Abstract
Purpose
The author investigates whether investors’ online information demand measured by Google search query and the changes in the numbers of Wikipedia page view can explain and predict stock return, trading volume and volatility dynamics of companies listed on the Nigerian Stock Exchange.
Design/methodology/approach
The multiple regression model which encompasses both the univariate and multivariate regression framework was employed as the research methodology. As part of our pre-analysis, we test for multicollinearity and applied the Wu/Hausman specification test to detect whether endogeneity exist in the regression model.
Findings
We provide novel and robust evidence that Google searches neither explain the contemporaneous nor predict stock return, trading volume and volatility dynamics. Similarly, results also indicate that trading volume and volatility dynamics have no relationship with changes in the numbers of Wikipedia pages view related to stock activities.
Originality/value
This study opens new strand of empirical literature of “investors' attention” in the context of African stock markets as empirical evidence. No evidence from previous studies on investors' attention exist, whether in Google search query or Wikipedia page view, with respect to African stock markets, particularly the Nigerian stock market. This study seeks to bridge these knowledge gaps by examining these relations.
Keywords
Citation
Osabuohien-Irabor, O. (2021), "Investors’ attention: does it impact the Nigerian stock market activities?", Journal of Economics and Development, Vol. 23 No. 1, pp. 59-76. https://doi.org/10.1108/JED-02-2020-0015
Publisher
:Emerald Publishing Limited
Copyright © 2020, Osarumwense Osabuohien-Irabor
License
Published in Journal of Economics and Development. Published by Emerald Publishing Limited. This article is published under the Creative Commons Attribution (CC BY 4.0) license. Anyone may reproduce, distribute, translate and create derivative works of this article (for both commercial and non-commercial purposes), subject to full attribution to the original publication and authors. The full terms of this license may be seen at http://creativecommons.org/licences/by/4.0/legalcode
1. Introduction
Attention is a key aspect of our visual experience closely relates to perception (Zagoruyko and Komodakis, 2019). And as humans, we need to pay attention in order to adequately perceive our surroundings. Attention is crucial but difficult to measure directly, hence previous studies could not provide direct measurement. For example; Takeda and Yamazaki (2006), Kim and Meschke (2011) used the appearances on TV programs as a proxy for investors’ attention. The paper by Antweiler and Frank (2004) reports that Internet stock messages posted on Yahoo! Finance and raging bull can help predict stock market volatility, and they found that posted stock messages impact stock returns significantly. Fehle et al. (2005) used product marketing advertisements as a surrogate for investors’ attention. Their results suggest that firms that spend more on advertising, attracts a larger number of both individual and institutional investors. Fang and Peress (2009) examined the relationship between media coverage (e.g. newspaper) and the stock returns. Their analysis shows that those with high media coverage have higher returns. But what is the guarantee that these investors pay attention to such advertisements and news media? This is one of the limitation studies using the traditional indirect measure for attention suffers, hence, the need for a more direct proxy to measure investors’ attention.
The popularity of Google searches among Internet users to collect information is one of the major reasons Google aggregate search frequency is use as a direct measure of attention. And as of 2019, Google receives 75,000 searches per second on any given day and 2 trillion searches per year [1]. This makes Google searches a direct proxy for attention much more valued than the earlier indirect measures. Hence, volumes of research literatures using Google searches, particularly studies related to financial markets have continue to increase. Since becoming public in 2006, Google Trends (GT) service has allowed users to see and use the aggregate search volume for any defined term or topic. Google search volume (GSV) can be used to examine the relationship between investors’ attention and the financial market, particularly the stock market activities. Search word(s) undoubtedly shows attention, and the aggregate of search frequency is a direct measure of active attention (Da et al., 2011). In other words, quest for information demand on a specific topic implies investors’ attention and time. Since the emergence of the famous seminal paper of Da et al. (2011), large volumes of empirical literature on Google search intensity as a measure of attention demands have been spotted in many branches of knowledge. However, despite the wide usage and popularity of Google search query in relation to stock market activities, no previous study exists with evidence from African stock markets. And to the best of our knowledge, most of the existing literature primarily focused on evidence from the emerged markets and developed economies. This contrasts with the dearth of literatures with evidence from African stock markets.
