they were constrained. However, they do find that privately held firms, in particular small ones, and listed firms with below investment grade ratings seem to be financially constrained. Therefore, and due to the lack of credit rating availability on firms, this study will mark unlisted firms as financially constrained and listed firms as financially
unconstrained. The study of Farre-Mensa & Ljungqvist (2016) can be seen as a point of critique on the study of Fazzari & Petersen (1993). The latter look at a sample of listed firms, denoting no dividend paying firms as financially constrained and dividend paying firms as unconstrained. Given that I look at listed as well as unlisted firms, this study is less sensitive to the criticism of Farre-Mensa & Ljungqvist (2016) on constrainedness.
5
the relative thin and highly regulated banking and equity markets. And, third, the relatively small amount of venture capital financing in Europe compared to the US.
2.2. Investment smoothing
Fazzari & Petersen (1993) give two reasons why firms would want to smooth their investments. First, because firms face rising marginal adjustment costs when their rates of investment
increase. In order to reduce the long-run costs for their capital accumulation, firms will try to keep a stable investment pattern over time. Casalin & Dia (2014) also find strong evidence for these adjustments costs in capital accumulation which arise from financial frictions. They argue that these costs cause firms to smooth investment over time, not only to minimize adjustment cost on the capital of stock, but also to minimize the cost of adjusting the stock of external finance.
The second reason why firms want to keep a smooth investment pattern is because firms cannot store or delay investments without costs. Fazzari & Petersen (1993) describe how firms have to invest constantly because own-firm innovation and innovation spill-overs from other firms generate new investment opportunities constantly. If these opportunities are not undertaken at the moment they are observed, the possible value in these investments will rapidly decrease because of short product life cycles and first-mover advantage from commercializing new technologies. Myers and Majluf (1984) also indicate that the full investment opportunity evaporates if a firm does not go ahead at the time of recognition. Or in other words, the delay of investments reduces the project’s net present value, so taking on investments immediately gives the highest value for the firm.
2.3. The role of working capital
In modern firms, working capital often is very often quite large compared to fixed capital. In some manufacturing firms, working capital is even more than half as large as the fixed capital stock. (Fazzari & Petersen, 1993). The most general description of working capital is current assets minus current liabilities. Current assets consists of accounts receivable, inventories and cash and cash equivalents. Inventories are then further divided into materials, work-in-process, and finished goods. Current liabilities consist of accounts payable and debt due in less than one year.
6
through the liquidity of the firm. In that way cash balances will reduce financing costs and early payment of accounts payables will result in discounts.
The main difference between fixed capital and working capital is that working capital is much more liquid. Non-cash working capital can be reversed into cash more easily than fixed capital. For example when firms consume raw materials or finished goods faster than they are replaced, non-cash working capital investments become negative. Also when the accounts receivables collection period is shortened or the collection itself is intensified, cash is freed up. Meltzer (1960) also finds that current assets are used as collateral for short-term borrowing, reducing non-cash working capital through increasing current liabilities. This is another way to generate cash out of it the non-cash working capital. By increasing short term debt, current liabilities increase, generating cash to be used where needed (e.g. financing fixed investment) .
2.4. Fixed investment smoothing with working capital
Fazzari & Petersen (1993) describe how firms facing a binding finance constraint may not be able to equate the discounted marginal rate of return on assets across time. So constrained firms may have a (temporary) difficulty to equate marginal returns on investment to the market cost of capital. However, firms can compare marginal returns across different assets using a shadow value of finance at each moment in time. Fazzari & Petersen describe a situation where there is a negative shock in cash flow, other things constant. The shadow costs of finance will rise for financially constrained firms, and they may have to respond by reducing their rate of asset accumulation. However, to equate returns across assets, firms should not cut investments proportionately in both fixed and working capital, because working capital is relatively liquid. This means that firms will try to avoid adjustment costs of fixed investments by choosing working capital investment to absorb a larger share of temporary cash-flow fluctuations than they choose to do with fixed investments. Because working capital is reversible, it can thus become a source of funds when firms choose to disinvest in working capital. This source of funds can relax a firm’s short-run financing constraints and helps the firm to be able to use fixed investment opportunities.
The amount of “smoothing” that is applied is likely to depend on a firm’s initial stock of working capital. The higher the level of working capital, the lower its marginal value to the firm and the more willing the firm will become to offset negative shocks to cash flow by forgoing or reducing working capital investment. However, if the stock of working capital is too low, its marginal value to the firm will be higher than when it is high so in this case less fixed-investment
smoothing is expected. This implies a positive relationship between the level of working capital and investment smoothing. In other words a positive coefficient can be expected because
7
sense, the strength of a firm’s balance sheet, measured by its working capital level, affects the link between fixed investment and cash flow.
Fazzari and Petersen (1993) also distinguish between internal finance shocks and external demand shocks. Both of them may have different consequences for constrained and
unconstrained firms. If there is a negative cash flow shocks due to an increase in fixed costs, unconstrained firms’ investment is unaffected while constrained firms will adjust their
investment in both fixed and working capital along the lines described above. When a demand shock occurs, things get more complicated. Due to the lower demand for goods, production will be lower. Lower production reduces desired work-in-process and materials inventories.
Furthermore, the drop in sales will cause lower accounts receivables and lower cash holdings. Finally the finished goods stock will rise temporarily, but will stabilize soon after production is adjusted to demand. The influence of a drop in demand described above holds for constrained as well as unconstrained firms. However, for financially constrained firms the demand shock has an additional impact. If they choose to smooth fixed investment, working capital is expected to drop even further.
2.5. Economic background in the UK
Campello et al. (2009) describe how the last decade is mainly marked by the global credit crisis of 2008. Firms link the availability of funds to the ability to pursue investment opportunities. Due to the crisis, credit is drained from the market, making less finance available. Constrained firms suffer more from this crisis than unconstrained firms. This is confirmed by the fact that constrained firms made deeper cuts in tech spending, capital spending and employment. They also used more of their cash balance, drew more heavily on lines of credit as they feared future restriction to capital by banks and they sold more assets to fund operations. This should
underline the relevance of being or not being financially constrained.
2.6. Hypotheses
Three predictions can be made when considering firms’ investments in fixed and working capital when firms face financial constraints. First, according to previous literature, fixed investment depends heavily on previous year’s cash flow for unlisted firms. This effect is expected to be positive, so if cash flows in year t-1 go up (down) year t’s fixed investments are expected to go up (down) as well. Based on evidence provided by the previous literature I hypothesize that: Hypothesis 1: Previous year’s cash flows have a positive effect on investment
8
Secondly, working capital investment will have a negative relationship with fixed-investment when included as an endogenous variable for unconstrained firms. This negative relationship can be explained because working capital competes with fixed investment for the limited pool of finance in a firm. So, when a firm chooses to decrease (increase) working capital investments, fixed investments should rise (fall).
