The impact of ownership structure on stock liquidity: Nonlinear approaches

186
DOI: 10.22099/JAA.2020.32584.1836
 
Journal of Accounting Advances (JAA)
Journal homepage: www.jaa.shirazu.ac.ir/?lang=en
The Impact of Ownership Structure on Stock Liquidity:
Nonlinear approaches

ABSTRACT
Received: 2019-6-29
Accepted: 2020-6-25
  According to the literature, we can predict a positive or negative relationship between ownership structure and stock liquidity. Thus, present study aimed to document the asymmetric relationship between ownership structure and stock liquidity. Using a panel smooth transition regression model, as a new econometric technique, we examined the data (135 firms for 2009-2018) to explore the asymmetric impact of blockholders ownership on liquidity. In addition, the prediction error rate was compared on the base of neural network models and logistic regression. The results show that there is a positive and significant relation between major shareholders and stock liquidity in the first regime (threshold level 34%), while in the second regime, the relationship between them is negative and significant. Also, present study shows that neural networks’ mean-squared error (MSE) is lower than logistic regression.

1- Introduction
       Explanation of ownership structure and its effects on disclosure of information to reduce information asymmetry and subsequent effects on the enhancement of liquidity is of a great importance for investors. Review of numerous studies conducted on the relationship between major shareholders and liquidity implies an asymmetric relationship (Jacoby and Zheng, 2010; Cueto, 2009; Magu, 1998). Given this relationship, researchers focus on two hypotheses: adverse selection hypothesis and efficient monitoring hypothesis. The former says that when major shareholders have more information than other shareholders, information asymmetry arises and as a result, market liquidity is reduced (Kyle, 1985; Easly and O’Hara, 2004). On the contrary, the latter says that an institution rather wants to monitor the managers because of its risk consideration. The logic of this hypothesis is based on this idea that, due to high costs of monitoring, only major shareholders such as institutional ones can achieve sufficient advantages so they would have motives for monitoring. In fact, institutional and major shareholders have opportunities, resources, expertise, and abilities to monitor and influence managers (Cornet et al. 2007). Using the Panel Smooth Transition Regression Model, the current study aims to determine a threshold to investigate non-linear behaviors of major shareholders and stocks liquidity. Then, we will determine whether neural networks’ mean-squared error (MSE) is lower than logistic regression.
 
2- Hypothesis
According to efficient monitoring hypothesis, major shareholders actively manage their investments based on the amount of the invested capital. According to Transaction Cost Theory, an active company management leads to reduced transaction costs and consequently, , reduction of the difference between buy and sell prices of the stocks, subsequently increasing company stocks liquidity. On the contrary, an increase in major ownership indicates information asymmetry, as, with the presence of the major owners, a few informed shareholders can transact based on their information advantages. Concentrated ownership shows motives of a few shareholders to collect and analyze information and ultimately, transact based on it. This imposes adverse selection risk on other shareholders, thereby reducing shareholders’ motives to trade the stocks and lowering liquidity (Rubin, 2007). Our hypotheses is as follows:
        The asymmetric impact of blockholders ownership on firm performance follows the blockholders ownership level. And neural networks’ mean-squared error (MSE) is lower than logistic regression.
 
3- Methods
       We employed a panel smooth transition regression model, recently developed by Gonzalez et al. (2005), to model a nonlinear relationship between Institutional ownership, and firm performance. The simplest case of a PSTR model with two extreme regimes is defined as follows:
 
       The data derives mainly from audited financial statements and board's reports of the TSE, and Reheard Novin software. The population of the study encompasses all TSE firms for the period 2008–2017. However, the study compiles a purposive sampling; thus, financial firms such as banks and insurance firms are absent because they have different conditions in relation to firm characteristics. Listing firms must also have continuous operations during the period of the study, and their information must be available. Following these criteria, the study includes 135 firms (1350 firm-year). Then we will deterimine whether neural networks’ mean-squared error (MSE) is lower than logistic regression.
 
4- Results
Slope parameter, which indicates the speed of transiting from one regime to another, was estimated to be 2242.412, with the major shareholders’ percentage threshold being 34. A threshold is actually a landmark that makes the two mentioned regimes distinct in the panel smooth transition regression model. Depending on the estimated value of the slope parameter and the changeable transition values (percentages of major shareholders), the estimated model coefficients changed from one regime to  another. It should be noted that the first and second regimes were thresholds of the panel smooth transition regression model. In fact, given the observed transition variable, values of regression coefficients changed between the two thresholds. Also, present study shows that neural networks’ mean-squared error (MSE) is lower than logistic regression.
 
5- Conclusion
       Ownership percentages of the major shareholders had positive effects on liquidity in regime 1 and negative effects on performance in regime 2, such that a certain increase in the ownership percentages of major shareholders increases company liquidity and then, an increase in ownership percentages of major shareholders reduces company liquidity. Results of regime 2 were in line with Mendelson and Tunca (2004), Jacoby and Zheng (2010) Chung et al. (2018), Maharani et al. (2019), Rahmani et al. (2010) and Yaghubnajad et al. (2012). Moreover, results of regime 2 were in line with Jennings et al. (2002), Cueto (2009), Etemadi et al. (2011) and Mehrani and Nasiri Forouzi (2017). The results of the current study support the idea that major investors are not a homogeneous group and there are differences in their characteristics that produce an asymmetric effect on liquidity. Thus, it cannot be determined whether the major investors are good or bad for the market, because their effects are different from one regime to another. It seems to be dependent on ownership percentages of the major investors and company features. The results also show that legal institutions and information transparency have to be improved in the financial market to increase the efficiencies of the major shareholders and produce an increasing effect on stocks liquidity. Furthermore, legal institutions are recommended to develop regulations with respect to supporting the investors and of course, prior to it, financial information transparency in the financial markets. The current study also had limitations that were: (1) determination of 5 percent as the criterion for being a major shareholder, a change in which could affect the results; (2) we don’t classified institutional shareholder can lead to false conclusion; (3) inherent limitations of the use of panel smooth transition regression model; (4). A variety of neural networks methods are not seen. This can lead to an unreasonable conclusion.
 
Keywords: Adverse Selection Hypothesis, Neural Network Models, Nonliner Approaches, Ownership Structure, Stock Liquidity and Transaction Cost Theory.