Microstructures, Financial Reforms and Stock Market Efficiency.

Vaalmikki Arjoona b

Chris Milnera

Spiros Bougheasa

Abstract

This paper analyses the effects of microstructures and fianncial reforms on time-varying efficiency in an emerging market setting. For this puspose, we use a multivariate panel regression framework. Our data comprises of firm level data from the Trinidad and Tobago Stock Exchange. Our empirical results show that microstructures have an important role in influencing the time-varying efficiency of this emerging market, while financial reforms do not. We also consider heterogeneity at the firm level, finding that the microstructures of the Banking stocks in this stock have a greater impact on market efficiency, in relation to the stocks of other sectors.

  1. GEP, School of Economics, University of Nottingham
  2. Corresponding author: Vaalmikki Arjoon, GEP, School of Economics, University of Nottingham, University Park Campus, Nottingham. NG7 2RD.
  1. Introduction.

An extensive literature has explored the informational efficiency of emerging equity market prices. The bulk of this literature investigates whether or not markets are efficient over a predetermined sample.[1] The conclusions of these studies are mixed, generally supporting the idea that emerging markets are inefficient. This suggests that prices in these markets do not reflect their full information (true) value and are not accurate signals for resource allocation. Such findings could arise as these markets are generally less developed and illiquid, with poor market structures, less sophisticated investors and inadequate regulatory standards. A major limitation of this group of studies is that they implicitly presume efficiency to be a static feature over the sample period. However, it is reasonable to expect that within the sample, there may be some temporary periods of efficiency, as microstructures change and the market becomes more mature. To this end, a growing body of literature considers the evolution of emerging market efficiency, rather than taking a snap shot of efficiency for a given sample period.[2] A natural extension to these analyses is to investigate what are those specific variables that influence the evolution of emerging market efficiency?

The present study is a continuation along the line of research on the evolution of emerging market efficiency, by adding to the literature the special focus of identifying the within-market variables that are driving the evolution of market efficiency overtime. In particular, we consider the effects that changes in several microstructure and financial reform variables have on the time-varying efficiency in an emerging market, namely the Trinidad and Tobago Stock Exchange (TTSE). It is integral to consider the effects of microstructures on time-varying efficiency, as microstructures influence the trading mechanism and the price formation process, which should in turn affect the transmission of information into equity prices. We also expore the implications for financial reforms as these alter the manner in which the market and investors behave overtime. These lines of enquiry not only establishes the role of microstructures and financial reforms in advancing emerging market efficiency, but also enhances our understanding of how information is transmitted into prices and the process by which markets become efficient overtime.

A number of studies have assessed the impact of microstructures and financial reforms on stock market efficiency. Among these studies, the majority focus on the effect of a single microstructure or reform variable. The microstructures commonly considered include the introduction of automated trading (for example Naidu and Rozeff (1994), Freund et al. (1997), Freund and Panago (2000) and Maghyereh (2005)) and the adoption of a call auction market design (for example...). Recently, Chordia et al. (2008) and Chung and Hradzil (2010) analysed the relation between efficiency and another key microstructure variable, liquidity. Most studies on the effects of financial reforms on efficiency focus on the effects of financial liberalisation (for example Kim and Singal (2000a, b), Groenewold and Ariff (1998), Kawakatsu and Morey (1999a, b), Laopodis (2004) and Cuajuero et al. (2009)). Other studies in this strand of the literature consider the impacts of regulation (for example Antoniou et al. (1997) and Fernandes and Ferreira (2008)).

The findings of several studies in the above strands of the literature are largely inconsistent. For instance, consider those studies that explore the effects of automation. These studies hypothesise that it leads to informational efficiency, as it fosters a rapid execution of trades with lower execution costs, causing prices to reflect a broader set of information at a faster rate. The results reported in Naidu and Rozeff (1994) confirm this hypothesis. Freund et al. (1997), Freund and Panago (2000) and Maghyereh (2005), however, find no evidence to suggest that automation stimulates efficiency. It can be argued that these inconsistent findings arise due to several limitations inherent in the literature, which puts into question the reliability of these findings.

Firstly, most of the literature assesses the imapact of only one microstructure or financial reform on efficiency, failing to consider that there may be other variables driving the eficiency/inefficiency of the exchange. For instance, Kim and Singal (2000) report that the Venezuelan stock market remains inefficient after financial liberalisation in 1990. This result, however, could be due to a lack of regulations to prevent unethical trading practices, which was only set up in 1998.[3] The effect of liberalisation may have had a positive effect on efficency, but the lack of regulation could have outweighed these effects. Therefore, it is fundamental to control for the effects of other microstructure or reform variables. Secondly, the empirical methodology used by the majority of studies is limiting. Specifically, these studies divide the overall sample based on the occurrence of a postulated microstructure or reform event. The efficiency of the market is then analysed in the pre and post-event sub-samples. If the pre-event sample is inefficient, while the post event sample is efficient, the study concludes that the event caused the market to move from an inefficient to efficient state. This method wrongly assumes that the event has an immediate effect on efficiency and any movement to efficiency is discrete. This effect, however, is likely to be gradual. For example, Cajueiro et al. (2009) point out that emerging markets take time to respond to liberalisation policies, as foreign investors may not enter the market immediately after liberalisation.