Hypothesis 2: Working capital investment has a negative effect on fixed investment
Lastly, the literature shows that financially constrained firms rely relatively heavily on internal funds (from cash flows) compared to financially unconstrained firms. The pool of capital in listed for unconstrained firms is less limited since they can attract additional finance to meet investment goals. Therefore unlisted firms should apply less investment smoothing using working capital. So, if hypothesis 2 cannot be rejected, hypothesis 3 can be tested: Hypothesis 3: Working capital investment has a stronger effect on fixed investment
in constrained firms than in unconstrained firms
3.
Methodology
The start of this section will show the most conventional regression of cash flows on fixed investment. Then, an equation will be presented that also usable for unlisted firms and less prone to measurement errors. For this equation, fixed, working capital and total investment (fixed plus working capital) regressions will be done. Then a fixed investment regression that controls for the smoothing role of working capital will be done. This regression will then be modified to analyse the difference between listed and unlisted firms.
3.1. Conventional regression of investment on cash flow
To test hypothesis 1, an investment equation needs to be formulated. At first, this section will focus on the effect of cash flow on fixed investment. Previous neoclassical models were only based on current year and previous year cost of capital or sales. Looking only at current year and past year’s variables, these models failed to include variables that were able to tell something about future investment opportunities. Fazzari & Petersen (1993) use a function based on the Q model (from Tobin’s q) similar to the following:
(𝐾𝐼)𝑗,𝑡= 𝛾0+ 𝛾1𝑄𝑗,𝑡−1+ 𝛾2(𝐶𝐹𝐾)
9
The cash flow coefficient 𝛾2shows how much fixed investments depend on internal funds. Therefore, it can be seen as an indicator of the degree of financing constraints faced by firms. In the case of a negative cash flow shock, a financially constrained firm will in fact be forced to reduce or postpone their investments. Therefore, 𝛾2 is expected to be positive. Unlisted firm’s fixed investment will be less dependent on cash flow, as they are expected to be less financially constrained, leaving a smaller coefficient.
Tobin’s q is defined as the ratio of market values of equity and debt over the replacement value of the firm’s capital stock. (Fazzari & Petersen, 1993). The idea of the Q model is that an
investment will be profitable if it adds more to the firm’s market value than it costs to undertake it. These profitable investments will be higher if a firm’s actual capital stock is below its optimal capital stock. Investments are therefore likely to go up when the market valuation of the firm’s capital is high relative to its replacement costs, which makes issuing shares profitable (Bond and Meghir, 1994). 𝛾1indicates how much the Q ratio affects investments. It is expected to be positive since a Q ratio above one makes investment in the firm profitable, whereas a Q ratio of lower than one makes investments to generate negative net present value.
Including the q-ratio, however, comes along with various problems according to Goergen & Renneboog (2001). First, Tobin’s q is difficult to measure. As the denominator, the replacement value of assets is not reported in most European countries. The numerator, the book value of assets, is also hard to estimate due to the measurement of intangibles. Second, Tobin’s q will only include future expectations if the firms is a price taker in perfectly competitive industries with constant returns to scale and if the stock market value correctly measures the present value of the expected future cash flows of the firm. In practice these conditions may not be fulfilled for example when stock markets display excessive volatility relative to fundamental value of the companies. A third problem is that the q-ratio can only be calculated for listed firms as it . Ding et al. (2012) look at listed as well as unlisted firms and therefore they have to leave out the Q ratio. This study looks at unlisted and listed firms as well, which means the q-ratio has to be left out as well.
10 (𝐾𝐼)𝑗𝑡 = 𝛾0+ 𝛾1(𝐾𝐼)𝑗,𝑡−1+ 𝛾2(𝐾𝐼)𝑗,𝑡−12 + 𝛾
3(𝐶𝐹𝐾)
𝑗,𝑡−1+ 𝛾𝑗 + 𝛾𝑡+ 𝑢𝑗𝑡 (2) all variables is this equation are already mentioned in the description of equation (1).
The investment rate of period t-1 and the investment rate squared of period t-1 are included. The reason for the lagged investment rate is that the rate of investment in the previous period is positively related to the rate of investment in the next period because firms want to keep their investment rates stable over time. Therefore, in the absence of financial constraints, Bond and Meghir (1994a) argue that 𝛾1is assumed to be positive and bigger than one. Firms that are financial constraint will show different investment behaviour. Therefore, 𝛾1 will be smaller if a firm is financially constraint. Bond and Meghir (1994a) also state that the second term, the one year lagged investment rate squared, reflects the assumed quadratic form of the adjustment cost function. The assumed coefficient is negative and less than minus one since the function is expected to have an inversed U-shape because of the assumption that there is an optimal rate of investment. Furthermore 𝛾3 is expected is to be positive since positive cash flows in the previous year should increase interest rates in the next. However, for unconstrained firms the coefficient is expected to be smaller than for unconstraint firms since these firms depend less on internal funds.
The same formula will be used to look at investment in working capital and total investment (fixed investment plus working capital investment). I do this, to test whether this new model has predictive power that is comparable to the q model used by Fazzari & Petersen (1993). Although the dependent variable changes in these equations, the lagged term of fixed investment and fixed investment squared will remain the same for consistency reasons and because no stable
relationship is assumed in working capital investment.
To focus on the effect of cash flow on working capital investment, the dependent variable in equation (2) needs to be replaced. As working capital faces lower adjustment costs then fixed capital, firms should adjust working capital earlier than fixed capital in the presence of
fluctuations in cash flows. This way, fixed investments can be kept high and be smoothed. To test this, equation (2) will be regressed again. However the investment terms in the dependent variable and in the independent variables are replaced by working capital investment (∆𝑊𝐶𝐾 ) leading to equation (3):
(∆𝑊𝐶𝐾 )𝑗,𝑡 = 𝛾0+ 𝛾1(𝐾𝐼)𝑗,𝑡−1+ 𝛾2(𝐾𝐼)𝑗,𝑡−12 + 𝛾 3(𝐶𝐹𝐾)
11
A stronger effect is expected since working capital investment is expected to increase (decrease) rapidly when rises (drops) in cash flow occur. Furthermore, if working capital is used to smooth investment in financially constrained firms, working capital should drop even harder in order to smooth investment. Also, a symmetric effect can be expected when cash flows go up (Fazzari & Petersen, 1993).
Finally, a comparable equation is also used to estimate the impact of total investment. Again, the investment variables are replaced, this time by (𝐾𝐼 +∆𝑊𝐶𝐾 ) which is fixed investment plus
working capital investment.