Based on empirical evidence, we categorize studies of investors’ attention and stock market activities in two strands − the US stock market related studies and the non-US stock market related studies. Dimpfl and Jank (2016) investigate the dynamics of Dow Jones Industrial Average volatility and retail investors' attention. Their empirical results show that there exist a strong co-movement between the realized volatility and investors’ attention. Huang et al. (2019) study indicates that the directional movement of S&P 500 stock index due to changes in search volume is dependent on the specific term being searched for and the sentiment itself. Their study also validates GT as a measure of investors’ attention. Using eighteen (18) different US companies from New York Stock Exchange (NYSE), Nasdaq Stock Market and the Dow Jones Industrial Average Index, Audrino et al. (2019) show that investors’ attention and sentiment variables improve volatility forecasts significantly. Besides, their results also identify the investors’ attention variables as the most relevant measure of active attention in stock market. Da et al. (2011) explore the US Russel 3,000 data, and their results shows that increase in Google searches predict a higher stock price.
Compare to avalanche of empirical evidence from US studies, not many studies have examined the relationship between investors’ search intensity and stock market activities for non-US stock markets. Dimpfl and Kleiman (2019) empirical studies investigate the relationship between online search intensity and the German stock market. Their results show that an increase in retail investors’ pessimism leads to an increase in volatility and trading volume but a decrease in stock return. Kim et al. (2019), examine the relationship between Norwegian stock market and GSV. Their results show no relationship between investors’ attention and stock returns. The study by Aouadi et al. (2013) reveals that the investors’ attention is a significant determinant of French stock market illiquidity and volatility, which is strongly correlated with trading volume. The papers by Takeda and Wakao (2014) and Adachi et al. (2017) also investigate the Japanese stock market. Whilst the former empirical results show online search intensity is weakly related to stock returns but strongly related to trading volume, the latter results indicate that the search intensity is positively related stock returns and liquidity. However, evidence from the Chinese stock market using Google search query as proxy for attention is quite lows. This is because most Chinese academia and investors prefer the use the Baidu index as attention proxy. That notwithstanding, Zhang et al. (2018) examined the impact of firms’ sustainability on stock return and volatility. They find evidence that firms’ sustainability news decreases volatility persistence and an increase in Google search engine.
Records of data from the history of viewership of Wikipedia pages (hereafter referred to as, Wiki) is another source of information demand use as proxy to measure investors’ attention. Studies on the relationship between the numbers of Wiki page views and stock markets activities have not attracted much attention like the Google search query. Moat et al. (2013) seminal paper first commenced investigation to check if changes in financially related Wiki page views impact the stock market before trading decisions are taken. Their results suggest that changes in Wiki provide information for investors to make decision before stock trading. Herve et al. (2019) used the Google search query and Wiki page views’ traffic respectively as a proxy for noisy traders and smart investors’ attentions to investigate the stock market activities. Their study indicates that only noise traders' attention impacts stock returns and increases volatility. However, activities of smart investors' attention show to decrease volatility. Cergol and Omladič (2015) employed the Google search query and Wiki page views to investigate substantial sample of firms listed on S&P 500 index. The Wiki records showed an inverse pattern of merry frown for the bear market and sour smile for the bull market in stock prices. The paper by Yoshida et al. (2015) indicates that frequently searched keywords are highly correlated with changes in the numbers of Wiki page views. Their results suggest that the numbers of Wiki page views can be used as an effective tool to determine popular web search trends.
