Corporate Cash Hoarding: The Role of Just-in-Time Adoption Xiaodan Gao∗ National University of Singapore
Abstract I explore the role of the just-in-time (JIT) inventory system in the increase in cash holdings by U.S. manufacturing firms. I develop a model to illustrate the mechanism through which JIT affects cash, and quantify its impact. In the model, both cash and inventory can serve as working capital. As firms switch from the traditional system to JIT, they shift resources from inventory to cash to facilitate transactions with suppliers. On average, this switchover accounts for a 4.1 percentage point increase in the cash-to-assets ratio, which is approximately 28% of the change observed in the data.
JEL Classification: E22; G31; G32; L60 Keywords: Cash holdings; Inventory; Just-in-time; Costly external financing; Trade credit.
∗
Corresponding author. Department of Strategy and Policy, NUS Business School, 15 Kent Ridge Drive, Singapore. Email address:
[email protected]. I would like to thank Viktoria Hnatkovska and Henry Siu for advice, encouragement, and guidance. I am also grateful to Hiroyuki Kasahara for insightful suggestions, Paul Beaudry, Frederico Belo, Murray Carlson, Mick Devereux, Vadim Marmer, Ivan Png, Martin Schmalz, Toni Whited, Jun Yang, Yaniv Yedid-Levi and Xiaoyun Yu for comments, as well as to the audiences at UBC, Bank of Canada, Columbia University, National University of Singapore, University of Minnesota, Indiana University, and the ASSA meeting (AFA, ISIR). I would like to thank Michael Kinney and William Wempe for generously providing data on JIT adoption. I gratefully acknowledge funding from the Singapore Ministry of Education (AcRF Tier 1, Grant R-313-000-105-133). All errors are mine.
1
1
Introduction
The build-up of cash reserves over the past few years by U.S. businesses has captured considerable attention from academic researchers, policy makers, and financial practitioners. It was also one of the most hotly debated issues during the recent recession, because it raises concerns about resource misallocation within firms from high-productivity assets (physical capital) to low-productivity assets (cash).1 This paper aims to understand the causes of the rise in cash holdings in the U.S. manufacturing sector. I start by documenting the stylized facts about the simultaneous changes in cash and inventory holdings in the U.S. and show that the same patterns are found in other major developed economies. I then use firm-level panel analysis to estimate the conditional correlation between these two margins, and find that a 1% drop in the inventory ratio is related to a 0.73% rise in the cash ratio, which is economically sizeable. After demonstrating the close association between changes in cash and inventory, I develop a model to rationalize the phenomenon. The model is motivated by the findings in inventory literature: The substantial reduction in inventory is most commonly attributed to the widespread adoption of the just-in-time (JIT) inventory system.2 I therefore construct a model to explore the role of JIT implementation in shaping cash holdings and quantify its effects. In the model, a firm purchases material inputs for production. It holds cash to facilitate transactions with suppliers and holds inventory to avoid stockouts and absorb cost shocks. More specifically, in each period, the firm makes decisions about inventory adjustment, material input use, cash savings, and dividend distributions, and faces productivity uncertainty and capital market frictions. I model two inventory supply systems: the traditional just-in-case (JIC) system and the new JIT system, which was introduced to the U.S. in the early 1980s. Under the JIC system, there is a lag between material orders and delivery. Although the firm makes input orders before production, new orders do not arrive until current-period production is complete. As a result, the firm adjusts input inventory to anticipate future demand and carries inventory 1
See, for instance, “Companies’ cash piles: Show us the money,” The Economist, July 1, 2010, which states that “if cautious firms pile up more savings, the prospects for recovery are poor.” See also “Corporate savings: Dead money,” The Economist, November 3, 2012. 2 See, for example, Chen, Frank, and Wu (2005).
2
forward to avoid a stockout. In contrast, JIT allows the firm to respond contemporaneously to unexpected demand. With JIT, the firm adjusts inventory holdings with full information about the state of the economy and receives new purchases before current-period production starts. Therefore, under the JIT system, the stockout motive for holding inventory is absent, which explains the decline in inventory as the firm shifts from JIC to JIT. How does JIT influence a firm’s cash policy? Under JIT, the firm is able to adjust and receive inputs before organizing production for each period and thereby avoid inventory carrying costs, which effectively implies lower real prices of material inputs. The firm, therefore, chooses to purchase and use more materials for production each period. To facilitate transactions needed for future production in the presence of costly external finance, the firm carries more cash forward (Baumol, 1952). As a result, cash replaces inventory as the main component of a firm’s working capital. My model suggests that as firms switch from JIC to JIT, they free up internal resources from inventory and shift them to cash reserves to facilitate operations. This implication is strongly supported by data when I test it using a sample of JIT users. I then estimate the model and explore the problem quantitatively. The model is able to deliver a negative cash-inventory correlation of a magnitude similar to that found in the data. Moreover, it predicts that if all firms in the economy switch from JIC to JIT and achieve supply-chain perfection, the average cash-to-assets ratio will rise by 5.9%. Taking into account the fact that around two-thirds of U.S. manufacturers had adopted JIT by 2008 and controlling for self-selection bias, I find that the average cash ratio increases by 4.1 percentage points. That is, approximately 28% of the observed cash increase in manufacturing is attributable to JIT adoption. Results are similar when I consider two extended versions of the benchmark model and a different sample. This paper fits into three broad strands of literature. First, it helps to understand the reasons behind the significant rise in corporate cash holdings over the past 30 years.3 It explores the role of JIT and finds that it can explain roughly 28% of the trend in average cash movements in the U.S. manufacturing sector. While Bates, Kahle, and Stulz (2009) also highlight the change 3
A number of papers examine the cash-hoarding behavior of U.S. firms. An incomplete list includes Armenter and Hnatkovska (2016), Azar, Kagy, and Schmalz (2016), Bates, Kahle, and Stulz (2009), Boileau and Moyen (2016), Falato, Kadyrzhanova, and Sim (2013), Karabarbounis and Neiman (2012), Lyandres and Palazzo (2016), Ma, Mello, and Wu (2013), Morellec, Nikolov, and Zucchi (2013), and Zhao (2016).
3
in inventory holdings as an important factor in understanding cash hoarding, they neither estimate its effects nor explore the mechanism that underlies the opposite trends in cash and inventory holdings. I propose a channel through which cash and inventory behave in a way that is consistent with the empirical facts and evaluate its contribution. Second, this paper complements the corporate cash literature. A number of structural cash models focus on the nonoperational use of cash (Riddick and Whited, 2009; Bolton, Chen, and Wang, 2011; Hugonnier, Malamud, and Morellec, 2015). In those studies, cash is modeled as a precautionary hedge against future uncertainty in the presence of financial frictions. That is, in the face of costly external finance, holding cash today can reduce the probability of tapping external borrowing tomorrow when firms experience cash flow shortfalls and need funds for capital investment. In my model, cash serves two purposes: nonoperational use (precautionary savings) and operational use (working capital). The former is the same as the motive highlighted in previous studies, while the latter is new in the literature. Even in the absence of uncertainty, firms have incentives to hold cash as a result of financial frictions. That is, firms choose to accumulate cash because more external borrowing costs can be saved when they purchase production inputs and facilitate operations before current-period cash flows are realized. This is consistent with the survey evidence that a large portion of corporate cash savings is held for operational purposes (Lins, Servaes, and Tufano, 2010). Accordingly, it is of great importance to model operational cash. Lastly, this paper adds to the JIT literature by relating JIT to firms’ financial policies. To the best of my knowledge, no previous work links JIT with cash management, despite abundant evidence that JIT is an efficient way to reduce inventory. How do firms allocate resources released from inventory after switching to JIT? My model suggests that firms choose to augment their cash stocks to maintain smooth operations. The remainder of the paper is organized as follows. Section 2 provides empirical facts about the close link between cash and inventory policies. Section 3 proposes a rationale for the empirical observations and presents a dynamic stochastic model in which a firm manages its cash and inventory. Section 4 discusses the estimation of model parameters and presents simulation results to evaluate JIT’s role in corporate cash-hoarding behavior. Section 5 concludes.
4
2
Empirical Facts
In this section, I use firm-level data to document the empirical facts regarding the relationship between firms’ cash and inventory holdings. I first show the unconditional correlation by presenting time-series dynamics. I then control for the effects of other firm-level characteristics and estimate the conditional correlation between cash and inventory holdings.
2.1
Time Series Dynamics
Figure 1 plots the time-series dynamics of the average cash-to-assets ratio (line connected with circles), the average inventory-to-assets ratio (line connected with triangles), and their sum (line connected with diamonds) for manufacturing firms that are publicly traded in the U.S., Japan, Germany, France, Italy, and Canada. The sample of U.S. companies is constructed from Compustat Annual Files for the period 1970-2009. The sample of Japanese companies is constructed from PACAP for the period 1975-1990. The samples for European countries are constructed from Compustat Global for the period 1988-2009, and the sample for Canada is from Compustat Global for the period 1980-2009. Both cash and inventory ratios are winsorized between zero and one. As shown in the top left panel, among publicly traded U.S. manufacturing firms, the average cash-to-assets ratio increased from 8.3% in the 1970s to 26.7% in 2009, and the inventory-toassets ratio decreased from 27.2% to 12.7%. Despite the striking changes in both, the sum of these two ratios was relatively stable over the entire 30-year period. This indicates an almost perfect negative correlation between cash and inventory. I then investigate whether firms in Japan manage their inventory and cash holdings in the same way as their counterparts in the U.S. The top middle panel suggests that similar patterns are present. More precisely, since the mid 1970s, the inventory ratio has been reduced from 20% to 12%. From 1988 on, the inventory ratio stopped decreasing and became stable. The average cash ratio during the same period moved in the opposite direction, except for the last three years. The decrease in cash ratio starting from the end of the 1980s is attributed to banks’ weakened power (Pinkowitz and Williamson, 2001). Also examined are Germany, France, Italy, and Canada. As shown in the remaining four 5
U.S.