(𝐾𝐼 +∆𝑊𝐶𝐾 )𝑗,𝑡 = 𝛾0+ 𝛾1(𝐾𝐼)𝑗,𝑡−1+ 𝛾2(𝐾𝐼)𝑗,𝑡−12 + 𝛾3(𝐶𝐹𝐾)
𝑗,𝑡−1+ 𝛾𝑗+ 𝛾𝑡+ 𝑢𝑗𝑡 (4) For the coefficients in equation (4) (which is a combination of equation (2) and (3)) the expected coefficients are also expected to have a combined effect. So a stronger effect is expected when two positive or two negative effects are combined and a weaker effect is expected when a positive and a negative effect are combined.
Nevertheless, these models (2) to (4) probably also underestimate the effect of cash flows on investments as working capital investment is left out as suggested by Fazzari & Petersen (1993) and their Q model. The models should be seen as models that look for the effect of cash flow on investment after the firm already took investment smoothing measures. The next section will show the model that controls for investment smoothing with working capital undertaken by the firm.
3.2. Regressions with working capital as a source and use of funds
From now on, the change in working capital (or working capital investment) (∆𝑊𝐶) is included in regression (1). However, including this variable exogenously in the regression would cause an endogeneity problem since the change in working capital is a decision variable for the firm (Fazzari & Petersen, 1993). Therefore, (∆𝑊𝐶𝐾 )
𝑗,𝑡 is estimated with instrumental variables. The instruments are previous year’s fixed investment (𝐾𝐼)𝑗,𝑡−1, previous year’s fixed investment squared (𝐾𝐼)𝑗,𝑡−12 previous year’s cash flow (𝐶𝐹𝐾)𝑗,𝑡−1 and previous year stock of working capital divided by fixed capital (𝑊𝐶𝐾 )
12 (𝐾𝐼)𝑗𝑡= 𝛽0+ 𝛽1(𝐾𝐼)𝑗,𝑡−1+ 𝛽2(𝐾𝐼)𝑗,𝑡−12 + 𝛽 3(𝐶𝐹𝐾) 𝑗,𝑡−1+ 𝛽4( ∆𝑊𝐶 𝐾 ) ̂ 𝑗,𝑡+ 𝛽𝑗+ 𝛽𝑡+ 𝑢𝑗𝑡 (5) which is similar to equation (2) except for the fact that change in working capital is now
included. In other words, coefficient 𝛽1 to 𝛽3 are similar to the expectations of 𝛾1to 𝛾3in equation (2). For the change in working capital, coefficient 𝛽4 is expected to be negative since firms in can invest more (less) in fixed assets if working capital decreases (increases).
In equation (6) working capital investment (̂∆𝑊𝐶𝐾 )
𝑗,𝑡 is estimated using: (∆𝑊𝐶𝐾 )𝑗,𝑡 = 𝛿0+ 𝛿1(𝐾𝐼)𝑗,𝑡−1+ 𝛿2(𝐾𝐼)𝑗,𝑡−12 + 𝛿
3(𝐶𝐹𝐾)𝑗,𝑡−1+ 𝛿4(𝑊𝐶𝐾 )
𝑗,𝑡−1+ 𝜃𝑗 + 𝜃𝑡+ 𝑢𝑗𝑡 (6) which is similar to equation (3) except for the fact that the level of working capital is added. This means expectations for 𝛿1 to 𝛿3 are similar to the expectations for 𝛾1to 𝛾3 in equation (3).
Fazzari & Petersen (1993) describe the coefficient of the level of working capital , 𝛿4 is expected to be negative. This, because the choice of a firm to invest in working capital should depend negatively on its stock of working capital as the marginal value of working capital falls as its stock rises. In other words, one can say that a firm with a high (low) stock of working capital is more likely to disinvest (invest) in its working capital. The outcomes of these first stage
regressions are then uses as input for the second stage regressions.
3.3. The difference between listed and unlisted firms
The two stage regression is then modified with dummy variables in order to explore a difference between investment smoothing with working capital between listed and unlisted firms. This results in equation (7) which is derived from equation (5):
13 (∆𝑊𝐶𝐾 )𝑗,𝑡 = 𝛿0+ 𝛿1∗ 𝑙𝑗,𝑡+ 𝛿2(𝐾𝐼) 𝑗,𝑡−1+ 𝛿3( 𝐼 𝐾)𝑗,𝑡−1∗ 𝑙𝑗,𝑡+ 𝛿4( 𝐼 𝐾)𝑗,𝑡−1 2 + 𝛿5(𝐾𝐼) 𝑗,𝑡−1 2 ∗ 𝑙𝑗,𝑡+ 𝛿6(𝐶𝐹𝐾)𝑗,𝑡−1+ 𝛿7(𝐶𝐹𝐾)𝑗,𝑡−1∗ 𝑙𝑗,𝑡+ 𝛿8(𝑊𝐶𝐾 )𝑗,𝑡−1+ 𝛿9(𝑊𝐶𝐾 ) 𝑗,𝑡−1∗ 𝑙𝑗,𝑡+ 𝜃𝑗+ 𝜃𝑡+ 𝑢𝑗𝑡 (8) in this equation, every variable shows up with a dummy variable each time which makes this equation specific for listed and unlisted firms. This is necessary because of the expected difference between listed and unlisted firms’ investment smoothing. If a standard regression would have been used in the first stage to estimate working capital investment, biased estimates of working capital investment would be plugged into the second stage of the least squares regression. Note that, although all independent variables are chosen with a one year lag, the dummy variable is shown for period t. I did this because I assume that working capital
investment in period t depends on whether the firm is listed in period t and not on its status in period t-1.
4.
Data
The dataset for this study shows panel data of UK manufacturing firms with total assets higher than 500 million GBP over a period of 2008-2016. Panel data has some advantages in
combination with fixed effects regressions. First of all, when doing fixed investment regressions, heterogeneity is expected as one firm is expected to invest more than another. Also, the time period is expected to have influence on investment as investment opportunities arise at certain moments and economic trends vary over time. Panel data allows to do regressions controlling for these effects that differ per time period for all firms and effects that differ per firm over the sample period. This way, the risk of unobservable heterogeneity related to firm characteristics is not included in the model (Baños-Caballero et al., 2014).
Because this study focusses on investments of firms and financial constraints, the data is drawn from the Orbis database by Bureau van Dijk. The advantage of using this database is that it contains unlisted as well as listed firms. The importance of including unlisted firms is described in section 2.1. The Orbis data is combined with data from the Zephyr database, also by Bureau van Dijk, to find initial public offering (IPO) dates of the firms. Because, if firms are listed or delisted in the period of 2009-2016, their indicator will switch from listed to unlisted or vice versa for the year in which they switch. This is why the listed dummy is also time specific. To control for the potential influence of outliers, observations are winsorized at a 1% level. As mentioned before, distinction between constrained and unconstrained firms will be based on whether firms are listed or not.