Our study is related to the recent works of Fernandes and Ferreira (2009) and Lagoarde Segot (2009), which provide evidence on the effects of certain financial reforms on emerging equity market efficiency. Fernandes and Ferreira (2009)

  1. Empirical Methodology

Lim et al. (2006) points out that there is no accepted theoretical model, which provides a clear explanation of the causal determinants of market efficiency. Consequently, most studies that investigate what influences efficiency achieve this empirically, by analysing the impact of a single exogenous factor on efficiency. This factor is usually an event, such as the implamentation of financial liberalisation policies, the introduction of automated trading on the exchange and financial crises. As is noted in Chapter 2, the conventional methodology, which most of these studies employ, divides the estimation sample on the basis of the occurrence of the postulated exogenous event. Tests of the random walk are then applied to both sub-samples to determine whether there is a difference in the efficiency of the market in the pre-event and post event sub-samples. If the pre-event sub-sample is found to be inconsistent with the random walk, while the post-event sub-sample adheres to the random walk, then these studies infer that the postulated exogenous event has a positive impact on efficiency and causes the market to move to an efficient state overtime. For example, Naidu and Rozeff (1994) use this approach to examine the effect of automated trading on the efficiency of the Singapore stock exchange. The authors divide the overall sample into pre and post-automation sub-samples, and apply serial correlation tests to both sub-samples. The results reveal lower serial correlation in the stock returns of the post-automation sub-sample, leading the authors to infer that the introduction of automated trading improved the market’s efficiency.

However, this traditional approach to assessing the influence of exogenous events on efficiency is limiting and can produce misleading results. To begin with, this framework assumes that the exogenous event has an immediate effect on informational efficiency. However, recent studies point out that in many cases, the effect is gradual. For instance, Kim and Shansuddin (2008), Cajueiro et al. (2009) and Hung (2009) show that the efficiency of emerging markets take time to respond to financial liberalisation and deregulation policies. Another major drawback with this conventional approach is that it only considers the effect of one variable on efficiency. It fails to consider that there may be other variables driving the efficiency of the exchange, apart from the exogenous event, such as microstructure and instututional features of the market.

The present study avoids these limitations involved with the traditional approach, by applying a regression framework to investigate the determinants of time-varying informational efficiency on the TTSE. Unlike the traditional approach, this framework enables us to detect and control for the influence of a multitude of variables and exogenous events on market efficiency. Moreover, the gradual effects of exogenous events on efficiency, such as financial liberalisation, are easily identified using this technique. The subsequent section outlines the regression framework and discusses the variables used to explain time-varying efficiency on the TTSE.

3.1. Model Specification

We assess the determinants of informational efficiency on the TTSE by applying two regression models: a baseline regression and an extension to the baseline specification.

3.1.1. Baseline specification

The bulk of the empirical stock market literature may be divided into two strands of studies: those that place emphasis on price efficiency (random walk behaviour of stock prices) and those that examine the link between microstructure factors and the level of stock prices/returns. We attempt to bridge these two strands, by estimating a baseline model, which examines the effect of key market microstructure variables on the efficiency of the TTSE. This model is motivated by the expectation that the institutional features and trading mechanisms of the market drive the process of information incorporation in security prices. In particular, the baseline model takes the form:

wherei indicates a particular stock and t refers to the time period. are the beta coefficient attached to the explanatory variables, where j = 1,...,n. Further, is a set of control variables, is an individual (stock) effect and is a year effect. The subsequent section provides a concise explanation of how we measure each of the variables and discusses the potential effects of each explanatory microstructure variable on informational efficiency. In the case of Liquidity and Volatility, we use multiple proxies. These proxies are applied interchangeably in the model to avoid any multicollinearity issues. The variables in the model are provided on an annual basis (156 trading days). Where data are available, we also consider semi-annual (78 trading days) data.

3.1.1.2. Description of the variables

3.1.1.2.1. Dependent Variable: Informational Efficiency (Efficiency)

The dependent variable denotes the degree to which stocks are weak-form efficient, and is denoted as Efficiency in. This variable is measured annually for each stock using the statistic of a random walk econometric test. Recall that a stock is weak-form efficient provided that its prices adhere to the random walk hypothesis (RWH) overtime. The statistic associated with a test of the RWH not only provides an indication as to whether or not a stock is weak-form efficient for a particular period, but also suggests the extent to which the stock is efficient. The magnitude of the test statistic can be interpreted as a relative indicator of efficiency: the higher the test statistic, the lower the probability of making a Type-1 error when rejecting the efficient market hypothesis (EMH). Thus, a higher statistic indicates a lower degree of stock return variation (randomness) and weak-form informational efficiency. This view is put forward by French and Roll (1986), who suggest that firm-specific return variation measures the degree of information incorporation into stock prices via trading. Therefore, a higher firm-specific return variation, as implied by a lower random walk statistic, indicates that the stock is tracking its fundamental value more closely.