Germany .4 .3 ratios
.1
.1
.1
.2
.2
.2
ratios
ratios
.3
.3
.4
.4
Japan
1970
1980
1990 year
average cash ratio average cash and inventroy ratio
2000
2010
1975
1980
1985
1990
1990
1995
year average inventory ratio
average cash ratio average cash and inventory ratio
2005
2010
average inventory ratio
Canada
.3 .1 0
.1
.1
.15
.15
ratios .2
ratios .2
ratios .2 .25
.25
.3
.3
.4
Italy
.35
France
2000 year
average cash ratio average cash and inventory ratio
average inventory ratio
1990
1995
2000 year
average cash ratio average cash and inventory ratio
2005
2010
average inventory ratio
1990
1995
2000 year
average cash ratio average cash and inventory ratio
2005
2010
1980
1990
2000
2010
year average inventory ratio
average cash ratio average cash and inventory ratio
average inventory ratio
Figure 1: Average Cash and Inventory Ratios in the U.S., Japan, Germany, France, Italy, and Canada. This figure summarizes the average cash-to-assets ratio, average inventory-to-assets ratio, and the sum of those two ratios over time in the U.S., Japan, Germany, France, Italy, and Canada. The sample of U.S. companies is constructed from Compustat Annual files, the sample of Japanese companies is constructed from PACAP, and the samples for European countries and Canada are constructed from Compustat Global. panels, the cash-to-assets and inventory-to-assets ratios all moved in opposite directions over time, with the sum of those two ratios remaining stationary. As a placebo test and to facilitate comparison, in Figure 2 I plot the time-series dynamics of average cash and inventory ratios in the manufacturing sector, mining sector, and petroleum refining industry. The mining sector is little concerned with inventory management. Over the past three decades, its inventory ratio has been quite stable, while its average cash ratio has fluctuated without a significant upward trend. However, mining is also a low R&D intensity sector, and therefore its unchanged average cash ratio may be caused by weak demand for internal cash to fund R&D investment. To alleviate this concern, I also examine the petroleum refining industry, which is a medium-high R&D intensive industry, yet is also little concerned with inventory management.4 As shown in the last panel of Figure 2, the average cash ratio did not move upwards. 4
All high-tech manufacturing industries—Chemical & Allied Products (SIC 28), Industrial Machinery & Equipment (SIC 35), Electronic & Other Electric Equipment (SIC 36), transportation-equipment industries (SIC 37), and Instruments & Related Products (SIC 38)—implement JIT, as shown in Panel B of Table A1.
6
Manufacturing
Petroleum Refining
ratios .15 .05
0
.1
.05
.1
.2
ratios
ratios .1
.3
.15
.2
.2
.4
.25
Mining
1980
1990
2000
2010
year average cash ratio average cash and inventory ratio
1980
1990
2000
2010
1980
1990
average cash ratio average cash and inventory ratio
2000
2010
year
year average inventory ratio
average inventory ratio
average cash ratio average cash and inventory ratio
average inventory ratio
Figure 2: Average Cash and Inventory Ratios in Manufacturing, Mining, and Petroleum Refining. This figure summarizes the average cash-to-assets ratio, average inventory-to-assets ratio, and the sum of those two ratios over time in the manufacturing sector, mining sector, and petroleum-refining industry. The sample is constructed from Compustat Annual files from 1980 to 2009.
2.2
Correlation between Cash and Inventory
To rule out the possibility that the patterns shown in Figure 1 could be a coincidence, I then estimate the correlation between cash and inventory by controlling for other factors that are usually taken into account to explain cash-holding behavior. To that end, I adopt the baseline cash regression used by Bates, Kahle, and Stulz (2009) and make three changes to it. First, I separate inventory from net working capital to explicitly gauge the importance of the former. Second, I replace industry-level risk with firm-specific risk and control for industry fixed effects, so that I can use within-industry variation to identify the effect of risk on cash. Lastly, I include cohort dummies—which are constructed based on firms’ IPO listing dates—and year dummies. Cohort fixed effects are motivated by the fact that more recently listed companies hold more cash than older cohorts (Bates, Kahle, and Stulz, 2009), and year fixed effects are used to capture the common macroeconomic shocks across firms. Cash regression is therefore specified as follows: cashi,t = α0 + α1 inventoryi,t + α2 f irm sizei,t + α3 riski,t + α40 Xi,t
+industryj + yeart + cohortl + i,t .
(1)
In this regression equation, cash is the ratio of cash and short-term investments to total assets; firm size is defined as the natural logarithm of total assets; risk is computed as the
7
standard deviation of the ratio of annual operating cash flow to total assets for the previous five periods; and inventory is measured as the ratio of inventory to total assets. Other explanatory variables, X, include market-to-book ratio, operating cash flow, working capital net of cash and inventory, capital investment, and so forth. A detailed description of these covariates is provided in Appendix A.1. The sample is constructed from the Compustat Fundamentals Annual files, which comprise an unbalanced panel of manufacturing firms (SIC 2000-3999) for the period 1980-2006.5 To control for outliers in the sample, I delete firms with negative total assets or negative sales, and winsorize continuous variables. Leverage, cash, and inventory ratios are winsorized between zero and one. R&D, acquisition, and capital investment ratios are winsorized at the top and bottom 1%. Cash-flow ratio and net working capital are winsorized at the bottom 1%, and market-to-book ratio is winsorized at the top 1%. Table 1 reports the descriptive statistics for these variables, which have similar characteristics to those used in previous studies.6 Table 1: Summary Statistics Table 1 presents the descriptive statistics for the variables used in the estimation. The sample is constructed from Compustat Fundamentals Annual files for the period 1980-2006. A detailed definition of variables is provided in Appendix A.1. Variables Mean Median Std. Dev. 25% 75% Obs. Cash 0.19 0.08 0.24 0.02 0.27 78055 Inventory 0.19 0.17 0.14 0.08 0.27 78006 Size 4.23 4.10 2.54 2.50 5.87 78055 Risk 0.11 0.04 0.22 0.02 0.10 53646 Market-to-Book 2.28 1.20 3.60 0.78 2.18 67494 Cash flow -0.12 0.06 0.62 -0.06 0.12 78055 Net working capital -0.14 -0.04 0.50 -0.14 0.04 77288 Capital investment 0.06 0.04 0.06 0.02 0.07 77154 Leverage 0.26 0.21 0.25 0.05 0.38 77921 R&D 0.13 0.05 0.21 0.02 0.13 54216 Dividend dummy 0.31 0 0.46 0 1 78180 Acquisition 0.02 0 0.05 0 0 74801 Concentration ratio-50 0.72 0.76 0.18 0.62 0.863 11211
Table 2 summarizes the estimation results of regression model (1) and its alternative specifications. Column (1) reports the pooled OLS regression results controlling for 3-digit SIC industry fixed effects, year fixed effects, and cohort fixed effects, whereas Columns (2) and (3) re-estimate regression equation (1) using 4-digit SIC industry dummy variables and firm fixed 5 6
I use a pre-crisis sample to ensure that estimation results are not driven by the Great Recession. See, for instance, Morellec, Nikolov, and Zucchi (2013).
8
effects, respectively. Table 2: Regression Results on Corporate Cash Holdings Table 2 reports the estimation results of the cash regression (1) on firms’ characteristics. Industry, cohort, and year fixed effects are included in the regressions. The heteroskedasticity-consistent standard errors reported in parentheses account for possible correlation within a firm cluster. Significance levels are indicated by ∗ ,∗∗ , and ∗∗∗ for 10%, 5%, and 1%, respectively. (1) (2) (3) (4) (5) Cash/Assets Cash/Assets Cash/Assets Cash/Assets log(Cash/Net Assets) Inventory -0.6928∗∗∗ -0.6894∗∗∗ -0.7299∗∗∗ -0.7453∗∗∗ (0.0210) (0.0201) (0.0134) (0.0299) Inventory/Net Assets -0.1128∗ (0.0646) Size -0.0101∗∗∗ -0.0107∗∗∗ -0.0059∗∗∗ -0.0132∗∗∗ 0.0208∗ (0.0008) (0.0012) (0.0018) (0.0018) (0.0113) Market-to-book 0.0067∗∗∗ 0.0067∗∗∗ 0.0047∗∗∗ 0.0083∗∗∗ 0.0729∗∗∗ (0.0008) (0.0008) (0.0006) (0.0017) (0.0058) Risk 0.0627∗∗ 0.0604∗∗∗ 0.0589∗∗∗ 0.0247 0.6105∗∗∗ (0.0114) (0.0111) (0.0089) (0.0200) (0.0841) Cash flow 0.0361∗∗∗ 0.0333∗∗∗ 0.0174∗∗∗ 0.0345∗∗∗ 0.2275∗∗∗ (0.0058) (0.0056) (0.0047) (0.0110) (0.0451) Net working capital -0.0096 -0.0077 -0.0340∗∗∗ -0.0053 -0.1874∗∗∗ (0.0067) (0.0067) (0.0052) (0.0122) (0.0565) Capital investment -0.7484∗∗∗ -0.7330∗∗∗ -0.4033∗∗∗ -0.8211∗∗∗ -4.3444∗∗∗ (0.0305) (0.0307) (0.0183) (0.0609) (0.2850) Leverage -0.2497∗∗∗ -0.2409∗∗∗ -0.1908∗∗∗ -0.2458∗∗∗ -2.9771∗∗∗ (0.0105) (0.0105) (0.0068) (0.0170) (0.1023) R&D 0.1397∗∗∗ 0.1160∗∗∗ -0.1160∗∗∗ 0.1054∗∗∗ 1.3098∗∗∗ (0.0185) (0.0182) (0.0139) (0.0293) (0.1295) Dividend -0.0291∗∗∗ -0.0253∗∗∗ 0.0077∗∗∗ -0.0320∗∗∗ -0.1894∗∗∗ (0.0047) (0.0045) (0.0025) (0.0065) (0.0459) Acquisition -0.3894∗∗∗ -0.3767∗∗∗ -0.2571∗∗∗ -0.3435∗∗∗ -2.7524∗∗∗ (0.0190) (0.0189) (0.0124) (0.0423) (0.1916) Concentration ratio-50 -0.0521∗ (0.0294) Industry FE (3-digit) Yes Yes Industry FE (4-digit) Yes Yes Firm FE Yes Year FE Yes Yes Yes Yes Yes Cohort dummy Yes Yes Yes Yes Observations 32,939 32,939 32,939 4,828 32,704 R-squared 0.572 0.585 0.279 0.578 0.416
The variable of particular interest here is the inventory ratio. According to Column (1), a 1% decrease in inventory is correlated with a 0.69% increase in a firm’s cash holdings, which is statistically and economically significant. The coefficients on other independent variables are consistent with those estimated by Bates, Kahle, and Stulz (2009). Larger firms, either because of economies of scale for transaction purposes or easier access to external capital, hold less cash. Firms facing higher risks tend to save more cash for precautionary motives. Firms expecting more future investment opportunities, proxied by market-to-book ratio and R&D spending,
9
accumulate more cash. Also, cash is consumed by paying off debt, investing in capital, and distributing dividends. Results are robust with respect to different specifications and regression methodologies, as shown in Columns (2) and (3). In particular, the coefficient on the inventory ratio varies within a fairly narrow interval [-0.69,-0.73], which is close to the values reported by Kulchania and Thomas (2017).7 Another motive for holding cash highlighted in the literature is the strategic motive (Lyandres and Palazzo, 2016). In Column (4), I control for market competition by including the 50-firm concentration ratio of each 4-digit industry which is directly collected from the Census of Manufacturers reports provided by the U.S. Census Bureau. Consistent with previous studies, firms that operate in more concentrated industries tend to hold less cash. Moreover, controlling for strategic interactions among firms strengthens the correlation between inventory and cash holdings. Cash and inventory are components of total assets. The negative correlation between cash and inventory ratios is possibly generated by variable construction. In an attempt to alleviate this concern, I also use the ratio of cash to total assets net of inventory and cash as the dependent variable, take the logarithm to deal with outliers, and run the same regression model. Results are reported in the last column of Table 2. The relationship between cash and inventory remains negative and statistically significant.