14
firms. There are about twice as many observations of unlisted firms vs listed firms (2169 vs 318) . These numbers represent reality in the sense that there are more unlisted firms that listed firms in the UK. Furthermore, regressions will always be done on separated sample or using
regressions with dummy variables to avoid unlisted firms having a bigger impact on the results than listed firms due to more observations.
Table 1 presents descriptive statistics of the variables used in this study. Firms are categorized into the group listed or unlisted. The mean and median values suggest that the investment ratios: (I/K), (ΔWC/K) and (ΔWC/K)+(I/K) for listed firms are still effected by outliers after deleting winsorizing the observations in 1% of their tails. Looking at cash flow to fixed capital medians there is a small difference (14.1% versus 10.8%) . Minimum values are exactly the same for listed and unlisted firms for all variables. This is caused by the winsorizing the 1% tails of the observations. It shows that all firms had minimum values in these tails. For the maximum values this only holds for fixed investment and the level of working capital. The other variables show different maximum values for listed and unlisted firms. Second, fixed investment is higher in listed firms compared to unlisted firms. This is also in line with the expectations as listed firms can acquire capital more easily, which means they can invest more.
Second, listed firms are listed for a reason, since it is a choice to go public. Which basically means that firms that are in a line of business which forces them to invest more and acquire big amounts of capital, possibly have made the choice to go public at some point in history.
Descriptive statistics of 2669 UK firms, 2009 – 2016: 15753 observations
Variable Mean Median Maximum Minimum Std. Dev.
Cash Flow / Fixed Capital
(CF/K)
Listed 0.235 0.141 12.172 -3.531 0.701
Unlisted 1.040 0.108 40.832 -3.531 5.004
Fixed Investment / Fixed Capital
(I/K)
Listed 0.035 0.037 1.000 -17.836 0.543
Unlisted -0.161 0.009 1.000 -17.836 1.637 Working Capital Investment /
Fixed Capital (ΔWC/K)
Listed 0.058 0.008 42.859 -33.463 1.413
Unlisted 1.200 0.011 119.755 -33.463 12.236 Total Investment / Fixed Capital
(ΔWC/K)+(I/K)
Listed 0.039 0.058 9.804 -72.472 2.156
Unlisted 0.132 0.044 60.790 -72.472 9.728 Working Capital / Fixed Capital
(WC/K)
Listed 0.044 0.004 9.807 -8.218 0.675
Unlisted 0.058 0.000 9.807 -8.218 1.826
15
Investment in working capital (ΔWC/K) is close to zero for listed firms as well as for unlisted firms. This can be explained because firms probably like to keep working capital stable in order to not disturb the business side (production function) on the one hand, not pulling on trade credit or (dis)investing inventories on one side or having to attract more capital to growing working capital due to sales growth. Investment and disinvestment in working capital should therefore only have a short run impact as firms are unlikely to be able to disinvest heavily in working capital year after year. Total investment (ΔWC/K)+(I/K) is quite similar to fixed investment in working capital which is logical given the fact that it is the sum of fixed investment and working capital investment, and given that the latter is almost equal to zero. Finally the level of working capital is close to zero, this is caused by the fact that there are firms that operate with negative working capital, so current liabilities exceed current assets in these firms.
5.
Results
First results for the regression are shown that examines the link between investment and cash flow without controlling for investment smoothing using working capital for listed as well as unlisted firms. Then, the two stage regression results controlling for investment smoothing are shown. Finally the results of the regression examining the difference between listed and unlisted firms are presented.
5.1. The effect of cash flow on investment
Listed firms are not expected to smooth investments using working capital as they are expected to be financially unconstrained. However, results of these firms will be shown for comparison reasons and because the last hypothesis aims at the difference in constrainedness between the two firm types.
16
Regression for Various Components of Investment Independent
Variable
Equation (2) Equation (3) Equation (4)
Dependent variable: Dependent Variable: Dependent Variable: Fixed Investment (I/K) Investment in Working
Capital (ΔWC/K)
Total Investment
(ΔWC/K)+(I/K)
Listed Unlisted Listed Unlisted Listed Unlisted
𝐶 0.001 -0.271*** -0.027 1.067*** -0.168*** -0.099 (0.069) (-17.823) (-0.900) (10.266) (-3.528) (-1.129) (𝐼/𝐾)𝑡−1 -0.192*** -0.170*** 0.559*** -0.525** 0.518*** -0.353* (-4.361) (-4.879) (5.894) (-2.194) (3.474) (-1.750) (𝐼/𝐾)𝑡−12 0.002 0.003 -0.001 -0.059*** 0.057*** -0.024* (0.514) (1.458) (-0.133) (-3.965) (4.831) (-1.931) (𝐶𝐹/𝐾)𝑡−1 0.117*** 0.051*** 0.428*** 0.177*** 0.854*** 0.147*** (3.774) (10.125) (6.370) (5.106) (8.090) (5.050) 𝑅2 0.334 0.321 0.310 0.401 0.263 0.343
Panel data cross-section and time fixed effects. The variable data are retrieved from Orbis and Zephyr including 2617 firms consisting of 2230 unlisted firms and 387 listed firms, totalling 14306 firm-year observations, within the United Kingdom for the period 2008-2016. This table provides the least squares regression output for the existence of a linear relation with firm and time fixed effects. Columns 1 and 2 show the linear relation with fixed investment as the dependent variable, columns 3 and 4 show the linear relation with working capital investment as the
dependent variable and columns 5 and 6 show the linear relation with total investment as the dependent variable. For each two columns the first column shows the regression on the sample of firms that are listed at time t, and the second columns shows the regression on the firms that are unlisted at time t. Fixed investment (I/K) is measured as the change in fixed assets in a year. Change in working capital (ΔWC/K) is measured as the change in working capital (current assets minus current liabilities) in a year. Total investment ((ΔWC/K)+(I/K)) is calculated as the change in working capital plus the change in fixed assets in a year. Cash flows in the previous year is shown by
(CF/K). Variables showing t-1 are lagged one year. Data are winsorized at a 1% level. The t-values are given in
parentheses below the coefficient. R-squared is included as goodness of fit. ***, ** and * stand respectively for statistical significance within the 1%, 5% and 10% confidence levels respectively. Fixed firm and time effects are not reported.
17
in their fixed investments. The positive relationship between cash flow and fixed investment shows that unlisted firms also rely on internal funds to finance investment as no effect would be expected if firms were completely unconstrained. One remarkable thing is that the smaller coefficient of 0.051 at 1% significance for unlisted firms could mean that firms depend less on internal funds than unlisted firms. However, since there is a recession in the time window of the data, it might also indicate that unlisted firms did not want to invest as much due to lower demand in this period.
