**3. Literature Review and Hypothesis Development**

According to Keynesian theory, firms value liquidity for three reasons: first, the transaction motive, which facilitates the routine daily business transactions that meet its short-term obligations without facing the risk of insolvency; second, to provide for unforeseen expenditures and contingencies (as explained by the precaution motive); and third, to save opportunity costs associated with lost investments in the case of insufficient funds (a speculative motive), consistent with the pecking order and trade-off theories, otherwise known as the financing friction hypothesis.

Pecking order theory expects firms with expensive external financing and higher capital costs to depend on internal financing and to hold higher levels of cash, or viceversa [3]. In addition, trade-off theory consistently expects that firms with higher capital costs will accrue more benefits from holding higher levels of cash, i.e., their optimal level of cash is expected to be higher [27]. Second, managers might hold cash to exploit it for personal interests, which is what the free cash flow theory expects; they will hold higher levels of cash to increase the assets under their control and to reduce the need for external financing, hence leading to less market discipline [28].

Studies on cash holdings provide support for the above-mentioned reasons. For example, regarding the transaction motive, the objective of corporate cash holdings is to ensure the liquidity levels required to meet short-term obligations, thus avoiding the risk of insolvency and the cost of short-term borrowing. In the Jordanian context, this argument is consistent with the findings of Shubita and Shubita [4]. With regard to the precautionary motive, firms have been found to hold higher cash levels when faced with difficulties in raising external capital [27] and by higher cash-flow uncertainty [29]. Al-Amarneh [30] and Iskandrani et al. [31] found that firms in Jordan hold cash for precautionary reasons. In the same vein, McLean [32] found that firms faced with higher R&D expenses increase their cash holdings as a precaution. Firms in Jordan have been found to hold higher levels of cash when they have higher growth opportunities [33,34], consistent with the speculative motive. On the contrary, it has been found that firms hold more cash in case of higher agency problems at the firm and country levels [32,35].

Government ownership, as previously explained, can affect the level of cash holding through the predictions of Keynesian theory. Simply put, firms hold cash for speculation, precaution, and transaction motives. If the government provides different kinds of support, such as laxed taxes, direct finance of capital as part of its investment policy, or by guaranteeing firms will receive preferential loans [36,37], firms with government ownership will have easier access to cash, and will therefore have less motive to hold it. This is consistent with the expectations of the soft budget constraints theory [38], which expects that when a government owns shares in a firm, it will provide different kinds of support, such as laxed taxes and access to credit; moreover, in cases of financial distress, the government might intervene to save the firm [39]. If this holds true, government-owned firms will make fewer transactions and have fewer precautionary motives. As expected by pecking order and trade-off theories, such firms will therefore hold less cash.

On the other hand, agency theory predicts that government ownership will increase agency costs. Such costs are partly expected to rise because managers, under the pressure of government, might serve political interests rather than those of shareholders. Moreover, the managers in such firms are less subject to effective monitoring because, with the role played by government and politicians, fewer owners will engage in such action [40,41]. In fact, it has been found that state ownership increases agency problems because of ineffective monitoring [11,42,43]. Research on the effect of government ownership on cash holdings

has provided mixed results. Megginson et al. [44] found that it was negatively related to cash holdings in China, while Abramov et al. [45] demonstrated that government-owned firms increased their cash holdings to serve political interests. Chen et al. [14] argued that government ownership is positively related to cash holdings. Based on the mixed empirical and theoretical evidence, the following non-directional hypothesis is formulated:

**Hypothesis 1 (H1).** *Government ownership is significantly related to the level of cash holdings of Jordanian firms.*

#### **4. Data and Methodology**

#### *4.1. Data*

The study employs a panel dataset for 107 Jordanian listed companies in the service and industrial sectors covering the period 2009 to 2018. Financial data were gathered from the official website of the Amman Stock Exchange (ASE). The operational measures of the variables utilized in the analysis are shown in Table 1, and their summary statistics and correlation matrix can be seen in Table 2. The average value of cash holdings is 16.1%; compared to other developing countries, this is considered to be high. For example, Al-Najjar [46] reported an average CH of 5.6% for Brazil, Russia, India, and China over the period 2002 to 2008. For the period 2007 to 2012, Maheshwari and Rao [47] reported an average CH of 14.4% in India. This average is also high if compared to developed countries; for example, it was 5.9% for a sample of UK-listed firms [48] and 10.19% for Spanish firms [49].

