Does Cash Hoarding Represent Excessive Conservatism or Opportunism of CEOs in pre-and post- crisis of 2008?

The aim of this paper is to define and isolate the drivers of speculative cash hoarding phenomena in publicly traded non-financial American firms before and after the financial crisis of 2008. The model presented in this paper uses previous literature defined drivers of transaction and precautionary motives for holding cash as control variables, and introduces new measures for market endorsement and economic conditions, that were never used before in this context. Empirical results show that cash accumulation is prominent among profitable firms, and that cash accumulation in excess of ongoing and precautionary needs of companies was endorsed by the financial markets both before and after the financial crisis of 2008. Results also support previous findings that financially unconstrained firms have a lesser need for market endorsement for cash hoarding. The contribution of this paper to the body of literature related to cash management and cash hoarding is five-fold: (i) it provides empirical evidence of market-endorsed cash hoarding in the US non-financial companies; (ii) it extends the existing models of transaction-related cash holdings to include, formally define and test the drivers of precautionary cash accumulation in the non-financial sector; (iii) it formally defines and tests a metric for market sentiment related to cash hoarding before and after the financial crisis of 2008; (iv) provides empirical evidence that the crisis of 2008 was anticipated by the corporate sector as early as 2005, and recovery from the crisis was predicted as early as Q2:08; and (v) defines and quantifies the border separating the precautionary and speculative motives for cash accumulation.


I. Introduction
For the last three decades, American non-financial companies have been choosing to hold immense reserves of cash. The motivations behind such behaviour, although of great interest to researchers, have eluded explanation. Underlying the research aimed to explain this phenomenon is the seminal idea proposed by Keynes (1936): firms may hold cash reserves for transaction, precautionary, and/or speculative motives.
More specifically, I will answer to the following two questions: (i) how does timevarying uncertainty about financial conditions affect cash holdings? and (ii) does cash hoarding occur with the blessing of the capital market?
To answer the first research question this dissertation will test whether the motivation for hoarding cash changed after the 2008 crisis.
This study will contribute to the existing body of literature by testing for (i) the existence of the speculative motive for holding cash, and (ii) the potential change in opportunistic behavior of CEOs in the non-financial companies before and after the crisis.
Only one theoretical paper regarding the banking sector's speculative cash management has been recently published (D. Gale and T. Yorumazer, 2013). Most of the existing literature has addressed only transaction and precautionary reasons, based on works by Opler, Pinkowitz, Stulz, and Williamson. (1999), and Bates, Kahle, and Stulz, (2009).
Under the second research question, the empirical investigation will illustrate benefits to stock price, associated with high cash positions, while simultaneously addressing the issue of the potential agency problem associated with excessive cash holdings. Bates, Kahle and Stultz (2009) briefly addressed the agency issue without directly testing for its impact. This paper directly tests for the effect of the agency problem in the context of cash hoarding. Moreover, empirical results show that most capital markets supported such a cash management practice.
The logic behind the course of action and the methodology in this paper will rely on Keynes' three definitions of motivations for holding cash: once optimal cash holdings Electronic copy available at: https://ssrn.com/abstract=3927550 required for ongoing operations (the transaction motive) and factors related to the precautionary motive are accounted for, the residuals must be attributed to the speculative motive. The extended model presented in this paper treats variables used in Opler et al. (1999) and Bates et al. (2009) models as controls for the transaction motive for holding cash, while those identified in Garcia-Appendini, and Montoriaol-Garriga, (2010), Duchin, Ozbas, and Sensoy, (2010) and Dittmar and Mahrt-Smith (2007) among others, are treated as control variables for the precautionary motive. An additional variable attributed to speculative motive is introduced and tested for its power to explain additional cash accumulation in excess of the two abovementioned motives.
As expected, data shows the presence of opportunistic behavior in cash management before the 2008 crisis. In the years after the crisis, the practice continued but precautionary motivation was also periodically present except among highly profitable corporations. I also find that cash accumulation was endorsed by the markets.

II. Literature review
The ultimate goal of financial management is the maximization of shareholders' wealth. Managerial actions that can lead to achieving this goal span a large number of functions, but one of the more important of them all -continuous search and investment to assure firms' growth, competitiveness and viability. Classic theory addresses the need in mixing various types of financing instruments such as debt and equity [ 7 ], while equity financing is subdivided into internally generated funds ( retained earnings) and externally raised equity. The source of retained earnings, ongoing operations and profitability of the firm, but is also subject to the dividend payout policy. The resulting cash reserves are used for financing new investments. The theory implies that in absence of investment opportunities the firm should distribute earnings to the shareholders, and since holding cash provides nearly zero return, firms should hold cash only to the point where the marginal benefit from holding cash and marketable securities equals the marginal cost of raising external financing. This concept may be extended to include net working capital management and not only cash and marketable securities [ 8 The reality is much more complex than the theory, as the presence of numerous market imperfections makes the task of managing internally and externally generated funds a challenge, and may also create conflicts of interests between the shareholders and the management, known as the Agency Problem [ ].