Looking at the results from equation (3) for listed firms, there is a strong positive relationship between fixed investment in the previous period and working capital investment in the next. Also, there is a strong positive relationship between cash flows in the previous period and working capital investment (0.559). So firms investing much in fixed assets in the previous period also invest much in working capital in this period. This might be caused by the fact that sales go up in the year after an investment, causing working capital to go up as well. Also, working capital investment depends heavily on cash flows. The coefficient of 0.428 at 1% significance is much bigger than for fixed investment which is in line with theory, as working capital is easily reversible. It may also indicate that firms save up financial slack in working capital, as suggested by Myers (1984). For unlisted firms, there is a negative relationship between fixed investment in the previous year and working capital investment (-0.525). This can be caused by cash holdings that are have be freed up after fixed investment or by remainders of disinvesting being held as cash by unlisted firms. Also, working capital investment depends strongly on cash flow in the previous period (0.177). As well as for listed firms, and in line with the theory, the coefficient for working capital is bigger than for fixed capital.
Looking at equation (4), for listed firms, there is a big positive relationship (0.518) between fixed investment in the previous year and total investment. So total assets in a firm tend to go up or down year after year when firms are listed. Furthermore, firms rely heavily on internal funds for these investments shown by the big coefficient for cash flows (0.854). For unlisted firms there is no significant relationship between total investment and fixed investment in the previous year. Only the cash flow coefficient is significant, showing that unlisted firms rely heavily on internal funds generated by the firm.
5.2. The effect of investment smoothing on fixed investment
To test hypothesis 2, a two stage least squares regression was done. Table 3 shows the first-stage OLS regression for equation (6) with dependent variable (∆𝑊𝐶𝐾 ). With these coefficients,
18
First-stage OLS regression, dependent variable: (∆𝑊𝐶
𝐾 )
Independent Variable Listed Firms Unlisted Firms Dummy
Regression 𝐶 -0.099*** 1.109*** 0.934*** (-3.299) (10.592) (3.721) 𝐶 ∗ 𝑙𝑗,𝑡 -0.095 (-0.064) (𝐼/𝐾)𝑡−1 0.850*** -0.560** -0.560** (8.939) (-2.340) (-2.554) (𝐼/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 1.377* (1.742) (𝐼/𝐾)𝑡−12 0.017** -0.059*** -0.059*** (2.293) (-4.011) (-4.379) (𝐼/𝐾)𝑡−12 ∗ 𝑙 𝑗,𝑡 0.074 (1.219) (𝐶𝐹/𝐾)𝑡−1 0.425*** 0.221*** 0.221*** (6.546) (5.934) (6.480) (𝐶𝐹/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.192 (0.369) (𝑊𝐶/𝐾)𝑡−1 0.094*** -0.005*** -0.005*** (11.515) (-3.208) (-3.505) (𝑊𝐶/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.099 (1.504) 𝑅2 0.355 0.401 0.402
Panel data cross-section and time fixed effects. The variable data are retrieved from Orbis and Zephyr including 2617 firms consisting of 2230 unlisted firms and 387 listed firms, totalling 14279 firm-year observations, within the United Kingdom for the period 2008-2016. This table provides the least squares regression output of equation (6) and (8) for the existence of a linear relation with firm and time fixed effects. Columns 1 and 2 show the linear relation of equation (6) for a sample of firms that are listed and unlisted at time t respectively, column 3 shows the linear relationship of equation (8) for a sample of all firms. The dependent variable working capital investment
(ΔWC/K) is measured as the change in fixed assets in a year. The independent variables fixed investment (I/K) is
measured as the ratio of the change in fixed assets to fixed assets, fixed investment squared (I/K)2 is measured as the
ratio of the change in fixed assets to fixed assets squared, cash flows to fixed assets in the previous year is shown by
(CF/K), the dummy variable Ij,t is equal to 1 if a firm was listed in a year and zero otherwise. Variables showing t-1
are lagged one year. Data are winsorized at a 1% level. The t-values are given in parentheses below the coefficient. R-squared is included as goodness of fit. ***, ** and * stand respectively for statistical significance within the 1%, 5% and 10% confidence levels respectively. Fixed firm and time effects are not reported.
19
Kan weg
Looking at working capital investment for listed firms, again there is a positive relationship with investment in the previous period (0.850). Also the lagged investment rate squared now gives significant coefficient, however it is economically small (0.017) and with 5% significance. Looking at cash flow, again, there is a positive relationship between cash flow in the previous period and working capital investment with a coefficient of 0.425 with 1% significance. Then the level of working capital, here, a negative relationship was expected due to the expected negative marginal value of working capital. However the regression shows a positive coefficients. This indicates that if working capital is high (low) in a firm, working capital investment is also high (low) in the following year. This might be explained by the fact that since listed firms are expected to be financially unconstrained, there is no need to manage working capital adequately in order to maintain good liquidity. This enables management to build up financial slack in their working capital. Although the term is much used in a negative way in agency theory, financial slack is important to the firm as described by Myers (1984). It can be held in the form of extra cash or reserved borrowing power which means the firm is able to issue safe debt when needed. There are two reasons why this is important to a firm. First, in order not to come in financial distress, which has high costs as a consequence. Second, the financial slack can be used to finance new investment opportunities. This reduces the probability that positive NPV projects will be passed because the firm will be unwilling to finance investment using equity or debt that is does not capture the NPV of the investment due to asymmetric information.
For unlisted firms, previous period’s fixed investment, fixed investment squared and cash flows do not show very big differences in coefficients either. The level of working capital has the expected negative coefficient with 1% significance although it is rather small economically. The dummy regression column shows equal coefficients for unlisted firms. When adding up the coefficient with the dummy variable, showing the additional effect listed firms have, all coefficients except for the constant, add up to the corresponding coefficient of the listed firm regression. For example, looking at (𝐼/𝐾)𝑡−1 and (𝐼/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 -0.560 + 1.377 = 0.817 which is close to 0.850 of the listed firm regression. What can be observed here is that cash flows coefficients are positive for listed and unlisted firms (0.425 and 0.221) with 1% significance. However, they are not significantly different from one another, as the difference is 0.192 with a probability of 71.2% which is not significant.