2.3
JIT Philosophy
The often-mentioned explanation for the observed reduction in inventory is the widespread adoption of the JIT inventory system, which therefore has the potential to explain the time patterns shown in Section 2.1 and the empirical findings in Section 2.2. Below, I give a brief introduction to the JIT philosophy. JIT is a philosophy of efficiency improvement that emphasizes the performance of activities based on immediate needs. Narrowly defined, it strives to eliminate excess inventory that results from overproduction and waiting. JIT can be applied at both the purchasing stage and the production stage. JIT purchasing involves the speedy delivery of materials from suppliers once 7
In addition to changes in inventory, Bates, Kahle, and Stulz (2009) also link the rise in cash with increasing R&D spending. I estimate the conditional correlation between cash and inventory for both high and low R&D firms. Interestingly, the negative correlation is stronger for high-R&D groups than for low ones, -0.83 vs. -0.6.
10
they have been ordered; the materials required for purchases arise from the manufacturing process. JIT manufacturing involves the production of goods to meet current needs, rather than to anticipate future demand. JIT purchasing is essential for firms that implement JIT manufacturing, because a delay in material delivery will slow the entire production process. Speedy deliveries are made possible by improved freight transportation systems and reduced transportation costs. In addition, firms often build long-term relationships with reliable suppliers, preferably with warehouses close to firms, and order several parts from the same supplier, even if the parts might be cheaper through other vendors. Firms also work closely with suppliers and exchange proprietary information to improve the quality of parts. This reduces inspection time and effectively shortens the time delay between material order and material use in the production process. JIT strategy was first adopted by Toyota’s manufacturing plants, and by the mid-1970s had attracted many followers in Japan. With Japanese manufacturing firms achieving high levels of international competitiveness in the early 1980s, JIT began attracting considerable attention in the U.S. as well, and has gradually been adopted on a broad scale since then. Using survey data, Morgan (1991) finds that JIT adopters increased from 5% in early 1980s to more than 15% in 1990. Compdata’s Compensation Data Manufacturing & Distribution surveys, which cover nearly 1,000 manufacturing employers across the U.S., report that in 2010, 69.7% of manufacturing companies employed lean manufacturing practices that are often viewed as equivalent to JIT. This figure continued to increase, and reached 71.3% in 2015. An independent survey conducted by CFO Research Services in collaboration with FM Global in 2008 shows a similar number—that is, nearly two-thirds of manufacturing firms had implemented JIT by 2008. These surveys suggest a steady increase in the JIT adoption rate over time and widespread JIT implementation across the U.S. Prior to the introduction of JIT, U.S. firms embraced the just-in-case (JIC) philosophy, and held buffer stocks at every stage in the production process to meet unexpected demand fluctuations or production problems. I next construct a model to understand the role of JIT in explaining firms’ cash-hoarding behavior. In the model, JIT is narrowly defined as JIT purchasing and aims to eliminate input inventory. 11
3
Model
This section presents a partial equilibrium problem of a firm that faces uncertainty and financing frictions. I introduce inventory into an otherwise standard cash model by assuming raw materials as factors in production.8 I model two inventory supply systems: the traditional JIC system and the more recent JIT system. The lag in delivery of material purchases is the key difference between JIC and JIT. In particular, the delivery lag under JIC leads firms to purchase materials in anticipation of future demand and maintain inventory as buffers against uncertainty. Under the JIT system, since there is no lag in delivery, firms are able to respond contemporaneously to shocks, and place orders based on current-period demand. I begin by specifying a firm’s production technology and financing options, then describe the problems the firm faces when it operates as a JIC adopter and as a JIT adopter.
3.1
Technology
Consider a discrete time model of an infinitely lived firm. The firm uses capital K = 1 and materials N to produce output, and faces a combination of demand and productivity shocks, z1 . The revenue function, F (z, N ), is specified by F (z1 , N ) = z1 N θ ,
(2)
where curvature θ < 1 captures decreasing returns to scale in production or market power, or a combination of both. The revenue is subject to a shock z1 , following an AR(1) process in logs with persistence ρz1 and innovation εz1 , ln z10 = ρz1 ln z1 + ε0z1 .
(3)
A prime indicates a variable in the next period, and no prime indicates a variable in the current period. The innovation εz1 has a normal distribution, with mean 0 and variance σz21 , εz1 ∼ N (0, σz21 ). 8
See, for instance, Riddick and Whited (2009).
12
3.2
Inventory
In every period, the firm decides how many units of materials is to purchase based on the current state. The input price is subject to a cost shock z2 , which is independent of the revenue shock and follows the AR(1) process specified below: ln z20 = ρz2 ln z2 + ε0z2 ,
(4)
with εz2 ∼ N (0, σz22 ). After uncertainty is realized, the firm makes the inventory adjustment decision is and chooses how much material, N , to use for production. Under JIC, newly purchased materials arrive after production. The decision on N , therefore, is constrained by the material stock s at the beginning of the period, N < s. In contrast, under JIT, new material orders get delivered immediately and arrive before production starts. The material available for current-period production thus becomes s + is . Materials fully depreciate in use, and unused materials are held as inventory and depreciate at a rate δs . The end-of-period inventory holdings are therefore given by s0 = (1 − δs )(s + is − N ).
3.3
(5)
Financing
The firm must pay fixed operating costs and purchase input materials in advance of production. It has three sources of funds to finance these expenses: current operating cash inflow from sales, internal cash balance, and external funds. The internal cash balance stored by the firm, c, earns zero interest. The cost of holding cash, therefore, is the risk-free interest rate. The firm can use trade credit, T , extended by suppliers to finance its input purchases. If the firm chooses not to use it, it gets an early-payment discount λ1 , which is effectively the borrowing rate. As such, internal cash is relatively cheaper than trade credit. When the firm has internal liquidity, it has an incentive to take advantage of early-payment discounts rather
13
than using extended trade credit.9 Besides trade credit, the firm can resort to another source of external funds: The firm pays a linear cost λ2 for every dollar raised externally (Nikolov and Whited, 2014). This form of external borrowing is a combination of short-term debt and equity financing. I model debt and equity financing together for two reasons. First, this paper focuses on internal cash holdings in the presence of costly external finance, rather than the trade-off between debt and equity financing. Second, this assumption reduces the control space by one dimension, retains the model’s main insights, and preserves computational tractability. However, this form of external borrowing differs from traditional debt and equity financing: (i) it is unlimited, (ii) it is more expensive than debt, and (iii) it incurs zero fixed costs. I will separate debt and equity financing when I perform robustness checks by considering an extension of the model.
3.4
The Firm’s Problem
The timeline for the model—in both JIC and JIT environments—is illustrated below. At the beginning of period t, after observing the shock realizations z1 and z2 , the firm pays fixed operating costs cf and decides on material orders is in advance of production. The firm then produces goods using materials on hand, and decides how to spend the current-period cash flow from sales. The firm is required to pay for input materials when it makes orders before production, using internal cash, trade credit, and/or other external funds.10 The only difference between JIC and JIT lies in the timing of the delivery of material orders and, therefore, materials available for current-period production use. I model JIT as immediate delivery of material orders. Under JIT, therefore, there is no shipping lag. 9 For example, OEA, Inc., a supplier of automotive firms with JIT inventory environments, said in its Form 10-K for 1997 that “customer payments are due on a current basis” and “customer payments are reasonably prompt and extended terms or collateral are not required.” 10 This assumption is reasonable under JIT, because materials get delivered immediately before production. This is also reasonable under JIC when the operating cycle is longer than material delivery, so that current-period sales cannot be used to fund material purchases.