II-a The theoreticians
Many researches tried to define optimal amounts of liquidity a company should hold in their reserves. Keynes set the framework by defining three reasons for holding cash, and the literature that followed addressed various possibilities and suggested solutions. Early works by Baumol (1952) and Beranek (1963) were attempting to determine optimal cash balances needed for ongoing operations of the firm in the short term. The Miller-Orr (1966) model enhanced those works by both introducing uncertainty into firm's cash flows as well as accounting for the dynamic nature of the companies' inflows and outflows. Notably, Miller was among the first to discuss the possibility of speculative cash holdings, but framed it as a possible benefit for risk management, i.e. purchasing assets at attractive prices. Those models consider no minimal level of cash holdings, leaving to the managerial discretion, and are inventory management style models.
Edward Whalen (1965Whalen ( , 1966 directly followed Keynes' three motives for holding cash, and noted that both precautionary and speculative motives are driven by uncertainties, but while the precautionary is addressing the uncertainty associated with the pattern of receipts and distribution, the speculative, in his opinion, is driven by uncertainty in interest rates, and de facto is driven by desire to store wealth and is a reflection of the investment demand for cash . He concluded that based on Keynes' framework, transactional and precautionary motives are more closely related than precautionary and speculative motives of cash management. His argument contradicted Johnson (1961) which argued that the speculative motive is "really the precautionary in disguise" (p.8). Michalski (2008) revisited Whalen's concept and illustrated the interrelationship between reasons for holding cash and sources of risk in the chart below. Mork, Shleifer, and Vishny (1990) focused on the drivers behind investments-related activities in corporations. They looked at four different models/concepts: (i) the stock market is a passive predictor of future activity, i.e. managers do not learn from or rely on the market to make investment decisions; (ii) managers rely on the market only as a source of information which may or may not be correct about future fundamentals; (iii) the stock market affects the cost of financing and hence indirectly affects investment decisions; (iv) managers have to account for investors' opinions and hence the market has an impact on corporate investment decisions. They found that overall, very little if any explanatory power is attributed to the stock market sentiment and the most powerful predictors were the fundamentals.[ 10 A massive new wave of academic work on cash management was triggered by Kim, Mauer and Sherman [KMS] (1998) and Opler, Pinkowitz, Stulz and Williamson [OPSW] (1999). KMS argued that holdings in low return liquid assets can be explained by costly external financing, and as those costs increase, along with uncertainty of future cash flows and with higher expected returns of future investment opportunities, so will holdings in cash and marketable securities. They find that cash holdings are positively related to industry cash flow volatility. This result was supported by Harford (1999). They also noted that companies with higher market-to-book ratios will hold higher balances of liquid assets. In essence, they supported Whalen's (1966) position that the build up of liquidity is done in anticipation of future profitable investment opportunities. OPSW (1999) developed the Trade Off Theory that claims that firms trade off the costs and benefits of holding cash to derive optimal cash levels. They not only consider the transaction costs motive, but also the effect of asymmetric information and the agency costs of outside financing on the demand for cash holdings. They found that firms hold more cash when they are smaller, have higher investment and R&D expenditures, better investment opportunities, when they have higher and more volatile cash flows and lower net working capital. It was found that transaction costs and costs ] due to asymmetric information are important factors in the Trade off Model. The model supports transaction and precautionary motives for holding cash, but is silent about the speculative motive.

II -b The empiricists
The literature on cash management followed a number of directions in the years that followed. On one hand, one body of studies followed Keynes' three motives, seeking drivers explaining (mostly) the transaction and precautionary motives to support or dismiss arguments for cash accumulation practices. On the other hand, a parallel body of literature was dealing with company and economy specific factors that can explain cash accumulation, including corporate governance and potential agency problems. Finally, following the crisis of 2008, this new aspect in cash hoarding was examined from various points of view. Almeida, Campello and Weisbach (2002) claimed that financially constrained firms may decide to hoard cash as they are unable to raise external capital for future investments, while cash holding for unconstrained firms seems to be irrelevant. This tendency of constrained firms to accumulate cash is even stronger during economic downturns, while financially unconstrained firms maintain a zero cash flow sensitivity effect due to business cycles. Their conclusions were supported in principle by Faulkender and Wang ( 2006), who found that the marginal value of an extra dollar held declines with larger cash holdings, higher leverage and better access to capital markets.
They also found that financially stable and cash rich firms would rather distribute dividends then repurchase shares. Carroll and Kimball (2001) argued that due to the convexity of marginal utility of consumers, they would tend to save more as a precaution and anticipation of future shocks. They claim that the precautionary reason is the driving force behind cash accumulation, as well as the liquidity constraint. Same sentiment was echoed by Mikkelson and Partch ( 2003) who argued in favor of cash accumulation without creation of an agency problem in smaller firms, further arguing that cash accumulation enhances firm value through higher R&D and lower NWC. Another angle of the same claim was discussed by Acharya, Almeida and Campello (2007). While implicitly addressing the precautionary motive for cash accumulation, they argued that for financially constrained firms, the trade-off between cash and debt levels would depend on their temporal hedging needs, and that cash cannot be viewed as negative debt.
Their findings were supported and extended by Denis and Siblikov (2010) which looked into reasons and levels of cash holdings of financially constrained firms. They concluded that higher cash positions in financially constrained firms are value enhancing.
In contrast to Acharya, Almeida and Campello (2007) and Bates, Kahle andStultz (2009), Gao, Hardford andLi ( 2013) argued that private firms hold on average 50% less cash than public firms, and this is due to agency problems. They claim that well managed firms would use excess cash to reduce leverage, while poorly managed firms would spend on lower quality investments that would hurt the value of the firm. Baum, Calgayan, Ozkan and Talavera (2006) found that firms that pay high dividends have lower sensitivity to macro effects, and only financially constrained firms and rapid growth firms have high sensitivity to macro economic uncertainty. Their findings were in line with Almeida, Campello and Weisbach (2002), and also set the tone for further research come the 2008 crisis a few years later.[ 11 A widely quoted paper by Dittmar and Mahrt-Smith (2007) took the discussion about cash accumulation to the next level by defining two costs of holding cash and cash equivalents, as well as testing for the effect of corporate governance on cash management practices. They defined the costs for holding cash as either (i) lower return investment relative to existing alternatives ( i.e. the cost of carry), or (ii) wasteful capital spending and acquisitions by managers. The first definition is in line with Keynes (1936) which stated that lower returns are a reflection of lower liquidity premium. The second definition supported arguments by Harford (1999) and Jensen (1986) that additional costs of holding cash may cause sub-optimal investments due to agency problems.
While not defining the reasons for holding cash, they evaluated the cost of holding excess cash and concluded that cash policy has little value if the firms are well governed. They concluded that governance has "a greater influence on the use, rather that the accumulation, of cash holdings" (p.627) [ ] 12 ]. This conclusion by Dittmar and Mahrt-Smith (2007) was supported by findings of Fresard and Salva (2010) that claimed investors value excess cash based on their expectations for how cash will be used.[ 13 In another widely quoted paper, Bates, Kahle and Stultz (2009) were seeking explanation for steadily increasing cash ratios, and questioning whether cash accumulation can be strictly attributed to the precautionary motive or whether there is an element of speculation and subsequent agency problem. They claim that it is the ] 11 This paper directly addresses this point by including dividend payout filter in robustness tests section 12 For this reason model used in this paper does not include any governance variables. 13 This aspect will be further investigated in this paper. precautionary motive that drives the increase in cash ratios due to changes in firms' characteristics over time. They suggested a strong precautionary effect and state that on aggregate they found no evidence for speculation or presence of an agency problem.
Strong support for the precautionary motive for holding excess cash was also reported by Duchin, Ozbas and Sensoy (2010). Most impacted are firms with low cash reserves or high short term debt, and those that are financially constrained or highly dependent on external financing.[ 14 A rare attempt for dynamic modeling for optimal cash holdings was undertaken by Decams, Mariotti, Rochet and Villeneuve (2011). In essence, they support the concept presented by Faulkender and Wang (2006), and report an increase in marginal value of cash with higher issuance costs. They conclude that "marginal value of cash increases after decline in stock prices since cash holding decreases the risk of incurring issuance costs in the short term" (p.1503).
] On the other side of the argument, Duchin (2010) claimed that diversified firms would hold lower cash reserves than specialized firms due to a less than perfect correlation between subsidiaries resulting from "diversification in investment opportunity and cash flow".
Two papers highlighted the strategic value of cash holdings. Fresard (2010) claims that cash holdings affect future performance of the firm in product market. Falling short of making connection to Keynes' reasons for holding cash, he has shown that since companies do not exist in isolation but rather as a part of a market, cash holdings improve their competitive position and block new entrants into the industry. This can be interpreted as both precautionary and speculative per Keynes' definitions. Boutin, Cestone, Fumagalli and Pica (2013) arrived at the same conclusion, arguing that cash hoarding can play a protective role for firms in the industry against potential newcomers.