But do investors search intensity using Google search query and Wiki page views impact the Nigerian stock market? Does the dearth of research literature on investors’ attention and stock market activities in African stock markets connotes absence of interrelationship? These are some major question among Nigerian stock market investors, stockbrokers, policy makers and researchers interested in the Nigerian stock market. The Nigerian stock market was established in 1961 as the Lagos Stock Exchange and later renamed in 1977 as the Nigerian Stock Exchange with 19 securities listed for trading. As at November 2019 the Nigerian Stock Exchange has a total of number of 161 listed companies; 8 domestic companies on premium board, 144 companies on main board and 9 companies on alternative securities market board and a total market capitalization of over ₦28 trillion as at January 9, 2019. Nigeria remains Africa most populous country with a GDP of $376.284 billion in 2017(IMF) and the largest economy in Africa [2]. And as at July 2019, Nigeria’s Internet penetration reached 61.4 percent with 122 million active users which accounts for 23 percent of Internet users in Africa [3, 4]. The Nigerian stock market investors are made up of both the domestic investors and foreign investors, with the domestic investors sub divided into retail and institutional investors. With these characteristics, coupled with being the market that powers the growth of the largest economy in Africa, Nigerian stock market is an ideal context to apply and test the effects of online search intensity using GSV and Wiki as proxy for investors’ attention.
This paper pioneers a new strand of literatures on the relations between investors’ attention and the stock market activities in Africa context. Specifically, we examine whether GSV and changes in the numbers of Wiki page views can explain and/or predict the stock returns, volatility and trading volume in the Nigerian stock market. We find that Google searches neither correlate with contemporaneous nor predict stock return, trading volume and volatility dynamics. And changes in the numbers of Wiki page view records does not influence the trading volume and volatility dynamics of the Nigerian stock market. However, our results indicate that changes in the numbers of Wiki page view related to stock negatively explain the variations in the stock returns.
The study consists of six sections, and the rest is organized as follows. Sections 2 and 3 explain our data and methodology, respectively, and Section 4 discusses the empirical results and robustness of our methodology. The concluding remarks are provided in Section 5.
2. Data
2.1 Google search volume
This study examines the GSV as a direct proxy for investors’ attention to examine the Nigerian stock market trading activities such as stock returns, volumes and volatilities using the NGSE-30 index. The normalized weekly values of GSV from numbers of keywords from August 3, 2015 to August 2, 2019 are obtain, through GT. Normalization removes possible effect of the overall increase in the number of searches. Google search values range from 0 to 100. Whilst the value 0 means that the search volume is too low to provide meaningful statistics and does not literarily indicates no search, the value 100 represents the highest level of search activity during the sample specified period. GT calculates the GSV from a random subset of data. Thus, same data downloaded at different times, have slightly different results. To avoid this problem, we follow the paper by Da et al. (2011), by downloading all the GSV index data once and no data change thereafter. Downloading GSVs values several times, results in most correlations to exceed the value of 97%. We focus on firms listed on the NGSE-30 index on the floor of the Nigerian Stock Exchange as at June 2019. GT reports data from Sunday to Saturday. The raw downloaded GSV cannot be applied directly in analysis, because its value depends on the time period of downloaded data, hence, the need for standardization of GSV data relative to its past history.
2.2 Abnormal GSV
This study follows the standardization formula by Bijl et al. (2016) and Kim et al. (2019), where the average of the past 52 weeks is subtracted from the weekly raw
Note: The Wiki data are downloaded from the English language Wiki web, launched on the 15th of January 2001 and available on https://en.wikipedia.org. The data are converted to weekly data and transformed using Eqn (1) framework. Firm’s Wiki page views can be downloaded daily (both in weekdays and weekend) and months.
2.3 Selection of keywords
Previous papers, such as Da et al. (2011), etc., employed the ticker symbols of stocks instead of the firm name, based on the following reasons. First, user searching for firm’s name may seek irrelevant information to stock investments purposes. Second, a company search name might be ambiguous with several meanings (e.g. “Apple”). Third, GT input does not allow searches for alpha-arithmetic term. Hence, company’s names such as 3M, 3com, etc. cannot be search for. Conversely, several literature studies, such as; Kim et al. (2019), Bijl et al. (2016), Aouadi et al. (2013) and Bank et al. (2011); have shown preference to the use of firm’s name than tickers names. This is because, firm’s name searches have a stronger relationship to stock market returns than ticker searches. But sometimes, firms name might not return any raw data due to insufficient GSV but have sufficient GSV for firm’s tickers.