The average value of the main independent variable, government ownership (Govt.), is 7.5%. This is considered low compared to other developing countries, for example, 24.35% in Vietnam [52] and 25% in China [14]. However, it is still high compared to firms in Kuwait, another Middle Eastern country, where government ownership is on average 3% [52,53]. Averages for other types of ownership, as in Table 1, are block-holders (Block) at 63.7%, Individuals (Indiv) at 48%, and Foreigners (Foreign) at 19.2%. Cash flow from operations (CFlow) is on average 6.1%, although some firms had negative operating cash flows. Debt issues (DbtIssues) stood at 15.2%, meaning that Jordanian firms increase their debt by 15% on average. SGr is 13.6%, indicating that, on average, firms increase sales by approximately 14%. Finally, CapEx is on average of 24.7%, meaning that firms increase their capital expenditure on average by 25%. The criterion of non-multicollinearity was confirmed, and there was no evidence of multicollinearity among the variables in the correlation coefficient matrix.

**Table 1.** Operational measures of the variables.



**Table 2.** Pair-wise correlation matrix and descriptive statistics.

\*, \*\*, and \*\*\* reflect significance at levels of 10%, 5% and 1%, respectively.

#### *4.2. Methodology*

According to Roodman [54], the generalized method of moments (GMM) is the most appropriate econometric estimator for dynamic model estimation. The system-GMM estimator is designed to accommodate a variety of data-generation assumptions and to deal with the dynamic generating process, which occurs when lagged dependent variables affect the dependent variable. It also manages the existence of unobserved heterogeneity and takes into account unobserved time-invariant effects. Third, the endogeneity issue caused by the explanatory factors is addressed by this methodology. Fourth, it is specially developed to deal with panels that have many individuals and few time periods (large N and small T), as well as to deal with the assumption that good instruments are available internally based on the lags of the instrumented variables and are not available outside the immediate dataset. Accordingly, to examine how government ownership affects the level of cash holdings (the aim of this research), the analysis is based on the following regression model:

$$\begin{array}{ll} \text{CH}\_{\text{i,t}} = \beta\_1 \text{CH}\_{\text{i,t-1}} & + \beta\_2 \text{SGr}\_{\text{i,t}} + \beta\_3 \text{Dbfsues}\_{\text{i,t}} + \beta\_4 \text{CFlow}\_{\text{i,t}} + \beta\_5 \text{Block}\_{\text{i,t}} \\ & + \beta\_6 \text{Indiv}\_{\text{i,t}} + \beta\_7 \text{Covt}\_{\text{i,t}} + \beta\_8 \text{Foreign}\_{\text{i,t}} + \beta\_9 \text{CapEx}\_{\text{i,t}} + \mathbf{f}\_{\text{i}} + \mathbf{d}\_{\text{t}} + \varepsilon\_{\text{i,t}} \end{array} \tag{1}$$

In order to control for corporate growth and investment demand, as discussed in the accelerator theory, SGr has been added to the model, together with DbtIssues and CFlow, which are used to control for trade-off and pecking order theories, and CapEx, which is used to control for capital expenditures. The model also controls for firm-fixed effects (fi) and year-fixed effects (dt). However, due to the lack of information in Jordanian firms' annual reports, we were unable to add equity issues and research and development expenditure to the list of predictors.