II -c The post 2008 crisis analysts
The crisis of 2008 provided researches with a new angle for studies of corporate cash management. A number of studies looked at the crossover between financial and strategic considerations. Garcia-Appendini and Montoriaol-Garriga ( 2010) addressed the practice of more stable and liquid firms to support their clientele base by extending the pre-crisis terms of credit. This ability to help clients had its limitation, and at some point during the crisis those firms were forced to reduce trade credits to refill their own cash positions. In this case, the accumulation of cash is justified as a precautionary measure "just in case" firms would have to support their clients. In general, it is in line with the existing body of literature (Peterson andRajan (1997), Wilner (2000), Cunat ( 2007)) that shows that suppliers would be willing to support their clients due to a vested interest in maintaining long-term business relationships, as long as it does not jeopardize their own liquidity positions. Highly liquid firms did not reduce investments during the crisis either, according to Duchin, Ozbas and Sensoy (2010). They find 'seemingly excess cash' is positively related to post-crisis investment and "relation between seemingly excess cash and post crisis investment is stronger for financially constrained firms, and during the crisis, so called excess cash is not in fact excessive" (p.431). They claim that the findings support the interpretation that financial liquidity had a value enhancing impact on investments during the crisis. They tested and found no evidence of an agency problem. The authors were critical about the existing models and body of literature that failed to account for the risks of an extreme credit crunch that would significantly dry up external sources of funding. They also claimed that the crisis was NOT anticipated by the market. This argument was also supported by Fahlenbrach and Stultz (2011) who also claimed that bank CEOs did not reduce their holdings and suffered losses because they did not anticipate the crisis. Cornett, McNutt, Strahan and Tehranian (2011) concluded that in their attempt to manage liquidity during the crisis, banks limited lending and credit supply, and this may have caused the companies to retain cash for precautionary reasons even after the crisis. As noted by Barnes and Pancost (2010) [ 15 Chen and Shane (2014) examined the effects of sub-optimal cash holdings using Bates et al. (2009) framework, and concluded that while market reaction to abnormal increases or decreases in cash holdings is asymmetrical, and that markets tend to overreact to decline in cash balances, but not to increases. They also found that diversion from equilibrium in either direction has no positive effect on future earnings.
] , cash hoards before the recession were possible due to availability of external financing, while after recession hoarding was fuelled by internally generated funds.
An interesting result was reported by Acharya, Davydenko and Strbulaev (2012) : when cash-to-assets mean and median ratios were plotted against bond ratings of firms in their sample, the highest levels of cash were observed both for the strongest (AAA rated) firms and near-default (CC, C-rated) firms. Mid-range rated firms had lower cash-to-assets ratios. They argue that companies with higher probabilities of distress will have increasingly conservative cash management policies. This conclusion partially supports results reported by Acharya, Almeida and Campello (2007). Bolton, Chen and Wang (2013) found that firms raise equity funds in good economic conditions even if they do not require funds, but by doing so they hedge themselves against potential market downturns. Higher probabilities assigned to market downturns induce firms to hoard cash. Also, firms may cut down on investments in good times in preparation for worsening market conditions. They were first to mention changes in market sentiment as a possible factor explaining managerial action leading to raising additional equity in favourable economic conditions. Even in cases when the cost of carrying cash is high, corporate savings may be explained by increase risk aversion and precautionary considerations. They also claim " As long as corporations put a non-zero probability on financial crisis occurring, the real effects of the financing shock would already be present before the realization of the shock . Any real response following the shock would merely be a residual response" (p.52).
The results of Marcum, Martin and Strickland (2013) were in line with those of Kim, Mauer and Sherman (1998). They argue that after the crisis smaller firms were taking defensive positions and paying off debt, large and profitable firms were more confident and maintained their debt levels and even reduced cash holdings. They state that shareholders of companies with high growth opportunities would place higher value on every dollar held in cash, compared to shareholders of firms with a low growth and low investment opportunities. Pinkowitz, Stultz and Williamson (2013) argued that profitable companies ( especially multinationals) holding abnormally large cash positions before the crisis can be defined as a "US-specific puzzle", but the increase in cash holding after the crisis is not. They implicitly address speculative arguments as defined by Whalen (1966) about higher investment opportunities but stopped short of calling them 'speculative'. They claim that besides lacking investment opportunities, they could not find any other explanation for the increase in cash holdings from the period before to after the crisis. In the follow up paper in 2016, same authors find that there is "no evidence that crosssectional determinants of cash holdings have changed from before the crisis to after the crisis" (p.344).
Sanchez and Yurdagul (2012) [ 16 Hartford, Klasa and Maxwell (2014) suggested an interrelationship between firms cash holdings and refinancing risk. They stated that companies with high levels of refinancing risk may value cash balances higher, as they assure that those companies will not face underinvestment problems, and that companies with shorter maturity debt would hold higher cash balances. Both findings implied the precautionary motive for holding cash. In line with Hartford et al. (2014) and Han and Qui (2007), Breuer, ] tested the argument that since taxes play a significant role in corporate decision to repatriate profits from subsidiaries, companies would rather hold cash reserves abroad and by doing so show high cash holdings on consolidated balance sheets. They concluded that the tax factor is not the reason for dramatic increase of cash position in U.S. corporations. Oler and Picconi (2014) looked at the effect of a shortfall or excess cash on a company's performance. As opposed to most models, their model is forward-looking but in essence is a variant of the concept introduced by Opler (1999) and Kim (1998), despite the authors' claim that it is an extended and improved model. Their findings show that diversions from optimal cash holdings in either direction will eventually lead to lower levels of accounting profitability (return on net operating assets), and those results are stronger for cash-constrained companies. They claim that investors incorrectly price companies in the current period, but will adjust their valuation once the effect of underor over-cash holding relative to the optimal cash levels will become apparent in the future.
More recently, Duchin, Gilbert, Harford and Hrdlicka (2017) reported that for the period between 2009 and 2012, about 40% of total corporate reserves where invested in risky assets and 35% of total reserves are illiquid. They claim that the COMPUSTAT based measure of cash is an inefficient measure of cash reserves and once investments in risky assets are included, the total reserves are much higher than defined by the cash to assets ratio. They find that because investors discount the value of investments in risky reserves, it is a sub-optimal asset management strategy, but they do not claim that this is an agency problem. They argue that large firms with significant cash reserves would invest in risky and illiquid assets. While stating that precautionary reserves are supposedly non-risky, and risky investments fall "outside of the traditional", they stop short of defining them as 'speculative'.
A study by Duchin et al (2017) highlights one of the fundamental problems in research on cash management -the inconsistency in the definition of corporate cash across various studies. They conclude that actual total cash reserves are 17% higher than reported on balance sheets. Amess, Banerji and Lamousis (2015) point to the difference in definitions ranging from the classic cash+marketable short-term securities that are liquid and mostly risk-free, to the adjustment made by Duchin et al (2017) that includes information from annual reports footnotes, to alternative short-term risky investments that cannot be classified as liquid or cash substitutes, as described by Cardella, Fairhurst and Klasa (2015). They also note that companies without an immediate need for cash would practice investments in higher risk, lower liquidity assets as a higher yield remedy for low risk, low return Treasuries and repos. Moreover, Carella et al (2015) conclude that managers do not just park excess cash in short term investments, as they do not treat them as cash but rather do so as the result of an investment decision.
A number of recent studies researched additional factors that influence the level of cash held by American firms, such as the differences in cash holdings between financially solid and financially constrained firms and unlisted firms, as well as the impact of tax regulations. Graham and Leary (2018) concluded that the increase in cash holdings can be explained by cash trapped overseas due to repatriation taxes. Tax avoidance strategies cause cash balances to become trapped overseas where tax rates are lower than in the United States. This finding is in line with the results reported by Harford, Wang and Zhang (2017) definition of permanently reinvested earnings. Rieger and Soypak (2017) addressed the issue of the increased importance of cash for financially constrained firms, and deepened the analysis by addressing the effect of investors' preferences on companies' cash policies. Their model shows the direct connection between investors' behavioral bias and cash management. Asimakopoulos, Asimakopoulos and Fernandes (2019) compared cash management practices between listed and unlisted European firms. The found that unlisted firms hold higher cash balances due to precautionary reasons, and found a U-shaped relationship in cash holdings between listed and unlisted firms. They also found that unlisted firms are more likely to save cash from their cash flows than listed firms. This result echoes US-based findings of Acharaya et al. (2012) that also reported a U-shaped relationship between cash holdings of high risk and low risk firms, where high risk firms hold cash for precautionary reason, while low risk firms conserve cash due to pecking order considerations. Faulkender, Hankins and Petersen (2019) looked at the effect of tax laws on cash accumulation by multinationals. They separated domestic and foreign subsidiaries cash holdings and noted that due to lower taxes abroad, subsidiaries is where those corporations park most of their cash, since this cash is taxed only when it is repatriated. They state that " domestic cash is explained by precautionary savings variables, while taxes explain foreign cash. There is very little evidence of precautionary motives explaining foreign cash holdings" (p.3301). They claim that it was never the intent of U.S. corporations to hoard cash abroad, but it was a consequence of the American tax system and as such has nothing in common with the traditional precautionary cash accumulation.
In a recent article Cunha and Pollet (2020) build their case study in cash management around demographic patterns that affect demand for certain goods ( like bicycles and pharmaceuticals). By analysing changes in demand driven by changing demographic factors they claim to be able to isolate the precautionary needs in those industries as far as 10 years in advance. According to the authors, firms will increase cash holding in anticipation of increased demand for their goods, and add that financially constrained firms start saving earlier and keep high cash level longer due to inferior market access and market frictions. When faced with higher cash needs in the future, financially constrained firms faced with limited or expensive external financing would retain higher proportions of earnings to meet future investment needs implied by forecasted higher demand, often as early as 4 years ahead of financially solid firms in the same industry.
Boot and Vladimirov (2019) researched cash management practices from a new and surprising angle : firms hoarding cash in order to avoid dilution, and by doing so they would delay current investments. This approach goes against the classic corporate finance theory about value maximization. They found that having access to private financing would cause the firm to hoard less, and that product market competition would have a significant effect on the choice of source of financing. They conclude that private placements not only do not dilute ownership but also are less transparent, yet they fall short of calling such actions and agency problem.