To this end, this study uses aggregates GSV from both the firm’s names and its associated tickers names. First, we search for the individual firm’s names to get the GSV. And if the use of firm’s name does not return GSV, then we search using the tickers names. And if ticker searches also fail to return any aggregate GSV, such firms are deleted. However, problems identify by Da et al. (2011) particularly the creation of random noise arising from the use of firms name not relevant to investments purposes, are address with the following procedure;
We exclude terms that identify a firm’s legal status, (e.g. “Holdings”, “Group”, “LLP”, “INC”, “limited”, “LTD”, “international”, etc.) from the firm’s name, unless excluding these terms make the company name too general
We check and compare firms name as well as their associated ticker, stated in Wikipedia pages (https://en.wikipedia.org).
We exclude firm’s names and tickers that are found to be ambiguous or associated with searches other than firm’s financial information.
Previous research by Preis et al. (2013), indicates that data filtered according to geographic location or proximity can better explain movements in the specific geographic location. Hence, GT default filter is set to “All Categories” of “web search” conducted in Nigeria (locations) for the period of four (4) years, between August 3, 2015 and August 2, 2019.
2.4 Sample selections and elimination
This section reports sample selection and elimination of data used in this paper. The entire sample consist of 30 stocks trading firms on NGSE-30 index. Firm’s trading stock data as well as their associated Wiki page views data are obtained, with procedures outline in Section 2.3 as precept. However, firm’s corresponding tickers searches listed in Appendix 1Table A1, Panel A and B, either returned insufficient GSV or have ambiguous keywords. Hence, our study used the firm’s GSV data. After the selection and elimination process, a total number of twenty-four (24) firms are finally applied in our analysis. Table A1, panel B, shown the complete list of companies with no Wiki or Google search data. Thus, these companies were excluded in our final sample selection.
2.5 Stock data
We downloaded a total of 24,624 daily adjusted closed prices data. And used the weekly (week ending Friday) historical closing stock prices data between August 3, 2015 and August 2, 2019 obtained from the Nigerian Stock Exchange, Lagos. The NSE-30 index is a modified market capitalization index with a fixed number of thirty (30) included stocks. All firms listed on NGSE-30 as at June 1, 2019 were searched for data analysis, but firms which did not meet Section 3.3 and 3.4 selection criterion are expunge from our sample. Besides, missing data due to holidays and non-trading days are replaced by simply using the mean imputation procedures (Batista and Monard, 2003). The weekly return is calculated from the adjusted closing prices as;
2.6 Abnormal return
Following capital asset pricing model (CAPM) criticisms of unrealistic assumptions and the non-changing time frame of investment appraisals, Fama and French (1993) developed the 3-factor asset pricing model which covers some of the deficiency of CAPM. Again, Fama and French (2015) developed additional 2-factors (the profitability and investment factors), which bring the total factors to 5-factors. We calculate our abnormal returns as values obtain from the subtraction of stock benchmark (expected) from the actual stock returns shown in Eqn (2). In this paper, the Fama and French (2015) model shown in Eqn (3) generates our returns benchmark. And the daily values of the abnormal returns are converted to weekly data by compounding the returns from Monday to Monday as shown in Eqn (4), thus;
2.7 Abnormal trading volume
In the spirit of Chesney et al. (2015), we describe the abnormal trading volume as an unusual trade in option contracts which brings about large gains, usually made few days before the occurrence of a major event. Thus, to examine the relationship between the
2.8 Volatility
Financial market volatility as well as stock returns are variables of significant interest to the economists, investors and other market participants (Chung and Chuwonganant, 2018). This is because it measures the levels of risk associated with investments. For this reason, we include volatility as one of the control variables in our regression models that explains the stock returns and the trading volume. We apply the Parkinson (1980) range-based volatility estimator, which uses the high
Ct = Closing price on day t; Ot = Opening price on day t;
Ht = High price on day t; Lt = Low price on day t;
n = number of daily periods.