20
Working capital investment depends negatively on fixed investment in the previous period for both listed and unlisted firms (-0.072 and -0.022) be it on the 10% confidence interval. The investment rate squared, that looks for an optimal level of investment is negative for unlisted firms (-0.028) and significant on the 1% confidence interval. So the assumption that there is an optimal level of fixed investment also has impact on working capital investment. This seems logical as this study sees investment as a pool of capital that has to be divided between fixed and working capital. For unlisted firms no evidence of an optimal level of investment influence working capital investment can be found. This may be due to the fact that unlisted firms are more financially constrained and therefore less able to chase optimal investment rates. A positive effect can be observed between cash flow and working capital investment for both listed as well as unlisted firms (0.357 and 0.023 respectively, both significant on the 1% confidence interval). This shows that both firm types depend on internal funds for working capital investment. However the effect of cash flows is much smaller for unlisted firms than for listed firms. The level of working capital has a negative coefficient. This fits the expectations in described in section 2 since firms are more likely to disinvest (invest) in working capital if the stock of working capital is high (low) in the previous
Results of this second stage regression are shown in table 4. Looking at listed firms, some coefficients have changed compared to the regression of equation 2 in table 2 now that working capital investment is added.
Using this indicator, the second stage regression was done, the results are presented in table 4.
Investment equation estimates with working capital investment included, dependent variable: (𝐼/𝐾 )
Independent
Variable Listed Firms Unlisted Firms Dummy Regression
21 (4.087) (-9.475) (-9.800) (∆𝑊𝐶/𝐾)̂ 𝑡−1∗ 𝑙𝑗,𝑡 0.395*** (4.603) 𝑅2 0.340 0.321 0.321
Panel data cross-section and time fixed effects. The variable data are retrieved from Orbis and Zephyr including 2619 firms consisting of 2232 unlisted firms and 387 listed firms, totalling 14279 firm-year observations, within the United Kingdom for the period 2008-2016. This table provides the least squares regression output for the existence of a linear relation for equation (5) and (7) with firm and time fixed effects. Columns 1 and 2 show the linear relation of equation (5) for a sample of firms that are listed and unlisted at time t respectively, column 3 shows the linear relationship of equation (7) for a sample of all firms. The dependent variable fixed investment (I/K) is measured as the ratio of the change in fixed assets to fixed assets. The independent variables fixed investment (I/K) is measured as the ratio of the change in fixed assets to fixed assets, fixed investment squared (I/K)2 is measured as
the ratio of the change in fixed assets to fixed assets squared, working capital investment (ΔWC/K) is measured as the ratio of the change in working capital in a year to fixed assets, cash flows to fixed assets in the previous year is shown by (CF/K), the dummy variable lj,t is equal to 1 if a firm was listed in a year and zero otherwise. Variables
showing t-1 are lagged one year. Data are winsorized at a 1% level. The t-values are given in parentheses below the coefficient. R-squared is included as goodness of fit. ***, ** and * stand respectively for statistical significance within the 1%, 5% and 10% confidence levels respectively. Fixed firm and time effects are not reported.
What I do see now is that the effect of cash flows in the previous period has become much smaller. This was already predicted by Fazzari & Petersen (1993) as they said that the cash flow coefficient (as shown in table 2, equation 2) would probably overestimate the effect of cash flows and underestimate the effect of internal finance on cash flow. Looking at working capital investment, which was expected to be negative, shows to be positive for listed firms. This indicates that listed firms do not face financial constraints when they want to invest. The correlation of both working capital and fixed capital going up simultaneously, might be caused by the fact equity and debt are left out in the regression. So when listed firms want to invest, they attract enough capital to invest in fixed capital while also investing in working capital
simultaneously. However, if I combine the two working capital investment coefficients of the dummy regression, the effect seems to be close to zero. This might be caused by the fact that I used panel data, leading to some differences in estimates.
22
Investment in the previous period has a negative coefficient for unlisted firms whereas for listed firms no clear relationship can be found (-0.203 versus -0.056). This again proves that unlisted firms do not seem to be investing year after year but are more likely to lower fixed investment if fixed investment in the previous period was high. Also cash flow has a negative impact on fixed investment for listed firms (-0.335). On the contrary, unlisted firms tend to increase fixed
investment if cash flow in the previous period was high although this effect is economically very small (0.049). Finally working capital investment has a positive impact on fixed investment in listed firms. Although theory was not focussing on these firms, as they are expected to be
financially unconstrained, this results seems odd. Listed firms invest in working and fixed capital simultaneously. Looking at this last variable, and also considering the other variables for unlisted firms. This suggests that listed firms make investment decisions independently from their
financial performance in the past period. For unlisted firms however, the coefficient for working capital investment is negative and significant on the 1% confidence interval. This is also what theory expected, as these firms are expected to be financially constrained. So unlisted firms disinvest in working capital in order to be able to invest in fixed capital. These findings also gives some evidence for the third hypothesis, stating unlisted firms use investment smoothing more than listed firms. However, to test accurately for this difference, results in of the last two equations (7) and (8) will be given in the next part.
The difference in investment smoothing between listed and unlisted
Moet in het onderdeel hier boven behandeld wordenTo test hypothesis three, regression (7) and (8) are now performed. This regression can be seen as an estimation that is less fit for both firm types than the regression based on their individual sample. This is caused by the different coefficients for (I/K)t-1 ,(I/K)2t-1 and (CF/K)t-1 observed in
table 4 for both firm types, whereas this regressions predicts one coefficient for both types. However this also enables the regression to analyse the explicit difference for the difference between listed and unlisted firms in investment smoothing using working capital. Table 5 shows the first stage regression that accounts for the difference between listed and unlisted firms denoted by the dummy variable αj,t which is equal to 1 if a firm is unlisted and 0 otherwise.
Only few variables in this table show significant values. None of the investment terms in the previous period has a significant impact in this regression. Then, looking at cash flow, we find a positive relationship for listed firms (0.386) that is significant on the 1% confidence interval. For unlisted firms, there is an effect that is close to zero which was also (0.386 – 0.364 = 0.028) which is close to the value found for the split regression in table 3. Another thing is that
investment in working capital depends negatively on the level of working capital similarly to the previous regressions. For unlisted firms this effect is bigger (-0.080 - 0.038 = -0.118) and
23
The results of the second stage of the regression are shown in table 6. The coefficients for working capital investment are in line with the expectations in section 2 and previous findings. No evidence of investment smoothing can be found for listed firms (0.217) and insignificant. However, for unlisted firms the results are more clear (0.217-0.414=0.217) which is significantly smaller than 0 on the 1% confidence interval. This proves unlisted firms participate in
investment smoothing using working capital in order to reach investment goals. Another thing that can be noticed is the negative coefficient for investment in the previous period. It indicates that firms invest less (more) after a year in which investment was high (low). It is not in line with the investment smoothing and stable investment rates theory described in section two. However, this study finds working capital disinvestment in order to be able to invest in fixed capital. Therefore results should be interpreted as if working capital is used to reach investment goals for that year. But these goals are not based on keeping stable investment rates. If this was the case, the coefficient of previous period investment should have been close to one and certainly non negative. This would have resulted in investment rates equal to the previous period which could then be further affected by the other variable
5.4.5.3.