In the present study, we apply the non-parametric signed rank test of the random walk developed by Luger (2003) to the returns of each listed stock on a semi-annual and annual basis. The resultant statistic for each year is used as the indicator of efficiency in the regression given in. Recall that this test is applied in Chapter 2 in a fixed length moving sub-sample framework, to capture the evolution of the efficiency of each TTSE stock (refer to Section 4.2.2.4 in Chapter 2 for a description of this test). We retain the use of this test to measure time-varying efficency in the present study, since it is robust to outliers and many forms of heteroscedasticity frequently found in financial time series. Moreover, the non-parametric signed-rank test also performs well in most sample sizes, including small samples, unlike other modern-day tests of the randomn walk. This is revealed by Luger (2003), who uses monte carlo simulations to show that in sample sizes of 100 and 200, the non-parametric signed-rank test is significantly more powerful than the rank and sign tests of Cambell and Dufour (1997) and the contemporary variance ratio tests of Wright (2000).[4] As we require a measure of efficiency for smaller samples in this study, that is, semi-annual and annual periods, we maintain the use of the non-parametric signed-rank test as the proxy for informational efficiency of each TTSE stock.

3.1.1.2.2. Exogenous (microstructure) variables

A. Liquidity (Liquidity)

Liquidity is widely regarded as the most important property for well-functioning and efficient stock markets (O’Hara, 1995). A highly liquid market is one in which securities can be traded quickly, such that there can be alarge number of transactions,without incurring significant transaction costs (also known as trading or execution costs). We consider these aspects of liquidity, namely the number of transactions and transaction costs, in the present section.

Increased stock market transactions signal improved liquidity. This promotes efficiency, as it causes stock prices to move more frequently, and induces price variability. Prices are therefore likely to conform to the random walk, eliminating price predictability and improving market efficiency. Moreover, increased investor trading brings a greater amount of information (based on fundamentals) in the market, which is reflected in market prices. This further serves to stimulate informational efficiency on the stock exchange.

Kyle (1985) points out that another key driving force of stock market liquidity is the extent of transaction costs, measured by the price spread. Indeed, this view is also shared by Amihud and Mendelson (1986), O’Hara (1995) and Lesmond (2005), who note that liquidity can be measured by the cost of immediate execution, with higher transaction costs lowering the liquidity and efficiency of the market. Liquid markets with little trading costs attract investor activity. This enhances trading frequency. On the other hand, high trading costs tend to impede market efficiency, as it discourages investor participation and restricts the ability of emerging markets to mobilise investment resources. This is documented by Chuhan (1992), who notes that poor liquidity, prompted by high execution costs, is one of the main obstacles to foreign investment in emerging markets. In addition, a number of studies find that excessive transaction costs tend to make investments more risky, as traders are less inclined to acquire and sell assets rapidly. This causes investors to demand a higher liquidity premium as compensation for holding illiquid securities (see Amihud and Mendelson, 1989; Eleswarapu, 1997; and Chalmers and Kadalec, 1998).

O’Hara (1995), however, argues that there are potential negative aspects of market liquidity. In particular, while liquidity may benefit the individual investor, it may impose costs on the market by encouraging the flight of investors. This can lead to regulatory and stability problems for emerging securities markets, thereby hindering their efficiency.

Undoubtedly, liquidity plays an essential role in the efficiency of stock markets. However, it is an elusive concept that is difficult to measure (see Kyle, 1985; O’Hara 1995; Lesmond, 2005; Chan et al., 2009). This may be attributed to the lack or unavailability of data pertaining to the number of trades executed and transaction costs in stock markets, the key indicators of liquidity. As such, a menu of procedures,which proxy for these indicators is well documented in the academic literature. In practice, these procedures are applied in a plethora of empirical investigations in the dynamics of market liquidity, including studies that examine the implications of liquidity for security prices and returns and the influence of other microstructure variables on liquidity.

The subsequent sections provide a description of liquidity measures applied in the present study, namely the number of transactions, stock turnover and transaction costs.

A.1. Measuring liquidity

A.1.1. Number of transactions

The first liquidity variable used in the present study, the number of transactions, provides an indication of the number of times each TTSE stock is traded annually. This variable is regarded as the most direct measure of liquidity. Its application in this study is novel, as to the best of our knowledge, the number of transactions is not previously employed in the literature as an indicator of liquidity. We hypothesise that this variable promotes efficiency on the TTSE, as an increase in the number of transactions stimulates price movement, variability and improves the speed of price adjustment to market information. This suggests that the prices of stocks which are actively traded react to information on a timely basis, thereby promoting informational efficiency. In this regard, we expect that the coefficient associated with this variable has a negative sign in the empirical regressions. Note that the data for this variable is only available on an annual basis.