The weekly volatility as a square root of average daily variance is calculated thus;
2.9 Summary statistics
Table 1 presents present two parts; the panel A which is the summary statistics and Panel B which contains correlation analysis of
3. Methodology
Our methodological framework is the multiple regression analysis. Our results also reject the null hypothesis of the existence of a unit root in our variables when we employed the augmented Dickey–Fuller (ADF) test with deterministic constant. This justifies our model estimations in OLS regressions. Our pre-analysis result also reveals the absence of multicollinearity as shown in Table 1 panel B. In addition, we conduct the Wu/Hausman specification test for endogeneity which reject the null hypothesis that our endogenous variables are actually exogenous variable (for brevity, table not shown). We use the multivariate regression in Eqns (8)–(10) to study the relationship between investors’ attention and the stock market activities for NGSE-30 index. The
Our regression models are expressed in both contemporary explanatory and the predictive models. In Eqns (8) and (8′), we examine whether the current and lagged values of
Similarly, the trading volume is regressed on of
Eqns (10) and (10′) investigate whether there is a contemporary and predictive relationship between
4. Results of empirical analysis
The major objective of this study is to investigate the impact of investors’ online information query on large firms listed on the Nigerian stock exchange. Research gap on “investors’ attention” are identified and filled by providing empirical evidence from the Nigerian stock market. To evaluate the impact of investors’ online search intensity (using
Tables 2–4, columns (1)–(11) shows the coefficient estimations and the robust standard errors (in square bracket) of results from our analysis. The table displays results from both the descriptive and predictive model in univariate and multivariate form. The contemporaneous weekly abnormal stock returns are significantly positively correlated with its lag values due to the autoregressive nature of
Results from Eqns (9) and (9′) shown in Table A1, investigate the impact of search queries with the abnormal trading volume
4.1 Robustness check
In this section, we examine the adequacy of our models by performing robustness check. We used five lags predictive model to compare our main results in Tables 2–4. We apply the Bijl et al. (2016) standardization in our main model and Da et al. (2011) standardization for the new model. The models result from both standardizations are compared. Results from both models are expected to be the same.
For the alternative model, we used the detrended volume,
We also compare the results of trading volume for both SM (A) and (B), shown on Appendix 1, Table A2. The
5. Conclusion
In this paper, we examined empirically whether investors’ online search intensity measured by GSV and Wikipedia page views influences the stock return, trading volume and the volatility dynamics of the Nigerian stock market. Our analysis covers the top thirty (30) firms in Nigeria stock market exchange with respect to their market capitalization and liquidity. Our analyzed empirical results suggest that the use of data information offered by Wikipedia and Google services can be of great benefit to active and prospective investors as well as individual companies. Results show that the GSV does not influence firms listed on NGSE-30 index. Nevertheless, changes in the numbers of Wikipedia page views related to stock market activities negatively impact both the contemporary explanatory and predictive stock returns. Negative significant influence of attention on stock returns implies that, higher investors attentiveness by viewing firm’s Wikipedia pages related to stocks market might lead to lower stock returns. This suggest that investors of NGSE-30 index companies appear to be more sensitive to bad news than good news in their investment decision. In summary, this article suggests that Google query for searching stock related information is not a driving factor for price movement for companies in NGSE-30 and appears not to interact with return, trading volume and volatility when determining price dynamics.