14
JIC {z1 , z2 , c, s}
choose N ≤ s
choose is
is arrives choose c0
s
s
t| {z } | {z } | {z }t+1 pay cf and z2 is using cash, produce goods with N allocate resources; unused and/or trade credit, and/or materials are held as s0 other external borrowings
JIT {z1 , z2 , c, s}
s
choose is
choose N ≤ s + is
is arrives
choose c0
s
t| {z } | {z } | {z }t+1 pay cf and z2 is using cash, produce goods with N allocate resources; unused and/or trade credit, and/or materials are held as s0 other external borrowings
3.4.1
The Firm’s Problem under JIC
More specifically, after shocks are realized at the beginning of each period, the firm pays fixed operating costs cf and purchases materials z2 is prior to production. It can use its cash reserve c, and/or borrow externally if cash is insufficient to cover these expenses. The firm can also fund ordered materials with trade credit T , by giving up early-payment discounts. The net cash flow before production is given by:
e1 = c − cf − (1 − λ1 ) max[z2 is − T, 0].
(6)
A negative e1 indicates external borrowing, while a positive e1 denotes unused funds at this stage that will be carried forward and used for post-production operation. Under JIC, material delivery takes time. After paying the fixed operating costs and placing orders for input materials, the firm starts production. It decides on how many materials to use in producing goods, but is constrained by the quantity available in stock, N ≤ s. The remaining materials, together with new orders,
15
are stored as inventory s0 . They depreciate and are transferred to the next period. The firm also decides how much to save out of current-period cash flow, with a nonnegative cash balance constraint c0 ≥ 0. If current-period after-tax cash inflow is insufficient to meet cashsaving demands and repay trade credit, the firm borrows externally and pays borrowing costs; otherwise, the firm distributes dividends. The post-production net cash flow is then given by e2 = F (z1 , N ) − τc [F (z1 , N ) − N ] + (1 − φe1 )e1 − c0 − T.
(7)
In the expression of e2 , the first two terms represent the after-tax profit, and the third term is the funds left after paying expenses prior to production. The last two terms are cash holdings carried to the next period and trade credit that the firm is required to repay at the end of the current period. The indicator function φe1 equals zero if e1 is nonnegative, and one otherwise. The firm is risk-neutral. It aims to maximize its value, which is discounted at the risk-free rate r, by choosing material orders is , trade credit T , material use N , and cash holdings c0 . Its problem can be written as:
V (z1 , z2 , c, s) =
max
{g(e1 ) + g(e2 ) + βEV (z10 , z20 , c0 , s0 )},
is ≥0,T ≥0,N ≤s,c0 ≥0
(8)
g(e1 ) = φe1 (1 + φe1 λ2 )e1 , g(e2 ) = (1 + φe2 λ2 )e2 , s0 = (1 − δs )(s + is − N ), where β =
3.4.2
1 1+r .
The Firm’s Problem under JIT
As above, the firm’s problem under JIT can also be viewed in two stages. The first stage is exactly the same as the JIC problem. The difference is that under the new system, materials get delivered immediately. With the newly purchased materials, the firm enters the second stage. Materials available for production under the new system are the sum of the initial stock and the new purchase. That is, material use is subject to a new resource constraint, N ≤ s + is . Upon
16
receiving after-tax revenue, the firm chooses the best way to allocate it. The corresponding second-stage net cash flow is e2 = F (z1 , N ) − τc [F (z1 , N ) − N ] + (1 − φe1 )e1 − c0 − T,
(9)
where N ≤ s + is . The firm’s problem can then be summarized by the following Bellman equation:
V (z1 , z2 , c, s) =
max
{g(e1 ) + g(e2 ) + βEV (z10 , z20 , c0 , s0 )},
is ≥0,T ≥0,N ≤s+is ,c0 ≥0
(10)
with the law of motion for inventory holdings, s0 = (1 − δs )(s + is − N ).
3.5
Optimal Policy Rules
In this subsection, I characterize the firm’s optimal cash and inventory policies under JIC and JIT and give the intuition behind them.
3.5.1
Optimal Cash Holdings under JIC and JIT
Solving the firm’s optimization problem under JIC gives the optimal cash level, which satisfies 0
0
0
1 + φe2 λ2 = β + βλ2 E[φe21 + (1 − φe1 )φe2 ] + µ1 .
(11)
The term µ1 is the Lagrange multiplier of the nonnegativity constraint on cash and gives the shadow price of cash holdings. The left-hand side of equation (11) represents the cost of saving an additional unit of cash, that is, forgone dividends or the sum of external borrowing and the corresponding costs in the case of borrowing externally to save cash during this period. The right-hand side of the equation is the marginal benefit of cash savings. It has two components: (i) the discounted expected value, captured by the first term; and (ii) the discounted expected reduction in external borrowing costs, represented by the second term. The firm anticipates two scenarios in which it might be short of internal liquidity and need to borrow externally. One scenario is when the firm needs funds to pay operating costs and finance material purchases before production. Its 17
0
corresponding probability of borrowing externally is Eφe1 . The other is when the firm needs 0
funds to repay trade credit and finance cash investment, which occurs with probability Eφe2 . The unit of cash at the margin can be used in either the first scenario or the second, captured by the indicator function φ0e1 . With everything else remaining the same, the optimal cash policy derived under JIT differs from equation (11) through the likelihood of borrowing externally before production, that is, 0
Eφe1 . Implementing JIT lowers the real price of materials by saving inventory-depreciation costs, which in turn prompts firms to purchase and use more materials in productions. Higher demand for material inputs translates into higher demand for liquidity. To avoid tapping into expensive external funds, the firm carries more cash forward. In other words, as the firm expects a higher likelihood of the insufficiency of internal funds to finance operating costs and purchase material inputs under JIT, it retains more cash because additional borrowing costs are likely to be saved.
3.5.2
Optimal Inventory Policy under JIC and JIT
I next compare the optimal conditions for inventory holdings under the two inventory systems. Under JIC, the optimal condition is given by
n o 1 − λ1 0 0 0 0 z2 = βE {1 + λ2 [φe21 + (1 − φe1 )φe2 ]}(1 − λ1 )z2 + βEµ03 + µ2 . 1 + λ2 [φ2e1 + (1 − φe1 )φe2 ] 1 − δs (12)
The marginal cost on the left-hand side is the input price plus borrowing costs in the case of borrowing externally to pay bills, plus inventory-depreciation costs. The marginal benefit on the right-hand side is the discounted expected value of an additional unit of end-of-period inventory for the subsequent period, which has three components. The first is the discounted expected value generated by saving materials-purchase costs. The second term is the discounted expected value of relaxing the materials-use constraint, N 0 ≤ s0 , by carrying one additional unit of inventory into the subsequent period. The last term is the Lagrange multiplier of the nonnegativity constraint on inventory holdings.
18
Under JIT, the optimal condition becomes
n o 1 − λ1 0 0 0 0 z2 = βE {1 + λ2 [φe21 + (1 − φe1 )φe2 ]}(1 − λ1 )z2 + µ2 . (13) 1 + λ2 [φ2e1 + (1 − φe1 )φe2 ] 1 − δs
Compared to equation (12), the marginal cost stays the same. The marginal benefit, however, decreases by the term βEµ03 . This term captures the value of holding an additional unit of inventory to relax the future materials-use constraint, which is absent under JIT. This is because JIT allows the firm to flexibly adjust its inventory stock before organizing production in each period. The drop in the marginal benefit of inventory holdings implies that the firm has a weaker motive to store inventory under JIT.
3.5.3
Discussion
The results obtained above allow us to understand the different incentives for carrying cash and inventory under different inventory systems. Under JIC, inventory functions as working capital and is not adjustable before the next period’s production. Anticipating the level of material use, the firm holds inventory forward to smooth future manufacturing. In contrast, under JIT, the firm is able to adjust inputs before organizing production for each period and save inventory carrying costs, which effectively implies lower costs of material inputs. The firm, therefore, no longer has a motive to store inventory, aside from taking advantage of favorable cost shocks. Instead, the firm chooses to purchase and employ more materials in each period, and transfer more cash forward to facilitate transactions with suppliers in the presence of financing frictions. Unlike the JIC system, cash in this new environment takes the place of inventory, acting as working capital to ensure smooth operations.
4
Empirical Results
The model presented in the previous section proposes an explanation for the time-series patterns plotted in Figure 1 and provides intuition on how implementing JIT drives down inventories and drives up cash holdings.
19
In this section, I explore the model’s quantitative predictions. I first provide suggestive evidence to support the model’s implications for the role of JIT in explaining the changes in cash and inventory observed in the data. I then parameterize the model and quantitatively assess the impact of JIT on firms’ cash and inventory management. Lastly, I examine the robustness of the main results by extending the baseline model in two ways.
4.1
Suggestive Evidence
The model implies that as firms switch from JIC to JIT, they reduce inventory to cut production costs and increase cash to preserve operating flexibility. I provide suggestive evidence below to support this implication by using a sample of JIT adopters and non-adopters and employing a difference-in-differences (DID) approach. The identification comes from the differences between pre-adoption and post-adoption, within-firm differences of JIT adopters and non-adopters. The initial JIT adopter sample of 201 firms was kindly provided by Kinney and Wempe (2002). I search for key words “JIT,” “Just in Time,” and “lean” in LexisNexis Academic SEC filings and find additional 79 adopters. I screen these 280 adopters with the following three requirements: (i) they operate in the manufacturing sector; (ii) quarterly data are comparable across time—that is, the variable compstq has an empty string; and (iii) there are no missing quarterly data with which to calculate the cash ratio for each of the five consecutive years from the two years prior to JIT adoption to the two years after. The final sample has 149 JIT adopters. Following Kinney and Wempe (2002), I construct a control group based on industry and total assets in the year preceding JIT adoption. The characteristics of the JIT adopter sample and matched control group in the year before adoption are provided in Table 3. On average, JIT adopters and control firms have similar observable characteristics, except that adopters are slightly larger, have higher market-to-book ratios, and carry less debt. I plot the dynamics of average cash ratios of the JIT sample and control group from a two-year pre-adoption period to a two-year post-adoption period in Figure 3, where Year 0 is the year that adopters start implementing JIT. The pre-trends of the JIT sample and control group are parallel, which lends support for performing DID analyses in the subsections below.