Robustness check – the influence of cash holdings
As explained in sections 1 and 2, working capital consists of current assets and current liabilities, which can be subdivided into smaller components. To look further into the effect different components of working capital have on the fixed investment of a firm, I split up working capital in two parts. Namely, cash and cash equivalents and short term debt, and inventories, accounts receivables and accounts payables. The argumentation for this subdivision of working capital is that the latter has an effect on the production function. Where, in order to free up cash,
inventories have to decrease, suppliers have to be paid later or customers need to pay earlier. In contrast, the short term debt only shows the proceeds of a contract with a supplier of finance. Also for cash holdings, these do not prevent stock-outs, do not allow to check bought goods after purchasing, and do not boost sales like inventories, accounts payables and accounts
receivables do respectively. In other words, as long as a firm can pay its expenses, cash holdings and short term debt have a much smaller impact on the production function than inventories, receivables and payables.
To check the impact these two part of working capital have on fixed investment smoothing, the estimated working capital to fixed assets (̂∆𝑊𝐶𝐾 ) in equation (7) will be split up into estimated cash and cash equivalents and short term debt (∆𝐶&𝑆𝑇𝐷̂𝐾 ) and estimated inventories, receivables and payables (∆𝐶&𝑆𝑇𝐷̂𝐾 ), leading to equation (9):
24
again, the estimated variables will be estimated using instrumental variables. The expected coefficients, for the previous year’s fixed investment rate, fixed investment rate squared and cash flows, were already described. For the estimated cash and short term debt (∆𝐶&𝑆𝑇𝐷̂𝐾 ) to fixed assets, as well as for the estimated change in inventories, accounts receivables and accounts payables (̂∆𝐼𝑅𝑃𝐾 ), negative coefficients are expected, as the limited pool of capital is expected to be divided over these two variables and the dependent variable. So fixed investment can be smoothed if at least one of the working capital components moves oppositely. To estimate the changes in the two
working capital components, the dependent variable in equation (8) is replaced with the two component indicators. Leading to equation (10):
(∆𝐶&𝑆𝑇𝐷𝐾 )𝑗,𝑡 = 𝛿0+ 𝛿1(𝐾𝐼) 𝑗,𝑡−1+ 𝛿2( 𝐼 𝐾)𝑗,𝑡−1 2 + 𝛿3(𝐶𝐹𝐾)𝑗,𝑡−1+ 𝛿4(𝐶&𝑆𝑇𝐷𝐾 ) 𝑗,𝑡−1+ 𝛿5( 𝐼𝑅𝑃 𝐾 )𝑗,𝑡−1+ 𝜃𝑗+ 𝜃𝑡+ 𝑢𝑗𝑡 (10) and (11): (∆𝐼𝑅𝑃𝐾 )𝑗,𝑡 = 𝛿0+ 𝛿1(𝐾𝐼) 𝑗,𝑡−1+ 𝛿2( 𝐼 𝐾)𝑗,𝑡−1 2 + 𝛿3(𝐶𝐹𝐾)𝑗,𝑡−1+ 𝛿4(𝐶&𝑆𝑇𝐷𝐾 ) 𝑗,𝑡−1+ 𝛿5( 𝐼𝑅𝑃 𝐾 )𝑗,𝑡−1+ 𝜃𝑗+ 𝜃𝑡+ 𝑢𝑗𝑡 (11)
in both these equations, the level of both indicators function as an instrument. This is done in order not to harm the order condition when doing two stage least squares regressions described by Phillips & Hansen (1990). Also, logically thinking, the level of cash and short term debt will have an impact on investment in inventories, receivables and payables because high levels of cash will cause management to use more of the cash balance and less of the inventories, payables and receivables. This also holds the other way around for the change in cash and short term debt depending on the level inventories, receivables and payables.
Equations (10) and (11) are also modified with a dummy variable that is equal to 1 if the firm is listed at the time t and 0 otherwise. This modification is identical to the modification of
equations (5) into (7) and (6) into (8). The formula is shown explicitly for three reasons. First, equations (7) and (8) already show what this modification would lead to. Second , the new equation would not be easy to read due to its length. Third, the presentation of the results will show the components of the equations if still unclear.
25
26
First stage regression testing relationship between investment in working capital category investment in the working capital component
Independent Variable Listed Firms Unlisted Firms Dummy Regressions
Independent
variable: Change in Cash and Short
term liabilities Independent variable: Change in Inventories, Receivables and Payables Independent variable: Change in
Cash and Short term liabilities Independent variable: Change in Inventories, Receivables and Payables Independent variable: Change in
Cash and Short term liabilities Independent variable: Change in Inventories, Receivables and Payables 𝐶 0.332* 0.001 0.288 0.267*** 0.190 0.235*** (1.837) (0.035) (1.603) (14.732) (0.485) (5.945) 𝐶 ∗ 𝑙𝑗,𝑡 0.757 -0.029 (0.287) (-0.108) (𝐼/𝐾)𝑡−1 (-0.557) -0.329 -0.179*** (-2.997) (0.781) 0.308 -0.158*** (-3.974) (0.828) 0.306 -0.158*** (-4.236) (𝐼/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 -0.591 -0.034 (-0.378) (-0.214) (𝐼/𝐾)𝑡−12 -0.130*** 0.001 -0.019 -0.009*** -0.019 -0.009*** (-3.072) (0.229) (-0.768) (-3.474) (-0.826) (-3.705) (𝐼/𝐾)𝑡−12 ∗ 𝑙𝑗,𝑡 (-0.954) -0.106 (0.772) 0.009 (𝐶𝐹/𝐾)𝑡−1 -4.252*** 0.416*** -0.042 0.029*** -0.042 0.029*** (-11.573) (11.239) (-0.665) (4.620) (-0.720) (4.944) (𝐶𝐹/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 -4.163*** 0.384*** (-4.391) (4.024) (𝐶&𝑆𝑇𝐷/𝐾)𝑡−1 (1.947) 0.578* -0.082*** (-2.752) (-0.019) 0.000 -0.001*** (-2.716) (-0.016) 0.000 -0.001*** (-2.902) (𝐶&𝑆𝑇𝐷/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.573 -0.083 (0.748) (-1.069) (𝐼𝑅𝑃/𝐾)𝑡−1 1.379*** -0.067*** 0.724*** -0.119*** 0.724*** -0.119*** (9.356) (-4.490) (16.393) (-26.687) (17.516) (-28.515) (𝐼𝑅𝑃/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.644* 0.051 (1.680) (1.313) 𝑅2 0.175 0.311 0.424 0.294 0.420 0.294
27
measured as the ratio of the change in fixed assets to fixed assets squared, cash flows in the previous year is shown by (CF/K), the level cash minus short term debt is measured (C&STD/K)is measured by working capital minus inventories, accounts receivables and accounts payables, the level of inventories, accounts receivables and accounts payables (IRP/K) is measured by the level of inventories plus accounts receivables minus accounts payables, dummy variable Ij,t is equal to 1 if a firm was listed in a year and zero otherwise. Variables showing
t-1 are lagged one year. Data are winsorized at a t-1% level. The t-values are given in parentheses below the coefficient. R-squared is included as goodness of fit. ***, ** and * stand respectively for statistical significance within the 1%, 5% and 10% confidence levels respectively. Fixed firm and time effects are not reported.