III. Theoretical background and motivation
The theory on the topic of cash management based on the initial definitions presented by Keynes (1934) was initially strictly on a philosophical level (Baumol (1952) and Beranek (1963), Whalen (1964,1965, Johnson (1961)) and was extended to a quantifiable model by Miller-Or (1966). Most of this body of work discussed the existence and the merits of motives for cash holdings in a general economic framework. While there was a general consensus on the existence of the transaction and precautionary reasons for holding cash, the line separating the precautionary and the speculative motives was blurred and poorly defined.
A breakthrough in empirical research came with the model introduced by Opler at al. (1999). Their model defined equilibrium levels of cash necessary for company operations, and de facto quantified the transaction motive for holding cash. A multiple regression model accounted for the drivers that determine cash flow needs for ongoing corporate operations, setting the stage for extensions that followed, and defined the drivers that explain the precautionary motive. The speculative motive was generally unaccounted for and, for most, excess cash held by corporations was attributed to the precautionary motive. To date, no empirical research concentrated or attempted to quantify and isolate the speculative motive for hoarding cash.
Since mid-1980s, the American non-financial sector displayed a steady trend of increasing cash reserves. As of the late 1990s, the American economy enjoyed an unusually prolonged upward sloping business cycle coupled with a similar trend in the stock market. This trend culminated and abruptly ended with the implosion of the housing market and a stock market collapse in 2008 followed by a global recession.
Leading up to the crisis, the American non-financial sector enjoyed both favorable access to financing sources and steady profitability, which translated into record high levels of cash holdings. This level of cash balances could be hardly explained by a combination of the transaction and precautionary reasons for holding cash and gave rise to suspicion of speculative activity fuelled by rising markets and sheer amounts of available liquidity. As extreme examples, Apple's cash balances exceeded the value of the American Treasury, and Google established an in-house trading floor stating it will increase operational efficiency of liquidity management and overly conservative investment strategies.
This period of extreme cash accumulation leading up to the 2008 crisis provides a unique opportunity for a case study on existence of the speculative motive, and is the core of this research. By quantifying the transaction and precautionary motives for holding cash, factors driving those two motives are accounted for and are removed from the data. Following Keynesian theory, any remaining unexplained variance, must be attributed to the speculative motive.
By creating an all-inclusive multi-factor regression model using variables used in prior research on the transaction and precautionary motives, and extracting the error term of this model, I obtain the basis for analysing the speculative motive in its purest form defined to date. In econometric terms, the error of a regression represents the unexplained variance of the dependent variable, and the null hypothesis #1 would imply that there is no market support for cash accumulation. Additionally, I define null hypothesis #2: Corporate CEOs were not able to foresee the upcoming financial crisis.
To assure statistical integrity of the model, potential problems with multicollinearity and time-dependency of the error term, are accounted for through appropriate statistical measures (adjusted R-sq. and the Durbin-Watson statistic, respectively). In my model, the dependent variable is the optimal cash balance for a firm, and the explanatory variables of the regression represent the combination of drivers of the transaction and the precautionary motives for holding cash, therefore the error term, in line with Keynes' definitions, would represent the speculative component of cash accumulation. This data is subsequently used to graphically show a clear trend of speculation-driven corporate actions leading up to the financial crisis and beyond.
Analysing this trend, I argue that the corporate managers expected the crisis significantly in advance of the event, but chose to ride the tide of speculative cash management to the last moment. Data also shows that with one exception, cash hoarding was blessed by the capital markets, and those results reject both null hypotheses as stated above. Both the concept and the methodology were never used before, and this is one of the major contributions of this research.
In addition, by analysing the plots produced by cumulative error terms of the regressions, I disambiguate the mystery associated with the definition of the border separating the precautionary and the speculative cash accumulation motives. This research proposes that the error term of the regression represents speculative activity in cash management. This is illustrated through the position of the points of the cumulative error terms relative to the vertical axis: the points positioned in the positive area should represent the speculative motive, while those in the negative area represent the excessive precautionary motive ("negative speculative"). The trend over time illustrates the shift from negative to speculative. In other words, the data shows the primary motive driving cash accumulation from the excessive precautionary to the speculative, as evidenced by the move from precautionary to speculative short-term investments, assuming the level of cash required to satisfy the transaction motive is fixed. This claim is in line with Duchin et al. (2017) and Cordella et al (2015) who claimed that there is a difference between the two types of short term investment assets in which companies Electronic copy available at: https://ssrn.com/abstract=3927550 park their cash reserves. The low-risk, low-return portion of those investments can indeed be attributed to the precautionary motive, even if it exceeds model-implied levels, while investments in high-risk, high-return short term assets are clearly driven by the speculative motive and plots of cumulative error term solidly places them in the positive area of the graphs.