20
Table 3: Summary Statistics Table 3 presents the descriptive statistics for the adopter sample and control group in the pre-adoption year. The sample is constructed from Compustat Fundamentals Quarterly files. A detailed definition of variables is provided in Appendix A.1. A: JIT Sample B: Control Sample Variables Mean Median Mean Median Cash 0.09 0.05 0.09 0.03 Inventory 0.24 0.22 0.24 0.23 Size 5.65 5.50 5.48 5.40 Risk 0.01 0.01 0.01 0.01 Market-to-Book 1.19 1.02 1.08 0.95 Cash flow 0.04 0.04 0.04 0.04 Net working capital -0.01 0.01 -0.02 -0.00 Capital investment 0.01 0.01 0.01 0.01 Leverage 0.23 0.20 0.27 0.25 R&D 0.02 0.02 0.02 0.02 Dividend dummy 0.49 0.00 0.47 0.00 Acquisition 0.005 0.00 0.004 0.00
4.1.1
Sample Validation
The role of JIT in reducing inventory holdings has been broadly documented in the inventory literature. I thus validate the sample by examining whether adopters manage their inventory in a way that is consistent with JIT philosophy. To this end, I consider the following regression specification: inventoryi,t = β0 + β1 JITi,t + β20 Xi,t + γi + σI,T + i,t .
(14)
Here, inventory is the ratio of inventory to total assets. Dummy variable JITi,t takes the value one if firm i at time t implements JIT and zero otherwise. The control variables, Xi,t , are similar to those included in the cash regression (1), and γi and σI,T are firm fixed effects and industry-specific year fixed effects, respectively. I identify β1 , the average effect of JIT on inventory holdings, by assuming that (i) all the unobserved heterogeneity that leads to the correlation between JIT adoption and the error term is captured by firm fixed effects, and (ii) variations in the dependent variable due to changes in the macroeconomic environment are captured by industry-specific year fixed effects, which are common to firms in both the treatment and control groups within the same industry. I estimate regression model (14) with a sample constructed from Compustat Fundamentals
21
.11 Average Cash Ratios .08 .09 .1 .07 −2
−1
0 Event Time JIT Adopters
1
2
Control Group
Figure 3: Changes in Average Cash Ratios for JIT and Control Firms. This figure summarizes the average cash-to-assets ratio for JIT adopters and matched control firms. The sample is constructed from Compustat Fundamentals Quarterly files. Quarterly.11 The advantage of using quarterly data, compared to annual data, is that they provide more information about how JIT adopters manage their inventory over time. The corresponding results are reported in Columns (1)-(2) of Panel A in Table 4. Evidently, JIT implementation indeed leads firms to reduce inventory, regardless of the inclusion of other control variables in the regression. Regression model (14) assumes that the effects of JIT on inventory are the same over time. Considering the fact that JIT implementation is a long-term process, I next allow for heterogeneous effects of JIT during the post-adoption period. I estimate the following first-difference specification and consider a 12-quarter (i.e., three-year) time lag. ∆inventoryi, t − t1 = β0 + β1 ∆JITi, t − t1 + β20 ∆Xi, t − t1 + σI,T + ∆i, t − t1 .
(15)
The variable ∆JITi, t − t1 is the change in the dummy variable JIT (whether implementing JIT 11 The theoretical setup in Section 3 focuses on input inventory, and accordingly, I am supposed to perform the empirical analysis using input inventory information. However, a limitation of Compustat Fundamentals Quarterly is that it does not provide disaggregated inventory data. In later subsections, in which I use annual data to estimate the structural model, I will focus on input inventory. In addition, when looking at the components of total inventory, Chen, Frank, and Wu (2005) find that there was no significant decline in finished-good inventory in U.S. publicly traded firms over the 1981-2000 period, and I find that the median input inventory holding period was shortened significantly from 43 days to 26 days. I also find that the patterns and estimates presented in Section 2 are similar when I replace total inventory with input inventory, which is measured as the sum of raw-material inventory and work-in-process inventory. Results are available upon request.
22
Table 4: Effects of JIT on Inventory and Cash Holdings (Quarterly) Table 4 reports the estimation results of the inventory and cash regressions on JIT-adoption dummy and firms’ characteristics, which include firm size, risk, market-to-book, cash flow, net working capital, capital investment, leverage, R&D, dividend payment, and acquisitions. Firm fixed effects and year fixed effects are included in the regressions, and the heteroskedasticity-consistent standard errors are reported in parentheses. Significance levels are indicated by ∗ , ∗∗ , and ∗∗∗ for 10%, 5%, and 1%, respectively. The sample is constructed from Compustat Fundamentals Quarterly for the period 1980Q1-2006Q4. (1) (2) (3) (4) Panel A: Constant Post-adoption Effect inventory inventory cash cash JIT -0.0285∗∗∗ -0.0252∗∗∗ 0.0238∗∗∗ 0.0309∗∗∗ (0.0016) (0.0060) (0.0021) (0.0097) X No Yes No Yes Firm FE Yes Yes Yes Yes Year FE × Industry FE (2-digit) Yes Yes Yes Yes Observations 22,868 4,088 22,890 4,088 Adj R-squared 0.751 0.828 0.581 0.725 Panel B: Heterogeneous Post-Adoption Effect ∆ inventoryt−(t−12) ∆ inventoryt−(t−12) ∆ casht−(t−12) ∆ casht−(t−12) ∆JIT -0.0187∗∗∗ -0.0315∗∗∗ 0.0143∗∗∗ 0.0412∗∗∗ (0.0062) (0.0062) (0.0024) (0.0121) ∆X No Yes No Yes Year FE × Industry FE (2-digit) Yes Yes Yes Yes Observations 21,143 2,370 21,175 2,370 R-squared 0.095 0.211 0.076 0.094 or not) for firm i between period t and period t1 . I report results in Panel B of Table 4. Column (2) suggests that in the first three years after adopting JIT, firms reduce inventory by 3.15%.
4.1.2
The Effect of JIT on Cash
After validating the sample, I then use it to test the model’s prediction for cash holdings. The first specification I consider is analogous to the regression equation (14), cashi,t = α0 + α1 JITi,t + α20 Xi,t + γi + σI,T + i,t ,
(16)
where the coefficient estimate on the dummy variable JITi,t is of particular interest. The variables in Xi,t are the ones used in the cash regression (1), including firm size, market-tobook ratio, cash-flow risk, etc. Columns (3)-(4) of Panel A in Table 4 report the results. The coefficient estimate of the dummy variable JITi,t is positive and statistically significant in both cases, as the model predicts. I also consider a first-difference specification, a counterpart of regression equation (15) that
23
allows for different effects of JIT on cash holdings in the post-adoption period: ∆cashi, t − t1 = α0 + α1 ∆JITi, t − t1 + α20 ∆Xi, t − t1 + σI,T + ∆i, t − t1 .
(17)
Estimation results are presented in Columns (3)-(4) of Panel B in Table 4. The results again support the model’s implication for cash holdings and are statistically significant: Firms increase cash by 4.12 percentage points within the first three years.12 Overall, the impacts found in this subsection have a correct sign and support model predictions. In the following subsections, I will numerically solve the model and quantitatively evaluate the effect of JIT on cash.
4.2
Quantitative Analysis
This subsection reports the numerical results of the dynamic stochastic model built in Section 3. I first estimate the firm’s problem in the JIC environment, using a sample of manufacturing firms constructed from the Compustat Fundamentals Annual files for the early 1980s. The sample period is chosen to ensure that JIT was not widely adopted in the U.S., as most firms used JIC during that time. I then perform counterfactual analyses to assess the impact of JIT on firms’ cash and inventory policies, by parameterizing the JIT environment with the same set of values used in JIC. Lastly, I examine the robustness of the results. I extend the baseline model in two ways. First, I separate equity and debt financing by introducing intra-period bank loans. Second, I add capital to the model and allow the firm to transfer resources across periods in the form of physical capital.
4.2.1
Model Parameterization
Considering the frequency of inventory adjustment, I solve the model at a quarterly frequency. The time period t in my model, therefore, corresponds to one quarter. I then aggregate the 12
To account for possible self-selection, I also employ the Heckman selection model and estimate the impact of JIT using the maximum-likelihood method. The estimated effect of JIT is 9%, larger than that reported in Table 4. This is possibly because the effect estimated with Heckman selection model is the local average treatment effect, reflecting the effect on the firm whose adoption decision is affected by the chosen variables (i.e., lagged independent variables X in the outcome equation). Results are available upon request.
24
simulated quarterly variables to the annual frequency. I estimate the parameters used in the JIC environment by matching firm-level and aggregatelevel annual moments in 1980s. The estimation strategy is explained below. Table 5: Model Parameterizations (Quarterly) Table 5 summarizes the parameters used to solve the JIC model at quarterly frequency. Panel A reports the parameters specifying revenue function and governing shocks, the parameters characterizing a firm’s external financing conditions, corporate income tax rate, and interest rate. Panel B presents estimation results by matching JIC-model moments to data moments in the early 1980s. Standard errors are reported in parentheses. Panel A: Parameters Calibrated Separately
risk-free rate (r) effective corporate income tax (τc ) curvature (θ) standard deviation of revenue shock (σz1 ) persistence of revenue shock (ρz1 ) standard deviation of cost shock (σz2 ) persistence of cost shock (ρz2 ) Panel B: Parameters Estimated by SMM inventory depreciation rate (δs ) fixed operating costs (cf ) costs of trade credit (λ1 ) linear costs of external finance (λ2 )
0.017 0.35 0.51 0.14 0.69 0.019 0.98 0.087 0.020 0.013 0.088
(0.004) (0.001) (0.000) (0.002)
The risk-free rate r at a quarterly frequency is set at 1.7%; that is, an annualized rate of 6.8%. It is chosen to match the interest rate adjusted for inflation in 1980s, where inflation is calculated using the GDP deflator. The corporate income tax τc is set at 35%, which was the average effective corporate income tax rate in 1980s. The rate is computed based on reports provided by the Economic Policy Institute. To calibrate the parameters that specify revenue function and govern revenue shocks, I construct a sample of manufacturing firms (SIC 2000-3999) for the period 1980-1989 from Compustat and use it to estimate the following regression model:
log Yi,t = α0 + α1 log Ni,t + f irmi + yeart + i,t .