Also, the level of inventories, receivables and payables shows positive coefficients for the change in cash, for both listed and unlisted firms, with coefficients of 1.379 and 0.724 respectively, both at 1% significance. This is probably caused by the fact that firms want to keep enough cash on hand to finance their less liquid current investments. What I also see is that the level of inventories, receivables and payables affects the change of it negatively in both listed and unlisted firms, showing negative coefficients of -0.067 and -0.119 respectively at 1% significance. Which corresponds to the main theory of the working capital level having a negative marginal value to the firm, as explained in section 3.2.
28
TSLS regression testing the effect of smoothing with cash and working capital without cash, for unlisted firms, dependent variable: (𝐼/𝐾 )
Independent Variable Listed Unlisted Dummy Regression
𝐶 0.041*** 0.652*** 0.509*** (2.660) (6.279) (5.982) 𝐶 ∗ 𝑙𝑗,𝑡 0.188 (0.788) (𝐼/𝐾)𝑡−1 -0.460*** -0.506*** -0.491*** (-6.025) (-10.276) (-10.940) (𝐼/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.071 (0.350) (𝐼/𝐾)𝑡−12 -0.004 -0.032*** -0.030*** (-1.060) (-7.204) (-7.620) (𝐼/𝐾)𝑡−12 ∗ 𝑙 𝑗,𝑡 0.024** (2.157) (𝐶𝐹/𝐾)𝑡−1 0.445*** 0.089*** 0.087*** (4.233) (12.097) (12.976) (𝐶𝐹/𝐾)𝑡−1∗ 𝑙𝑗,𝑡 0.205 (0.858) (𝐶&𝑆𝑇𝐷/𝐾)̂ 𝑡−1 -0.093*** -0.456*** -0.435*** (-6.028) (-8.590) (-9.252) (𝐶&𝑆𝑇𝐷/𝐾)̂ 𝑡−1∗ 𝑙𝑗,𝑡 0.351*** (5.734) (𝐼𝑅𝑃/𝐾)̂ 𝑡−1 -1.683*** -3.021*** -2.892*** (-4.303) (-9.290) (-10.036) (𝐼𝑅𝑃/𝐾)̂ 𝑡−1∗ 𝑙𝑗,𝑡 1.692* (1.758) 𝑅2 0.303 0.328 0.329
Panel data cross-section and time fixed effects. The variable data are retrieved from Orbis and Zephyr including 2468 firms consisting of 2169 unlisted firms and 318 listed firms within the United Kingdom for the period 2008-2016. This table provides the least squares regression output for the existence of a linear relation with firm and time fixed effects. The three columns show results for listed, unlisted and a regression with a dummy variable for listed and unlisted firms respectively. The dependent variables fixed investment (I/K) is measured as the ratio of the change in fixed assets to fixed assets. The independent variables fixed investment (I/K) is measured as the ratio of the change in fixed assets to fixed assets, fixed investment squared (I/K)2 is measured as
the ratio of the change in fixed assets to fixed assets squared, cash flows in the previous year is shown by
(CF/K), the change in cash minus short term debt is measured (ΔC&STD/K)is measured by the change in
working capital minus the change in inventories, accounts receivables and accounts payables, the change in inventories, accounts receivables and accounts payables (ΔIRP/K) is measured by the change in inventories plus the change in accounts receivables minus the change in accounts payables, the dummy variable Ij,t is equal to 1 if
29
Comparing listed and unlisted firms, I see that both firm types show a negative coefficient for fixed investment, pointing at a mean reverting process (-0.560 and -0.506 respectively at 1% significance) that are not significantly different from one another 0.071 with a probability of 35%. Furthermore, cash flows in the previous period show influence fixed investment
positively for both listed as well as unlisted firms, with coefficients of 0.445 and 0.089 at 1% significance. Then, looking at the subdivided working capital indicators, I see that both listed and unlisted firms types use their cash balance to invest in fixed assets with coefficients of -0.093 and -0.456 respectively at 1% significance. However, the difference between the two firm types is also significant. In the dummy regression the indicator for unlisted firms is -0.435 at 1% significance. Additionally for listed firms the extra coefficient is equal to 0.351 at 1% significance showing listed firms use much less of their cash balance. This is also in line with the described theory because listed firms were assumed to be financially unconstrained meaning they would be able to attract external financing easily. However, I have to note that listed firms are financially constrained to a certain extend as well, since their coefficient is equal to -0.093, whereas otherwise it would have been 0. Then, looking at inventories, accounts receivables and accounts payables, both listed and unlisted firms show negative coefficients of -1.683 and -3.021 respectively at 1% significance. So both firms free up cash out of this part of working capital. What strikes is that the coefficient of listed firms is less negative than the coefficient of unlisted firms. This can also be seen in the dummy regression where unlisted firms have a coefficient of -2.892 at 1% significance. The difference of listed firms is equal to 1.692 with 1% significance. Showing listed firms disinvest their inventories, accounts payables and accounts receivables much less than unlisted firms. This again
corresponds to the described theory, firms use their working capital to smooth fixed
30
6.
Conclusion
This study used panel data of 323 listed and 2346 unlisted firms in the UK, totalling 2669 firms and 15753 firm-year observations to study the role of investment smoothing on fixed investment rates when firms are financially constrained. The panel ranges from 2009 to 2016. Listed firms were expected to be financially unconstrained whereas unlisted firms were expected to be financially constrained. This study finds that unlisted as well as listed firms show to be smoothing fixed investment using working capital instead of only relying on cash flows as an internal source of financing. The effect of investment smoothing with working capital depends on the level of working capital for the firm. The higher (lower) the level of working capital in a firm, the lower (higher) its marginal value for the firm and therefore a firm is more likely to disinvest (invest) in it in order to smooth fixed capital. It also matters how financially constrained a firm is. Listed firms, which are expected to be less financially constrained, show to smooth fixed investments less using working capital.
Furthermore, the robustness check shows the effect of two components of working capital. These components are cash holdings minus short term debt, and inventories plus accounts receivables minus accounts payables. Cash holdings are the most liquid part of working capital and short term debt is not expected to influence the production function of a company much. This could mean that this part of working capital would be used most as it would be the easiest way to smooth fixed investment. However, results show that inventories, accounts payables and accounts receivables are also much used to smooth investment.