III -a. "Market sentiment" -definition.
Another novelty this research contributes to the body of literature on cash management is the introduction of a metric to quantify market sentiment. The most widely used measure of market opinion regarding future corporate performance is the stock price. In this study I needed to create a metric that would directly measure market reactions to corporate cash hoarding. The new variable follows the logic behind the price-to-earnings ratio, which can be interpreted as the dollar amount investors are willing to pay for every dollar of earnings. The new variable is defined as price per share divided by per cent change in the cash position of a firm, and can be interpreted as the price investors are willing to pay for every per cent change in cash holdings. Should the per cent change in cash be positive, and the regression estimate be significant and positive, it can be interpreted as market endorsement of cash accumulation. Should the change in cash be positive and regression estimate be significant and negative, it would attest to market dissatisfaction with cash hoarding and by definition would imply the existence of an agency problem, where the managerial decision to accumulate cash is not in the best interests of the shareholders, and is not supported by the markets. In the case where the change in cash position is negative, similar logic applies. Considering that estimated coefficients are not sensitive to the sign of the variable in the original data, but rather measure correlation, a positive sign of the estimate would still imply market support, while a negative sign would attest to lack of support and point to an existing agency problem.
The actual value of the regression estimate for this variable holds additional information. By splitting the sample into two parts, before and after the crisis, and running the model on two sub-samples, I measure the market sentiment in these two distinct periods, and can therefore draw conclusions about whether support for cash accumulation increased or declined following the financial crisis.
Additionally, the introduction of filters as a robustness check allows me to evaluate the effect of market sentiment for profitable companies, with increasingly strong criteria to measure corporate profitability. These filters address a company's ability to pay steady uninterrupted dividends, whether a company is ranked in the top profitability quartile (as defined by return on assets (ROA)) in the sample, having consecutive 8 quarters of positive earnings, and finally, whether a firm meets all these conditions, i.e. all filters combined.
The sign and the value of the estimate for the price-to-cash change ratio allow for a new interpretation of the impact of the crisis on financially unconstrained firms. Due to their financial position, those firms were most likely to be able to hoard cash.

IV. The data and the sample
For the empirical tests I use two datasets, drawn from COMPUSTAT quarterly databank, which exclude financial (SIC 6000-6999) and utilities  companies in the US for the period 1997Q1-2019Q4. This data is further split into preand post-crisis periods, 1997-2007 and 2009-2019, respectively. For the purpose of testing robustness, data for the quarters corresponding to the height of the crisis will be excluded in order to better distinguish the pre-and post-crisis effects. Overall, the total sample included 106231 company-quarter data points. A vast majority of prior research was conducted using annual observations. In my opinion, higher frequency quarterly data is capable of better capturing the dynamics of the decision-making process by corporate CEOs with regards to cash management policies.
Considering the variables proposed below in the model section, additional data was obtained from the Federal Reserve website, namely, data on T-bills and bond yields.
Summary statistics for the entire sample are presented in Table 1 (1,99), and, in the case of leverage (similar to Pinkowitz et al. (2013)), restricted to the interval of (0,1). Table 1 represents descriptive statistics for the entire period 1997Q:1-2019Q:4, Table 2 shows descriptive statistics for same set of variables but for positive changes in cash only, and Table 3 reports the same set of variables for cases where the change in cash was negative. Data selection was restricted to positive values of total assets (ATQ), sales (SALEQ), cash and marketable securities (CHEQ) and equity (SEQQ). Net assets are defined as total assets (ATQ) minus cash and marketable securities (CHEQ), net working capital to net assets ratio is defined as current assets (ATQ) minus current liabilities (LCTQ) minus cash (CHEQ), divided by total assets ( ATQ) net of cash (CHEQ). Capital expenditures to net assets is calculated as one fourth of annual capital expenditure (CAPXY) to net assets (ATQ-CHEQ). This approximation was required due to lack of quarterly data for capital expenditures. Total debt to total assets is determined as the ratio of the sum of short term debt (LCTQ) and long-term debt (DLTTQ) to total assets (ATQ). Research and development to sales ratio is calculated as XRDQ/SALEQ. For the Market-to-Book ratio, market value is determined as a multiplication of shares of common stock This ratio is also used to calculate the industry sigma, which is the industry average of the standard deviation of cash flows to net assets. Each standard deviation is calculated based on 5-year quarterly data but the restriction was imposed, following Pinkowitz et al. (2013), that a minimum of 3 data points are required for calculation. The process of sigma calculation was a rolling (one quarter at a time) standard deviation for each company, averaged by 2 digit SIC code. From a technical standpoint, I re-defined 'Industry sigma' as the average volatility within each sub-industry in the sample.
However, instead of calculating it on 20 years of data, as in the work by Opler et al. (1999), I limit the focus to 5 years of quarterly data in order to isolate recent, rather than long term, volatility effects, similar to Pinkowitz et al (2013). This means that sigma was calculated on a block of 20 (quarters) observations, rolling one quarter at a time. The dividend dummy is defined as "1" if dividend was paid during a particular quarter, and "0" otherwise. Industry dummy is based on Fama and French (1997)  ], excluding the financial sector and the utilities sector that are also absent from the sample. The crisis dummy was set to "1" for each quarter between 2007Q:3 and 2009Q:2, and "0" otherwise.