(18)
Here, Yi,t is the real sales of firm i in year t, and Ni,t is real material use measured as the difference between the cost of goods sold and labor costs divided by the material price index. Firm fixed effects and time fixed effects are also included to control for firm-specific time-
25
invariant characteristics and common macroeconomic shocks across firms, respectively. The coefficient of material use α1 corresponds to θ in the model. Estimating equation (18) yields an estimate of the curvature of the revenue function, α ˆ 1 = 0.51. I therefore set θ to be 0.51. The error term in the regression equation (18) is the empirical counterpart of the logarithm of the revenue shock in the model. To calibrate persistence ρz1 and volatility σz1 , I collect the predicted residuals and estimate the following regression:
ˆi,t = βˆ i,t−1 + εi,t .
(19)
This gives an estimate of persistence at an annual frequency, βˆ = 0.23, corresponding to the 1
persistence at a quarterly frequency of ρz1 = 0.23 4 = 0.69. The standard deviation of the revenue shock σz1 is computed as the average within-firm standard deviation of the residuals εˆi,t . This yields a standard deviation of 0.186 at an annual frequency, implying σz1 = 0.14 at a quarterly frequency. I calibrate the persistence and volatility of cost shocks z2 in a similar way. Using the price index for material costs reported in the NBER-CES Manufacturing Industry Database, I estimate an AR(1) regression model and find cost shocks to be highly persistent and relatively smooth, with ρz2 = 0.98 and σz2 = 0.019. The remaining parameters are estimated using simulated method of moments (SMM). That is, I estimate parameters by minimizing the distance between the moments constructed from model-simulated data and the moments computed with actual data. I choose the following moments to match: average cash-to-assets ratio, average inventory-to-assets ratio, average external financing-to-assets ratio, average trade credit-to-assets ratio, and within-firm standard deviation of cash-to-assets ratio. Average cash-to-assets ratio is informative about the fixed operating costs cf . Larger fixed operating costs drive up cash holdings, because the presence of the fixed costs makes the firm cautious and leads it to accumulate cash to facilitate operations. Average inventory-to-assets ratio can be used to infer inventory carrying costs, which are captured by depreciation rate δs . Intuitively, the firm is reluctant to hold inventory if it faces high carrying costs. Average external financing-to-assets ratio provides information on the parameter λ2 . Lower external borrowing costs relax financial frictions and encourage firms
26
to make more use of external funds. Similarly, the average trade credit-to-assets ratio can be used to infer average discount rates of trade credit λ1 . Lastly, I target within-firm standard deviation of cash-to-assets ratio. This moment is important to distinguish between discount rates λ1 and fixed operating costs cf . An increase in either of these parameters will drive up average cash ratio; however, they have different implications for within-firm cash volatility. Intuitively, fixed operating costs are constant across periods. Given external borrowing costs λ2 , the need for cash to fund fixed operating costs barely changes over time. In contrast, discount rates determine the choice between cash and trade credit for funding material purchases, which tend to move with shock realizations. The variation in need for cash to purchase materials over time can generate volatile cash holdings. Results are summarized in Table 5. Panel A presents the parameters that are directly calibrated from data, and Panel B reports parameters that are estimated jointly using SMM.
4.2.2
Simulated Moments under JIC
I simulate the JIC economy over 1,050 quarters, drop the first 50 observations to limit the effect of the initial condition, and construct the annual quantity from the quarterly. Moments are computed by averaging over 1,000 simulations. Table 6 reports the simulated moments. Panel A shows the selected moments used for estimation, and Panel B presents nontargeted moments for cash holdings, inventory holdings, trade credit, and external financing, which are used to validate the model. Cash holdings are the sum of cash, cash equivalents, and short-term investments (item che). Inventory is measured as the sum of raw-material inventory (item invrm) and work-in-process inventory (item invwip), given the theoretical setup in Section 3. Trade credit is measured as firms’ accounts payable (item ap). External financing includes both debt and equity issuance, constructed as the sum of the sale of common and preferred stock (item sstk) and debt in current liabilities (item dlc). The model performs well in reproducing data moments. The estimation targets the average cash-to-assets ratio, average inventory-to-assets ratio, average trade credit-to-assets ratio, average external financing-to-assets ratio, and standard deviation of cash-to-assets ratio. These model moments match data closely. In addition, the model-implied correlations between sales and variables of interest all have the correct signs and are consistent with their data coun27
Table 6: Moments under the JIC System Table 6 reports data and simulated moments from the JIC environment. Panel A shows the moments used for parameter estimation, and Panel B presents nontargeted moments for cash, inventory, trade credit, external financing, and sales. The data moments are constructed with the sample of manufacturing firms in Compustat from 1979 to 1980. Moments Data (1979-1980) JIC Panel A: Targeted Model Moments average cash to assets (ct+1 /(ct+1 + st+1 + Kt+1 )) 0.088 0.088 average inventory to assets (st+1 /(ct+1 + st+1 + Kt+1 )) 0.193 0.184 average trade credit to assets (Tt /(ct+1 + st+1 + Kt+1 )) 0.117 0.106 average external financing to assets (et /(ct+1 + st+1 + Kt+1 ) when et < 0) 0.095 0.085 standard deviation of cash to assets 0.035 0.032 Panel B: Nontargeted Model Moments (i) correlations correlation between cash savings and sales -0.02 -0.03 correlation between inventory investment and sales 0.19 0.18 correlation between trade credit and sales 0.14 0.38 (ii) standard deviation standard deviation of inventory to assets 0.02 0.05 standard deviation of trade credit to assets 0.02 0.09 standard deviation of external financing to assets 0.04 0.07 (iii) serial correlation sales 0.83 0.80
terparts. However, compared to the actual data, the model generates slightly more volatile inventory holdings, trade credit, and external financing. Overall, the model is able to replicate key features of the data. This strengthens the model’s reliability and the validity of using it to do counterfactual exercises.
4.2.3
Simulated Moments under JIT
In this subsection, I examine the quantitative implications of the JIT model and estimate the marginal effect of JIT adoption on firms’ cash and inventory management. To that end, I perform a counterfactual exercise by setting all parameters in the JIT environment to their values in the JIC environment. That is, I hold all factors unchanged except for inventory system. Results are reported in the column titled JIT of Table 7. JIT allows firms to access resources that would otherwise be tied up in materials and products, and save on the costs of utilities and space incurred in keeping warehoused inventory from perishing. In this environment, the firm, on average, has a cash-to-assets ratio of 16.8%, which implies that the production flexibility introduced by JIT leads the firm to increase its cash
28
Table 7: Implications under the JIT System Table 7 reports data and simulated moments from the JIT environment, including the moments for cash, inventory, trade credit, and sales. The data moments are constructed with the sample of manufacturing firms in Compustat during the period 2000-2009. JIT with Moments Data (2000-2009) JIT cheaper trade credit (i) cash average cash to assets 0.234 0.168 0.147 standard deviation of cash to assets 0.093 0.041 0.053 correlation between cash savings and sales 0.027 0.188 0.216 (ii) inventory average inventory to assets 0.084 0.001 0.001 standard deviation of inventory to assets 0.027 0.000 0.000 correlation between inventory investment and sales 0.104 -0.100 -0.096 (iii) trade credit average trade credit to assets 0.114 0.098 0.110 standard deviation of trade credit to assets 0.048 0.129 0.165 correlation between trade credit and sales 0.03 0.295 0.350 (iv) sales serial correlation of sales 0.81 0.66 0.72
ratio by 8%. In addition, the correlation between cash savings and sales changes from negative under JIC to positive under the new system, which is consistent with data. On the other hand, the firm, on average, holds zero inventory, which is the goal of JIT inventory management. The stockout motive for inventory holdings is absent under the new system, so the firm no longer needs to keep a large amount of inventory on hand. Instead, it holds cash and replenishes inventory stock before production at the beginning of each period. This, in turn, explains the nearly invariable inventory ratios, indicated by its low standard deviation. Firms implementing JIT usually try to develop a nurturing and reliable long-term relationship with their suppliers. More friendly terms and conditions of trade credit may be obtained by entering this kind of long-term commitment. I capture this feature by lowering the costs of trade credit. Due to the lack of information on trade credit contracts between JIT adopters and their suppliers, I cut the costs of trade credit by 10% and estimate the marginal effect of the long-term relationship on cash holdings. Results are summarized in the column JIT with cheaper trade credit of Table 7. Cheaper trade credit lowers the value of cash in preserving financial flexibility. The reduced costs, from 1.34% to 1.2%, lead to a drop in the average cash ratio by 2.1 percentage points.
29
In short, this subsection estimates the marginal effect of JIT on cash holdings. It suggests that if everything else remains unchanged and all firms in the economy switch from JIC to JIT and achieve supply-chain perfection, the average cash-to-assets ratio will rise by 8%. If this switch is accompanied by a 10% reduction in the cost of trade credit, the increase in the average cash ratio will drop slightly to 5.9%.