This table presents the summary statistics and the correlation matrix for the sample data as shown in Panel A and B respectively, for 24 firms from the NGSE-30 index
Panel A: Summary statistics | |||||||
---|---|---|---|---|---|---|---|
Variables | Sample sizes | Mean | St. Dev | Kurtosis (excess) | Skewness | Min | Max |
AGSV | 209 | 0.000 | 1.000 | 2.516 | 1.230 | −2.107 | 4.155 |
AWiki | 209 | −0.354 | 1.000 | 0.956 | 0.855 | −2.498 | 3.115 |
ARET | 244 | −0.016 | 0.023 | −0.112 | 0.608 | −0.066 | 0.048 |
ATVOL | 293 | −0.237 | 0.995 | 5.206 | 1.829 | −1.628 | 5.111 |
Volatility | 293 | 0.013 | 0.003 | −0.317 | −0.015 | 0.002 | 0.022 |
Panel B: correlation matrix analysis | |||||
---|---|---|---|---|---|
Correlation | AGSV | AWiki | ARET | ATVOL | Volatility |
AGSV | 1.000 | −0.329 | 0.162 | −0.014 | 0.00002 |
AWiki | −0.329 | 1.000 | −0.098 | 0.150 | −0.005 |
ARET | 0.162 | −0.098 | 1.000 | −0.079 | −0.480 |
ATVOL | −0.014 | 0.150 | −0.079 | 1.000 | 0.122 |
Volatility | 0.00002 | −0.005 | −0.480 | 0.122 | 1.000 |
Regression results from descriptive and predictive models of ARET, regress on AGSV, AWiki and other control variables. Columns (1)–(6) report results from a single regression models to explain the various independent variables. Columns (6) corresponds to multiple explanatory regressions model. Columns (7)–(11) display the corresponding results for predictive models
Dependent variable: ARET | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Explanatory models | Predictive models | |||||||||||
Variables | (1) | (2) | (3) | (4) | (5) | (6) | Variables | (7) | (8) | (9) | (10) | (11) |
ARETt−1 | 0.404*** [0.071] | 0.302*** [0.699] | ARETt−1 | 0.299*** [0.066] | ||||||||
AGSV | 0.004*** [0.001] | 0.001 [0.001] | AGSVt−1 | 0.002 [0.002] | −0.001 [0.001] | |||||||
Awiki | −0.007*** [0.001] | −0.004*** [0.001] | Awikit−1 | −0.008*** [0.001] | −0.006*** [0.002] | |||||||
ATVOL | −0.001 [0.001] | 0.0008 [0.001] | ATVOLt−1 | −0.001 [0.001] | −0.001*** [0.001] | |||||||
Volatility | 0.930** [0.428] | −0.046 [0.624] | Volatilityt−1 | 1.220*** [0.393] | 0.511 [0.612] | |||||||
Constant | −0.009*** [0.001] | −0.012** * [0.001] | −0.015*** [0.001] | −0.015** * [0.001] | −0.027** [0.005] | −0.004 [0.008] | Constant | −0.013*** [0.01] | −0.015*** [0.001] | −0.015*** [0.001] | −0.0317*** [0.005] | −0.0182** [0.008] |
R-Squaied | 0.165 | 0.034 | 0.099 | 0.001 | 0.019 | 0.182 | R-Squared | 0.008 | 0.120 | 0.004 | 0.033 | 0.206 |
Adjust. R2 | 0.161 | 0.029 | 0.094 | −0.002 | 0.015 | 0.161 | Adjust. R2 | 0.003 | 0.116 | 0.0002 | 0.029 | 0.187 |
Durbin-W | 2.076 | 1.276 | 1.461 | 1.197 | 1.235 | 1.986 | Durbin-W | 1.277 | 1.424 | 1.182 | 1.234 | 1.996 |
Note(s): 1. The symbols ***, ** and * indicates significance at the 1%, 5%, and 10% levels, respectively, 2. White heteroscedasticity-consistent (Eicker-Wliite) standard errors and covariances are applied to all of the models (1)–(11)
Regression results from descriptive and predictive models of ATVOL, regress on AGSV, Awiki and other control variables. Columns (1)–(6) report results from a single regression models to explain the various independent variables. Columns (6) corresponds to multiple explanatory regressions model. Columns (7)–(11) display the corresponding results for predictive models