V. Methodology
The methodology applied in this study will reflect the multitude of dimensions associated with both the explanatory variables and data treatment. The second, more technical separation of the sample, is temporal, distinguishing between the two time periods before and after the 2008 crisis. Both sample breakdowns described so far are needed to address the second research question regarding whether we can identify differences in rationales for hoarding cash before and after the crisis.
In the first stage the model was tested on an unrestricted sample, including the crisis dummy variable. Then the model was run on a sub-sample representing positive cash change only, followed by all periods run for all cases that displayed negative change in cash. All those regressions were run once without fixed effects and then repeated to include both industry and quarterly fixed effects.
In the next three steps, three different filters were applied, followed by the final step where all three filters were applied simultaneously. The filters were defined as dummy variables (flags) as following: (i) if no dividends were cut, flag#1 was set to "1" and "0" otherwise; (ii) if ROA was been positive for 8 consecutive quarters, flag#2 was set to "1" and "0" otherwise; (iii) if the company was in the top 25% of ROA in 10 consecutive quarters, flag#3 was set to "1" and "0" otherwise".
Finally, a sequence of regression runs were performed on all abovementioned cases twice again, to reflect the time periods before and after the financial crisis. The estimates of the market sentiment variable were extracted and reported in Table 7.

V-a. Model development and testing.
Before defining the actual structural model for testing, two terms from the research questions should be defined more precisely in the context of cash hoarding: conservatism and opportunism. The aforementioned "time varying uncertainty about economic conditions" are both represented by specifically defined variable and also built into the way the model is tested. The model will also include fixed effects controlling for industry, time and the crisis period.
Most previous studies used the natural log of cash-to-assets as the dependent variable and based on Opler et al. (1999). Bates et al (2009), used the following array of explanatory variables: market-to-book ratio, industry cash flow risk, R&D-to-sales, size, NWC-to-assets, CAPEX/Assets , leverage, a dividend payment dummy, a year dummy, and the lagged cash-to-net-assets ratio. Pinkowitz, Stultz and Williamson (2013) included the following additional variables: acquisitions, debt issuance, and equity issuance. Jankensgard and Andren (2014) ]:

VI. Results
The model was tested under a number of scenarios and with application of additional filters as a robustness check. Overall, for all variants and filters, the model produces highly significant statistical results. The goodness of fit tests are very strong, with F-stat values ranging from 140 to 3200, and adjusted R-sq. between 35 and 52.
While only the adjusted R-sq. were reported, as it is customary for multivariate regressions, the difference from the regular R-sq. was negligible, pointing to a lack of multicollinearity in the model, implying absence of interactive dependence between the explanatory variables. The model also does not display any problems with autocorrelation (lag dependency), as measured by the Durbin-Watson statistic.
[insert Table 4   Electronic copy available at: https://ssrn.com/abstract=3927550 When the sample was restricted to companies with positive changes in cash, as shown in columns (iii) and (iv) of  (2010) and Cleary (2013) which showed that portfolios based on cash-rich firms showed higher stock profitability as compared to portfolios comprised of cash-poor firms. Those studies, however, did not address the pre-and post-crisis differences. Tobin's Q, the measure of future earnings potential, is significant for the overall sample and the restricted sample, but only in regressions without fixed effects.
Addressing the financing activities of the firms in the sample, all types of financing activities show strongly significant results. Not surprisingly, the absolute values of the estimates for this sub-sample of cash-rich firms are higher than for the entire sample, probably because those firms had better and easier access to the capital market to raise capital, make acquisitions more easily, invest, repurchase shares and retire debt. Table 4.

An important but unsurprising result is reported in columns (v) and (vi) of
In the sub-sample that included firms with negative changes in cash positions, the estimate of the price-to-change in cash ratio is negative and highly significant. In its absolute value it is much larger than comparable estimates of the positive change in cash balances sub-sample as well. This result is definite proof to strong market objection to speculative cash management practices by cash poor-firms, and any speculative action by CEOs of such firms were not in the best interests of the shareholders. It seems like they were trying to take advantage of strong capital market trends to better their financial position, despite objection by the market, and by doing so, were clearly creating an agency problem. As expected, in this sub-sample the estimates of the drivers of financing activity are also highly significant, but in absolute values much lower than in the positive change sub-sample.
Overall, both the unrestrictive and the restricted model provide support to the notion of market -supported cash accumulation, thereby solving the "puzzle" defined by Pinkowitz at al. (2013). The puzzle was defined as unclear reasons for cash accumulation prior to the 2008 crisis. My results show that cash hoarding was driven by two interchanging motives (precautionary first, followed by speculative), under a support umbrella of the financial markets, but only for financially strong firms. For financially restricted companies market objection to speculative practices created an agency problem. All-in-all, significant estimates for the price-to-change in cash ratio imply that speculative cash accumulation is good, but only if you can really afford it, suggesting that financially constrained firms should stick to conservative cash management practices.