4.2.4
JIT Implementation and the Rise in Corporate Cash
The next exercise is to quantify the fraction of the observed cash increase that can be attributed to JIT implementation. To perform the analysis, I need to model the adoption decision to control for self-selection bias and determine the percentage of firms using JIT in the economy. One key barrier to JIT adoption is high costs—both the direct costs of system adoption and the indirect costs of adapting to the new system.13 As a result, resource-rich firms are more likely to be able to afford adoption and implement JIT systems than resource-poor firms (White, Pearson, and Wilson, 1999). Taking this into account, I control for self-selection by making the following assumption. I suppose that shifting from JIC to JIT requires a one-time fixed cost C, which is identical for all firms. Each period, after the realization of shocks, firms operating under JIC weigh the expected benefits of the switchover against the cost and make decisions. If the adoption benefits are greater relative to the cost C, firms decide to implement JIT. The switchover is irreversible.14 Moreover, according to the report Physical Risks to the Supply Chain by CFO Research Services, nearly two-thirds of manufacturing firms had implemented JIT by 2008. Compdata’s surveys provide figures close to this value. I then calibrate the adoption cost C to match the adoption rate in the data. Specifically, I simulate a sample of 3,000 firms for 212 periods, by starting from the same initial state {z1,1 , z2,1 , c1 , s1 } and drawing 3,000 sequences of revenue shocks εz1 and cost shocks εz2 from the distribution N (0, σz21 ) and N (0, σz22 ), respectively. For the first 100 periods (i.e., 25 years), all firms are restricted to operating under JIC. From the period 101, firms are 13
The costs, for example, include extensive training for employees in different areas and high initial investment—from purchasing more sophisticated equipment, information systems and software to changing the physical layout of facilities. 14 The assumption of irreversibility can be justified by the fact that multiple management practices are typically associated with JIT, which require changes throughout the entire organization, from the way inputs are ordered to the role of people working on the shop floor. Therefore, switching back to JIC from JIT is very costly, and rarely observed in the data.
30
allowed to select between JIC and JIT. Once they have switched over, firms operate under JIT permanently. Prospective adopters make adoption decisions in each period. I choose C such that two-thirds (approximately 66.7%) of firms are JIT adopters after 112 periods (i.e., 28 years) since JIT has become an option. Results are summarized in the column Self-selection of Table 8. Those moments are computed based on the simulated data from period 181 to period 212 (i.e., year 20 to year 28 after allowing a switchover). Table 8: Moments after Controlling for Self-selection Table 8 reports simulated model moments after controlling for self-selection bias and taking into account the fraction of firms using JIT in the economy. It presents moments for cash, inventory, trade credit, and sales. The data moments are constructed with the sample of manufacturing firms in Compustat from 2000 to 2009. Self-selection with Moments Data (2000-2009) Self-selection cheaper trade credit (i) cash average cash to assets 0.234 0.147 0.129 standard deviation of cash to assets 0.093 0.033 0.026 correlation between cash savings and sales 0.027 0.034 0.004 (ii) inventory average inventory to assets 0.084 0.062 0.062 standard deviation of inventory to assets 0.027 0.020 0.020 correlation between inventory investment and sales 0.104 0.052 0.104 (iii) trade credit average trade credit to assets 0.114 0.092 0.110 standard deviation of trade credit to assets 0.048 0.093 0.097 correlation between trade credit and sales 0.03 0.327 0.359 (iv) sales serial correlation of sales 0.81 0.81 0.75
After considering non-adopters in the economy and controlling for self-selection, the average cash ratio drops to 14.7%. In addition, cash savings are still positively correlated with sales, and the correlation becomes closer to that observed in the data. Inventory-related model moments also become quantitatively similar to their data counterparts. The mean and standard deviation of the inventory-to-assets ratio climb up, close to the actual values. The correlation between inventory investment and sales has the correct sign. I then perform the same exercise for the case with cheaper trade credit and report results in the column Self-selection with cheaper trade credit of Table 8. The average cash ratio drops to 12.9%, while the average trade credit ratio increases and moves closer to its data counterpart, which in turn provides confidence in the choice of the new costs of trade credit under JIT. Given that many factors that possibly affect firms’ real and financial decisions have been
31
changing over the last three decades, I do not expect the model to reproduce the data moments observed in the 2000s. However, Table 8 shows that the model performs reasonably well and suggests that 4.1 percentage points of the observed rise in cash can be attributed to JIT adoption and the associated long-term relationships between manufacturers and their suppliers.
4.3
Robustness
In this subsection, I evaluate the robustness of my main results to different model assumptions and different samples. I consider two model extensions: (i) separately modeling debt and equity financing, and (ii) adding physical capital to the baseline model. I also estimate the baseline model using the JIT sample used in Section 4.1.
4.3.1
Debt Financing
The firm is allowed to use intra-period bank loans to fund various expenses, subject to a borrowing constraint, b ≤ κ. Each period, the firm borrows a loan at a risk-free rate r before production and repays it at the end of the period. To examine the robustness of the main results, I follow the estimation procedures described in previous subsections and re-estimate all the parameters reported in Panel B of Table 5, along with borrowing constraint κ. I estimate λ2 and κ by matching average equity issuance-to-assets ratio and average short term debt-to-assets ratio. With the new parameter estimates, I then solve and simulate the JIC and JIT economies, construct corresponding model moments, and summarize them in the column titled Debt of Table 9. Separating cheaper debt from equity while matching moments on external funds leads to a higher estimate of λ2 , compared to the baseline model. Since debt is a limited source of external financing, the possibility of hitting the borrowing limit κ and resorting to expensive equity financing in the future makes the firm cautious, and induces it to accumulate cash (Armenter and Hnatkovska, 2016). The presence of a cheaper source of external financing, however, reduces the value of cash holdings and therefore requires a larger fixed operating cost cf to match the average cash ratio in the early 1980s. Implementing JIT provides the firm with cheaper materials by reducing the carrying costs of inventory holdings, and the firm therefore scales up operations and uses more materials. 32
Table 9: Model Robustness Table 9 summarizes estimation results for extended models and for different samples. Panel A reports the parameters calibrated separately from data, Panel B presents the parameters estimated by SMM, Panel C shows the moments selected to match, and Panel D reports the model-implied cash and inventory moments under JIT. Standard errors are reported in parentheses. Debt Capital JIT Sample Panel A: Parameters Calibrated Separately risk-free rate (r) 0.017 0.017 0.017 effective corporate income tax (τc ) 0.35 0.35 0.35 revenue elasticity of materials (θ) 0.51 0.43 0.55 standard deviation of revenue shock (σz1 ) 0.14 0.13 0.086 persistence of revenue shock (ρz1 ) 0.69 0.70 0.80 standard deviation of cost shock (σz2 ) 0.019 – 0.13 persistence of cost shock (ρz2 ) 0.98 – 0.94 revenue elasticity of capital (α) – 0.20 – capital depreciation rate (δk ) – 0.03 – capital linear adjustment costs (γ1 ) – 0.039 – capital convex adjustment costs (γ2 ) – 0.70 – Panel B: Parameters Estimated by SMM inventory depreciation rate (δs ) 0.071 (0.001) 0.032 (0.000) 0.118 (0.002) fixed operating costs (cf ) 0.046 (0.002) 0.035 (0.001) 0.020 (0.001) costs of trade credit (λ1 ) 0.013 (0.000) 0.010 (0.000) 0.015 (0.000) linear costs of external finance (λ2 ) 0.20 (0.020) 0.067 (0.001) 0.089 (0.002) borrowing constraint (κ) 0.08 (0.002) – – Panel C: Targeted Moments under JIC Data Model Data Model Data Model average cash to assets 0.088 0.077 0.088 0.088 0.067 0.070 average inventory to assets 0.193 0.194 0.193 0.188 0.204 0.195 average external financing to assets − − 0.095 0.084 0.058 0.114 average trade credit to assets 0.117 0.110 0.117 0.114 0.087 0.080 standard deviation of cash to assets 0.035 0.036 0.035 0.032 0.025 0.028 average short-term-debt to assets 0.084 0.055 – – – – average equity issuance to assets 0.015 0.053 – – – – Panel D: Implied Cash and Inventory under JIT Data Model Data Model Data Model average cash to assets 0.234 0.132 0.234 0.171 0.129 0.128 average inventory to assets 0.084 0.001 0.084 0.004 0.085 0.003
To facilitate purchases, the firm needs more liquidity composed of internal and external funds and chooses to raise its cash balance by 5.5% under JIT. The increase in cash ratio is smaller compared to the baseline model, because the availability of cheap debt for purchasing material inputs limits the role of internal cash balances. Note that a significant portion of short-term debt is lines of credit that have been used (Sufi, 2009). Lines of credit, however, are generally used to finance future growth options instead of serving operational purposes (Lins, Servaes, and Tufano, 2010). The estimate obtained here, therefore, can be treated as the lower bound of the marginal effect of JIT.
33
4.3.2
Physical Capital
The second extension concerns physical capital. Aside from cash and inventory holdings, the firm can transfer resources across periods in the form of physical capital. It combines capital K and materials N to produce goods, and faces a Cobb-Douglas revenue function:
F (z, k, N ) = z1 K α N θ .
(20)
Compared to the revenue function specified in the baseline model, this function allows capital to vary over time and deviate from 1, and introduces a new parameter, α < 1, which describes the elasticity of revenue with respect to capital. The firm augments its capital stock each period by capital investment, I, given as I = K 0 − (1 − δk )K.
(21)
The parameter δk is the capital-depreciation rate, 0 < δk < 1. Adjusting capital by purchasing or selling it incurs adjustment costs, which are defined by
A(K, I) = γ1 K1I6=0 +
γ2 I 2 ( ) K. 2 K
(22)
This specification includes both the linear and convex adjustment costs (Cooper and Haltiwanger, 2006). The parameter γ2 > 0 captures the smoothing effect. I estimate the new model as follows. I first estimate the following regression model to calibrate the parameters of the augmented revenue function:
log Yi,t = β0 + β1 log Ki,t + β2 log Ni,t + f irmi + yeart + i,t ,
(23)
and find that the estimates of the revenue elasticity of capital α and revenue elasticity of materials θ are 0.20 and 0.43, respectively. I then use the residuals ˆi,t to reparameterize the revenue shock process and get ρz1 = 0.70 and σz1 = 0.13. The ratio of the annual depreciation to the gross capital stock is 0.12, derived from the Compustat manufacturing sector in the 1980s. I therefore set δk equal to 3% per quarter. Following Cooper and Haltiwanger (2006) 34
and Nikolov and Whited (2014), I choose γ1 to be 0.039 and γ2 to be 0.7. Lastly, I estimate the remaining parameters and report simulated model moments in the column titled Capital of Table 9. Adding physical capital to the model introduces another use for cash: financing capital investment. It strengthens precautionary cash demands and raises the volatility of cash holdings. Relative to the baseline model, the estimated fixed operating cost is larger in order to smooth out the volatility and match the data counterpart, and the costs of trade credit and external financing, λ1 and λ2 , are lower in order to reduce the stronger cash demand. Switching to JIT lowers material costs and prompts firms to employ more materials in productions and augment their capital stocks, which therefore generates a higher demand for liquidity. On the other hand, lower external borrowing costs reduce the value of cash and, in turn, cash demand. These two effects offset each other and leave the net increase in cash ratio barely changed relative to the baseline model, 8.3% vs. 8.0%.