Due to the study model's dynamic structure and the endogeneity of its predictors, traditional least squares regressions produced somewhat inconsistent results. The association between the lagged dependent variable and the unobservable fixed effects, as well as the endogenous nature of the predictors, explain this inconsistency [55]. As a result, Arellano and Bond [56] introduced the differenced-GMM estimator and took the initial difference to solve this issue; nonetheless, this approach does not completely avoid the association between the disturbances and the lagged dependent variable. To solve the endogeneity problem, it is crucial to utilize instruments that are not correlated with the residuals, but with the explanatory factors. However, as Blundell and Bond [57] and Alkhataybeh [50] point out, in the presence of weak instruments, estimates of the difference-GMM are not totally reliable because estimations tend to be downward biased (According to Alkhataybeh [50], inconsistent difference-GMM estimates can be discovered if the coefficient of the lagged dependent variable falls between OLS (upward-biased) and fixed-effect (downward-biased) estimates, with being closer to the second).

Blundell and Bond [57] created the system GMM estimator, which involves a set of the moment conditions for the differenced equation as well as for the equation in level to improve the estimator. It is preferable to utilize one-step or two-step estimation while using it. Homoscedastic errors are assumed in the one-step estimator, while heteroscedastic ones are assumed in the two-step version. Flannery and Hankins [55] found that the two-step estimator was asymptotically more efficient in this setting, but that its standard error estimates were frequently biased downwards. As a result, the use of finite-sample standard error correction is encouraged. This study therefore considers the use of finite sample correction in the estimation of the two-step system-GMM. It should be emphasized that the instruments employed for the level equation in this study are the lagged difference and lagged level endogenous variables (dated t − 2 to t − 2) for the equation in difference.

#### **5. Results and Discussion**

The estimation results of the dynamic GMM model for the CH determinants are shown in Table 3. The lagged dependent variable (cash) has a positive and significantly different from zero coefficient, indicating that lagged cash levels positively influence current cash levels. Sgr, a control for growth opportunities, is almost zero. According to our results, sales growth does not have an impact on the level of CH. This is consistent with previous research [48,58,59] Therefore, growth opportunities do not play an important role in determining CH. This is inconsistent with the predictions of theories explaining CH levels. As mentioned previously, the financial system in Jordan is bank oriented, and government ownership also ensures preferential access to credit. Therefore, the opportunity costs of lost investments and growth opportunities are less relevant in Jordan.


**Table 3.** Estimation results of the dynamic cash holding model.

\*, \*\*, and \*\*\* indicate significance of the coefficients at levels of 10%, 5%, and 1%, respectively.

DbtIssues is found to be significantly and positively related to the level of *CH*, indicating that firms with new debt issues tend to hold more cash; this is in line with the findings of Maheshwari and Rao [47] and with trade-off theory. Such theory expects that firms with

higher debt levels will hold higher levels of cash because of the greater risk of bankruptcy. Companies with insolvency problems and facing the risk of bankruptcy tend to hold more cash [60] as a precaution. Therefore, such firms will raise more debt, if available, to increase their cash reserves.

CFlow has a positive significant effect on CH; in other words, firms tend to hold more cash when higher cash flows are in place, a finding that is consistent with free cash flow and pecking order theory. The positive result is consistent with the works of Ozkan and Ozkan [48] and Sher [51]. Capital expenditure (CapEx) is positively and statistically related to the levels of cash holdings, which is consistent with Jinkar [61] and Jebran et al. [62]. Supporting trade-off theory, this indicates that with higher capital expenditure in place, firms tend to hold higher levels of cash as a precaution in anticipation of investment frictions and time lags.

Finally, regarding our main variable, government ownership (Govt) was found to be significantly and positively associated with the levels of CH. Therefore, firms in Jordan with more shares owned by the government tend to hold higher levels of cash. The result is consistent with Gao et al. [63] and Chen et al. [14]. The results of the Arellano–Bond test of autocorrelation and Hansen's J-test for the validity of the used instruments confirm the consistency of the one-step and two-step system-GMM estimates.