VII-a. Market approval
For the purpose of testing the model under increasingly strict measures of profitability, and thereby answering the question whether only profitable companies were involved in hoarding cash, three filters were applied on the restricted model. First, a requirement was imposed that in order to be included in the sample, a company should not cut dividends for 10 consecutive quarters. Second, only companies that belong to the top 25% of ROA across the sample were included, and third, only companies with 10 consecutive quarters of positive ROA were included. Finally, all three filters were applied simultaneously.
Results of the restricted model with the no dividends cut filter are reported in Table 5 columns (i) and (ii). I consider this to be the weakest filter of the three due to the fact that, as a rule, companies would try to maintain a steady dividends pattern and cut dividends with great reluctance, and thereby the payment of dividends does not postulate a strong measure of profitability and ability to hoard cash.
[insert Table 5 here] As can be seen from Table 5, companies that continuously paid dividends had a strong incentive to accumulate cash, expressed by highly significant estimates of the market sentiment variable. However, in their absolute values, those estimates are the lowest among the three filters used in this study. While common shares issuance estimates are significant, dividend-paying companies were also actively re-purchasing shares, as well as active in the bond market, making acquisitions and increasing their investments. This attests to high levels of available liquidity and financing despite using profits to pay steady dividends. This is consistent with favourable stock market conditions, yet contrary to pecking order hypothesis. Also, while the dividend-paying companies sub-sample did produce highly significant estimates for R&D-to-total assets, the absolute values are the lowest among the three filters used. In view of significant increase in investments, coupled with highly significant estimates for capital expenditure to total assets ratio, this may be interpreted as a trend of buying rather than developing among this sub-set of corporations. Both Tobin's Q and the Confidence Index estimates were significant potentially implying that steady dividend-paying firms were indeed considering future profitability prospects and market sentiment regarding macro economic trends in their decision-making process. Industry sigma, a measure of riskiness of the cash flows, is significant as well, despite the findings of Han and Qiu (2007) who also used quarterly data and showed that unconstrained firms display no association between sigma and the level of cash held by corporations. Estimates for the ratio of accounts receivables-to-sales are highly significant, i.e. dividend-paying companies were also active in the preservation of their clientele base.
[insert Table 6 here] Table 6 columns (iii) and (iv) report the results of the restricted model with the profitability filter including only companies that belong to the top 25% of ROA in each quarter. For this subset of data, the price-to-cash change estimates again produced significant results, and the absolute value of the estimates of price-to-cash change ratio are the highest among the three filters. This can be interpreted as a market blessing for cash hoarding, since the most profitable companies are most able to afford it. Since companies in this sub-sample were actively investing, involved in acquisitions, invested in R&D, retired debt and taking overall advantage of favorable market conditions, it seems like the strong market endorsement for hoarding and speculation was based on the fact that this type of company could afford to do so. This follows the logic of Google's strategy whereby large amounts of cash are considered to provide greater benefit to the shareholders if invested in higher risk, higher potential return securities.
Results also show that those companies were acting to protect their clients as reflected by the significant estimate for the ratio of accounts receivable to sales.
The next profitability filter identified and included only companies that had eight consecutive profitable quarters measured in terms of a positive return on assets ratio.
These results are reported in Table 6 columns (i) and (ii). This sample differs from the previous case that isolated only the top 25% of the profitable firms by greater stability of the sample that includes steady performers rather than the variable rankings of top performers.
The results for the price-to-cash change ratio are consistent with previous findings and are highly significant. Both Tobin's Q and the Confidence Index are significant in the no fixed effects runs, but not in the regressions that include fixed effects. This may be interpreted as relative caution in CEOs attitude, where future profitability prospects were playing a role after all, alongside opportunistic behaviour.
Also, despite the steady show of profitability, significant shares and debt issuance were coupled with acquisitions, increases in investments and long term debt reduction. This may be the result of favorable market conditions and some degree of conservatism on Electronic copy available at: https://ssrn.com/abstract=3927550 behalf of the financial managers. As expected, those companies also show significant estimates for R&D and increases in investments, while acting to protect their clients.
Finally, when applying all three filters simultaneously, I obtain some very interesting results, as reported in Table 6 columns (iii) and (iv). The sample size shrunk dramatically to 5,606 observations from a total of 85,411 observations in the original sample. Hypothetically, this sample should represent the select group of elite companies that both produced steady positive return, remain in the top 25% of ROA and paid continuous dividends.
The model displays highly significant results for the market sentiment variable as well as the drivers of the transaction motive, and selected drivers of the precautionary motive, including highly significant estimates for increases in investments and acquisitions, but with some notable exceptions. CAPEX to total assets and R&D to sales are surprisingly insignificant, and there is no clientele support. Neither Tobin's Q nor Confidence Index produced significant estimates, but absolute values of the market sentiment variable are the highest of all previous cases. The fact that estimates for the Confidence index and Tobin's Q are insignificant, may suggest that CEOs accumulated cash disregarding the fundamentals, and the indicator of macro economic conditions, and did so enjoying the strongest market support for their actions. This attitude changed after the crisis when market sentiment stopped playing a role in CEOs decision-making process, pointing to a shift towards even higher levels of opportunism in their behaviour. As can be seen from the Standard & Poor's chart mentioned earlier, as well as the second chart with updated data to 2020, during the post-crisis period cash accumulation continued to grow, and this elite group of companies where on the forefront of amassing mountains of cash, with Apple leading the pack. Table 7 reports extracts of estimates of the price-to-cash change ratio from 12 different regression runs of the full model, and supports the analysis of shift in market support from before the crisis to post-crisis period for each one of the previously reported cases.
[insert Table 7 here] At this stage, the entire sample and subsequently filter-based sub-samples were further divided into temporal sub-samples -before and after the crisis of 2008. The estimates of the market support variable are highly significant in all cases and draw a clear picture.
Market sentiment towards cash hoarding was stronger before the crisis and weakened after the crisis in all cases except when the dividend filter was applied. Interestingly, market disapproval for companies with negative cash changes also weakened. The general trend shown by these results may be explained by some degree of caution in the market after the deep crisis, however, as will be discussed below, speculative actions by financial managers ignored weakening market support and, for as long as the support existed, they significantly increased speculative activity even more so than before the crisis. This can be clearly seen in residuals plots discussed in the next section.