4.3.3
Subsample Estimation
The last exercise is to estimate the baseline model with the JIT sample used in Section 4.1 and examine whether the model is able to explain their behavior. I first recalibrate revenue function, the revenue shock process, and the cost shock process, then re-estimate the remaining parameters. The column titled JIT Sample in Table 9 presents parameter estimates, targeted data moments, and simulated model moments. Relative to an average manufacturing firm, early JIT adopters face less volatile revenue shocks but more volatile cost shocks. The former weakens the precautionary motive for saving cash, while the latter prompts firms to hold more inventory to hedge against uncertainty about future input prices. In addition, early JIT adopters face higher inventory carrying costs, which, to some degree, explains why they become pioneers in adopting JIT. The model predicts that when these JIT pioneers shift from JIC to JIT, their average cash ratio will rise by 5.8%, conditional on the assumptions of perfect supply chains and complete implementation of JIT. The estimate is higher than that obtained using DID in Section 4.1.2. The difference in the estimated effects can possibly be reconciled with two facts: the partial 35
adoption of complete JIT systems (White, Pearson, and Wilson, 1999) and imperfect supply chains in reality, as indicated by the non-zero inventory holdings by JIT pioneers in 2000s.
5
Conclusion
Over the last three decades, the U.S. manufacturing sector has gradually shifted resources from inventory to cash. In this paper, I propose an explanation—the implementation of a JIT inventory system—to understand the phenomenon, and in turn shed light on cash hoarding, which has recently attracted extensive attention. I begin by documenting the stylized facts about the strong negative association between cash and inventory holdings. I then develop a structural model to explain the phenomenon by exploring how JIT influences inventory and cash policies and then quantify its effects. In the model, I emphasize the transaction motive for cash savings. Adopting JIT helps firms eliminate non-value-added inventory; it also allows them to allocate released resources to cash, in order to expand operations and facilitate transactions with suppliers without tapping into expensive external finance. I provide empirical evidence to support the role of JIT adoption in reducing inventory and raising cash, and show that the estimated structural model is able to reproduce a high negative correlation between cash and inventory, as well as other key empirical features. I then use the model to conduct a counterfactual exercise, which suggests that JIT adoption can account for a 4.1 percentage point increase in cash-to-assets ratio. There is lively debate about the causes of corporate cash hoarding, which raises concerns about possible resource misallocation from physical capital to cash. My results suggest that approximately 28% of the changes in the average cash ratio since 1980 can be rationalized as a normal cash investment.
36
References Armenter, R. and V. Hnatkovska (2016). Taxes and capital structure: Understanding firms’ savings. Manuscript, University of British Columbia. Azar, J., J.-F. Kagy, and M. Schmalz (2016). Can changes in the cost of carry explain the dynamics of corporate cashholdings? Review of Financial Studies 29 (5), 1–54. Bates, T., K. Kahle, and R. Stulz (2009). Why Do U.S. Firms Hold So Much More Cash than They Used To? Journal of Finance 64 (5), 1985–2021. Baumol, W. (1952). The Transactions Demand for Cash: An Inventory Theoretic Approach. Quarterly Journal of Economics 66 (4), 545–556. Boileau, M. and N. Moyen (2016). Corporate Cash Holdings and Credit Line Usage. International Economic Review 57 (4), 1481–1506. Bolton, P., H. Chen, and N. Wang (2011). A Unified Theory of Tobin’s q, Corporate Investment, Financing, and Risk Management. Journal of Finance 66 (5), 1545–1578. Chen, H., M. Frank, and O. Wu (2005). What Actually Happened to the Inventories of American Companies between 1981 and 2000? Management Science 51 (7), 1015–1031. Cooper, R. and J. Haltiwanger (2006). On the Nature of Capital Adjustment Costs. Review of Economic Studies 73 (3), 611–633. Falato, A., D. Kadyrzhanova, and J. W. Sim (2013). Rising Intangible Capital, Shrinking Debt Capacity, and the US Corporate Savings Glut. Manuscript, Federal Reserve Board. Hugonnier, J., S. Malamud, and E. Morellec (2015). Capital Supply Uncertainty, Cash Holdings, and Investment. Review of Financial Studies 28 (2), 391–445. Karabarbounis, L. and B. Neiman (2012). Declining Labor Shares and the Global Rise of Corporate Saving. Manuscript, University of Chicago. Kinney, M. and W. Wempe (2002). Further Evidence on the Extent and Origins of JIT’s Profitability Effects. The Accounting Review 77 (1), 203–225. 37
Kulchania, M. and S. Thomas (2017). Supply Chain Disruption Costs and the Substitution of Cash for Inventory. Manuscript, University of Pittsburgh. Lins, K. V., H. Servaes, and P. Tufano (2010). What Drives Corporate Liquidity? An International Survey of Cash Holdings and Lines of Credit. Journal of Financial Economics 98 (1), 160–176. Lyandres, E. and B. Palazzo (2016). Cash Holdings, Competition, and Innovation. Journal of Financial and Quantitative Analysis, forthcoming. Ma, L., A. Mello, and Y. Wu (2013). Industry Competition, Winner’s Advantage, and Cash Holdings. Manuscript, University of Wisconsin - Madison. Morellec, E., B. Nikolov, and F. Zucchi (2013). Competition, Cash Holdings, and Financing Decisions. Manuscript, Ecole Polytechnique Fdrale de Lausanne. Morgan, D. (1991). Will Just-In-Time Inventory Techniques Dampen Recessions?
Federal
Reserve Bank of Kansas City Economic Review 2, 21–33. Nikolov, B. and T. Whited (2014). Agency Conflicts and Cash: Estimates from a Dynamic Model. Journal of Finance 69 (5), 1883–1921. Pinkowitz, L. and R. Williamson (2001). Bank Power and Cash Holdings: Evidence From Japan. Review of Financial Studies 14 (4), 1059. Riddick, L. and T. Whited (2009). The Corporate Propensity to Save. Journal of Finance 64 (4), 1729–1766. Sufi, A. (2009). Bank Lines of Credit in Corporate Finance : An Empirical Analysis. Review of Financial Studies 22 (3), 1057–1088. White, R., J. Pearson, and J. Wilson (1999). JIT Manufacturing: A Survey of Implementations in Small and Large U.S. Manufacturers. Management Science 45 (1), 1–15. Zhao, J. (2016). Accounting for the Corporate Cash Increase. Manuscript, Peking University HSBC Business School.
38
A A.1
Appendix Variable Definitions
Following Bates, Kahle, and Stulz (2009), I construct the sample from Compustat and define the variables used in the cash and inventory regressions as follows: Cash is defined as the ratio of cash and short-term investments over total assets; Inventory is the ratio of total inventories over total assets; Firm size is the natural logarithm of total assets; Risk is computed as the standard deviation of operating cash flow-to-assets ratio in the previous five periods, with operating cash flow defined as earnings after interest, dividends, and taxes but before depreciation; Market-to-book ratio is the sum of market value and debt over total assets; Net working capital is equal to working capital net of cash and inventory over total assets; Capital investment is the ratio of capital expenditures over total assets; Leverage is the sum of long-term debt and debt in current liabilities normalized by total assets; R&D investment is the ratio of research and development expenses to total assets; Dividend is a dummy variable taking a value of one if the dividend payout (common) is nonzero; Acquisitions is the ratio of acquisitions over total assets; Capital investment (Quarterly) is the ratio of capital expenditures over total assets, with capital expenditures defined as the first difference in gross property, plant, and equipment. Concentration ratio-50 is the share of value of shipments accounted for by 50 largest firms in the industry.
A.2
Description of the JIT-adopter Sample
Panel A of Table A1 shows the distribution of JIT adoption year for the sample of 149 JIT adopters. About 8% of the firms in the sample adopted JIT in the first half of the 1980s (1980A1
1984), with the earliest in 1982. More than 40% of the sample firms implemented JIT in the second half of the 1980s. Panel B of Table A1 reports the distribution of adopters by two-digit SIC industry. A large proportion (approximately 70%) of adopters operate in four industries. In order by number, they are: electronic equipment (SIC 36, 26.8%), industrial equipment (SIC 35, 21.4%), instrumentation (SIC 38, 12.7%), and motor vehicles (SIC 37, 8.0%). The rest of the adopters in the sample are relatively evenly distributed among other industries. Table A1: Descriptive Statistics for 149 JIT Adopters Panel A: Distribution of JIT Adoption Years Year Number of Firms Distribution 1982 3 2.0% 1983 2 1.3% 1984 7 4.7% 1985 10 6.7% 1986 9 6.0% 1987 10 6.7% 1988 15 10.1% 1989 21 14.1% 1990 23 15.4% 1991 19 12.8% 1992 12 8.1% 1993 11 7.4% 1994 5 3.4% 1995 2 1.3% Total 149 100% Panel B: Distribution of Two-Digit Industry Classifications 2-Digit Industry Number of Firms Distribution 20 Food 1 0.7% 22 Textile 3 2.0% 23 Apparel 2 1.3% 25 Furniture 5 3.4% 26 Paper 3 2.0% 27 Printing, publishing 5 3.4% 28 Chemicals 1 0.7% 30 Rubber and plastics 4 2.7% 31 Leather 1 0.7% 32 Concrete Products 1 0.7% 33 Primary metals 8 5.4% 34 Fabricated metals 8 5.4% 35 Industrial equipment 32 21.4% 36 Electronic equipment 40 26.8% 37 Motor vehicles 12 8.0% 38 Instrumentation 19 12.7% 39 Other manufacturing 4 2.7% Total 149 100%
A2