Dependent variable: ATYOL | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Explanatory models | Predictive models | |||||||||||
Variables | (1) | (2) | (3) | (4) | (5) | (6) | Variables | (7) | (8) | (9) | (10) | (11) |
ATVOLt−1 | 0.334*** [0.068] | 0.397*** [0.080] | 0.380*** [0.083] | 0.375*** [0.076] | 0.270*** [0.054] | 0.242*** [0.061] | ATVOLt−1 | 0.397*** [0.081] | 0.379*** [0.081] | 0.369*** [0.072] | 0.362*** [0.075] | 0.456*** [0.107] |
AGSV | −0.039*** [0.097] | 0.009 [0.049] | AGSVt−1 | −0.089 [0.062] | −0.047 [0.063] | |||||||
Awiki | 0.073 [0.062] | 0.108*** [0.057] | Awikit−1 | 0.103 [0.061] | 0.011 [0.066] | |||||||
ARET | −1.176 [2.622] | −0.640 [2.697] | ARETt−1 | −7.141** [2.252] | −8.388*** [2.607] | |||||||
Volatility | 107.797*** [15.591] | 180.085*** [21.708] | Volatilityt−1 | −17.373* [17.065] | −147.160** [27.538] | |||||||
Constant | −0.022*** [0.055] | −0.009*** [0.089] | 0.016*** [0.069] | −0.052*** [0.076] | −1.349*** [0.191] | −2.460 [0.298] | Constant | −0.007*** [0.062] | 0.028*** [0.070] | −0.145*** [0.056] | 0.192 [0.229] | 0.541 [0.399] |
R-Squared | 0.111 | 0.159 | 0.162 | 0.144 | 0.267 | 0.422 | R-Squared | 0.165 | 0.168 | 0.175 | 0.114 | 0.216 |
Adjust. R2 | 0.108 | 0.151 | 0.154 | 0.137 | 0.262 | 0.407 | Adjust. R2 | 0.157 | 0.160 | 0.166 | 0.108 | 0.197 |
Durbin-W | 1.994 | 1.980 | 1.972 | 1.992 | 1.772 | 1.827 | Durbin-W | 1.981 | 1.985 | 1.982 | 1.995 | 1.992 |
Note(s): 1. The symbols ***, ** and * indicates significance at the 1%, 5%, and 10% levels, respectively, 2. White heteroscedasticity-consistent (Eicker-White) standard errors and covariances are applied to all of the models (1)–(11)
Regression results from descriptive and predictive models of Volatility, regressed on AGSV, AWiki and other control variables. Columns (1)–(6) report results from a single regression models to explain the various independent variables. Columns (6) corresponds to multiple explanatory regressions model. Columns (7)–(11) display the corresponding results for predictive model
Dependent variable: Volatility | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Explanatory models | Predictive models | |||||||||||
Variables | (1) | (2) | (3) | (4) | (5) | (6) | Variables | (7) | (8) | (9) | (10) | (11) |
Volatilityt−1 | 0.509*** [0.049] | 0.204*** [0.067] | 0.203*** [0.068] | 0.417*** [0.060] | 0.417*** [0.045] | 0.135*** [0.053] | Volatilityt−1 | 0.202*** [0.068] | 0.202*** [0.067] | 0.425*** [0.059] | 0.545*** [0.056] | 0.081* [0.083] |
AGSV | −0.00008 [0.000] | −0.0001 [0.000] | AGSVt−1 | 0.0001 [0.000] | 0.0001 [0.000] | |||||||
Awiki | −0.00007 [0.000] | −0.0004** [0.000] | Awikit−1 | −0.00005 [0.000] | −0.0002 [0.000] | |||||||
ARET | 0.007 [0.009] | −0.004 [0.008] | ARETt−1 | 0.009** [0.009] | −0.008 [0.009] | |||||||
ATVOL | 0.001*** [0.000] | 0.001*** [0.000] | ATVOLt−1 | −0.0003 [0.000] | 0.0006** [0.000] | |||||||