VII-b. Did CEOs know what was coming ?
One of the main contributions of this thesis is the evaluation of the role of the speculative motive for holding cash in the context of a financial crisis as a 'natural' negative shock experiment while using a different approach, variables and methodology compared to the existing body of literature.
The intent is to try to discover whether the investors and markets encouraged It was expected that the empirical results would show that some of the cash piles accumulated by American non-financial sector before the crisis may be attributed to the speculative behaviour by corporate CEOs, while in the post-crisis era cash hoarding may be explained by a precautionary motive. It was also expected that cash hoarding prior to the crisis was endorsed by financial markets, giving further support to corporate CEOs' cash management policies. While expectations regarding pre-crisis era were supported by the results, the expectations regarding increased post-crisis conservatism were rejected by the model that displayed an even higher level of speculation in the post crisis years. The emergent picture was even more complex, showing selective market support for cash hoarding and speculative market activity by financially strong companies, but not for companies with weak liquidity positions.
While the expectations regarding market endorsement of cash hoarding were supported by data, the test for change in behavior of CEOs pre-and post-2008 crisis produced a consistent and a surprising result. As the exhibits show, the speculative trend of cash accumulation was broken approximately 3 years before the crisis. The residual term from each regression reported so far, was assumed to represent cash accumulation attributed to speculative motive. Considering that both transaction and precautionary motives were accounted for in the model, the error term must represent what is left, i.e. the speculative motive for holding cash. In reality, I found that not all excessive cash hoarding was speculative after all, and some may be attributed to larger than implied by equilibrium conditions for precautionary accumulation, that eventually became speculative due to favorable market conditions.
The methodology used here mirrors that of Fama, Fisher, Jensen and Roll (1969). Across sample (or sub-sample) regression error terms were averaged and plotted against time, followed by a plot of cumulative error terms over time for each case discussed above.
Analysis of the residual plots (Exhibits 1 -7) produces a very clear and consistent picture, with the exception of the case where all three filters are combined. Prior to 2002 I observe "negative speculation" (i.e., excessive precautionary cash accumulation), followed by a rapid move towards speculative activity that peaked around 2005.
Speculative activity slows down towards the 2008 recession and consistently increases thereafter.
[insert Exhibits 1-7 here] As for the case when all three filters were applied simultaneously, the plot past

VIII. Summary
In conclusion, this paper's goal was to test two hypotheses in a case study framework. The massive cash accumulation by American non-financial firms that started in the early 1990s could no longer be justified by the reasoning of transaction and precautionary motives as defined by Keynes. The speculative motive could be expected to play a role in this cash hoarding, but it was never quantified and/or tested The results turned the initial expectations on their heads instead of displaying a change from opportunism to conservatism following the 2008 crisis, I found that the picture is more complex and a shift in the opposite direction first happened long before the crisis when companies moved from excessive conservatism to speculation. Starting in the late 1990s cash hoarding, according to the model was a testament to excessive precautionary cash accumulation, which soon turned into speculative behavior peaking in 2004. However, following the crisis financial managers did not change their practices back to conservatism as initially expected, but rather increased up their speculative practices while still enjoying market-based support.
The results showed high statistical significance of the model, and proved beyond reasonable doubt the existence of the speculative motive and its presence in the actions of American CEOs leading up to the crisis of 2008 and beyond. It was also shown that such actions were endorsed by the broad market. Also, results supported the findings of Duchin, Ozbas and Sensoy (2010) which showed that financially unconstrained firms did not reduce investments during the economic crisis. Finally, this study solved the puzzle defined by Pinkowitz et al (2013) and provided empirical explanation for the reasons for cash accumulation leading up to the financial crisis.
The table contains regressions of the firms natural LOG of the ratio of cash to net assets on a set of firm characteristics. Columns (i)+(ii) report the results of total sample with and without fixed effects, including only companies that reported 10 consecutive profitable quarters. Columns (iii)+(iv) report results with and without fixed effects for companies that both belonged to the top 25% profitability in terms of ROA, did not cut dividends and showed 10 consecutive quarters of profitability. (*** = Sig <.001, ** = Sig<.05, *=Sig<.10). This plot represents the behaviour of cumulative regression errors of the full sample. Each period error term is averaged across all companies in the sample and cumulative averages are plotted across the timeline. Horizontal (zero) line represents the border between excessive precautionary and speculative cash accumulations. The three vertical lines represent important dates: the beginning of the decline in speculative activity in anticipation of the 2008 crisis, the start and the end of the crisis.

_____________________________________________________________________________
Exhibit 5: Cumulative Average Residuals plot -top 25% ROA companies filter This plot represents the behaviour of cumulative regression errors of the sub-sample of firms that belong to top 25% performers in the sample, as measured by ROA. Each period error term is averaged across all companies in the sub-sample and cumulative averages are plotted across the timeline. Horizontal (zero) line represents the border between excessive precautionary and speculative cash accumulations. The three vertical lines represent important dates: the beginning of the decline in speculative activity in anticipation of the 2008 crisis, the start and the end of the crisis.

Variable Description and calculation
Ratio of price per share to per cent change of cash and marketable securities ratio of price per share (PRCCQ) to per cent change of cash and marketable securities {[CHEQ/LAG(CHEQ)-1]*100} Net assets total assets (ATQ) minus cash and marketable securities (CHEQ) Size (*) Natural log of Total Assets (ATQ) Net working capital to net assets(*) current assets (ATQ) minus current liabilities (LCTQ) minus cash (CHEQ), divided by total assets ( ATQ) net of cash (CHEQ) Capital expenditures to net assets (*) one fourth of annual capital expenditure (CAPXY) to net assets (ATQ-CHEQ). *This approximation was required due to lack of quarterly data for capital expenditures. Total debt to total assets (**) the ratio of the sum of short term debt (LCTQ) and long term debt (DLTTQ) to total assets (ATQ) Research and development to sales ratio (*) calculated as XRDQ/SALEQ Market-to-book ratio (*) market value is determined as a multiplication of shares of common stock outstanding (CSHOQ) by closing price per share ( PRCCQ), divided by the book value, calculated as the sum of common equity (SEQQ) and deferred taxes (TXDITCQ) minus preferred equity ( dividing the sum of net income before taxes (PIQ), depreciation (DPQ) and total interest (XINTQ), by net assets (ATQ-CHEQ) Industry sigma rolling industry average of the standard deviation of cash flows to net assets based on 5 years of quarterly date for each roll Confidence Index the spread between Moody's yield on seasoned corporate bonds -all industries, BAA and Moody's yield on seasoned corporate bonds -all industries, AAA Accounts Receivables -to -Sales ratio (*) Dividend dummy as "1" if dividend was paid during a quarter, and "0" otherwise Industry dummy based on Fama and French (1997) 12 industries classification excluding the financial sector and the utilities sector Crisis dummy set to "1" for each quarter between 2007Q:3 and 2009Q:2, and "0" otherwise (*) -variable winsorized within (1,99) interval; (**) restricted to (0,1) interval