Constant | 0.006*** [0.000] | 0.011*** [0.000] | 0.011 [0.000] | 0.007 [0.000] | 0.006*** [0.00] | 0.011 [0.000] | Constant | 0.011*** [0.000] | 0.011*** [0.00] | 0.007*** [0.000] | 0.005*** [0.00] | 0.012*** [0.001] |
R-Squared | 0.256 | 0.042 | 0.042 | 0.186 | 0.413 | 0.372 | R-Squared | 0.044 | 0.042 | 0.192 | 0.261 | 0.073 |
Adjust. R2 | 0.253 | 0.032 | 0.033 | 0.179 | 0.409 | 0.356 | Adjust. R2 | 0.034 | 0.032 | 0.185 | 0.256 | 0.051 |
Durbin-W | 2.106 | 1.927 | 1.930 | 2.057 | 1.927 | 1.866 | Durbin-W | 1.932 | 1.934 | 2.057 | 2.114 | 1.906 |
Note(s): 1. The symbols ***, ** and * indicates significance at the 1%, 5%, and 10% levels, respectively, 2. White heteroscedasticity-consistent (Eicker-White) standard errors and covariances are applied to all of the models (1)–(11)
Twenty-four (24) NGSE-30 stock companies index with tickers as well as their associated sector included in our sample for empirical analysis. This reduction is due to non-availability of stock data
S/N | Ticker | Company | Sector |
---|---|---|---|
Panel A | |||
1 | Access | Access bank PLC. | Financial services |
2 | ConoiL | Conoil PLC | Oil and gas |
3 | Dangcem | Dangote cement PLC | Industrial goods |
4 | FBNH | FBN holdings PLC | Financial services |
5 | Flourmill | Flour mills Nig. PLC. | Consumer goods |
6 | FO | Forte oil PLC. | Oil and gas |
7 | Guaranty | Guaranty trust bank PLC. | Financial services |
8 | Guinness | Guinness NIG PLC | Consumer goods |
9 | Intbrew | International breweries PLC. | Consumer goods |
10 | Jberger | Julius berger NIG. PLC. | Construction/real est. |
11 | Wapco | Lafarge Africa PLC. | Industrial goodS |
12 | Nestle | Nestle Nigeria PLC. | Consumer goods |
13 | NB | Nigerian brew. PLC. | Consumer goods |
14 | Oando | Oando PLC | Oil and gas |
15 | PZ | P Z Cussons Nigeria PLC. | Consumer goods |
16 | Seplat | Seplat petrol. dev. comp. PLC | Oil and gas |
17 | Stanbic | Stanbic IBTC holdings PLC | Financial services |
18 | Sterlnbank | Sterling bank PLC. | Financial services |
19 | Transcorp | Transnational corp. of Nig. PLC | Conglomerates |
20 | UACN | U A C N PLC. | Conglomerates |
21 | UBA | United bank for Africa PLC | Financial services |
22 | Unitybnk | Unity bank PLC | Financial services |
23 | Unilever | Unilever Nigeria PLC. | Consumer goods |
24 | Zenithbank | Zenith bank PLC | Financial services |
Panel B | |||
1 | Mobil | 11 PLC | Conglomerates |
2 | Dangsugar | DANGOTE SUGAR REFINERY PLC | Consumer goods |
3 | ETI | ECOBANK TRANSNATIONAL INCORP. | Financial services |
4 | Okomuoil | OKOMU OIL PALM PLC. | Agriculture |
5 | Presco | PRESCO PLC | Agriculture |
6 | Total | Total Nigeria PLC. | Oil and gas |
Estimation results for model robustness. SM (A) and (B) indicates standardization methods by Bijl et al. (2016) and Da et al. (2011) respectively
Variables | Returns | Trading volume | Volatility | |||
---|---|---|---|---|---|---|
SM: A | SM: B | SM: A | SM: B | SM: A | SM: B | |
−0.0420*** | −0.0447*** | 0.184 | −0.191 | 0.008*** | 0.009*** | |
0.2989** | 0.3136*** | −7.575*** | −1.864** | −0.015 | −0.011 | |
0.0063 | −0.0499 | −0.344 | −0.500 | 0.008 | 0.004 | |
−0.0382 | 0.0422 | −1.230 | 0.102 | 0.0002 | 0.002 | |
−0.023 | −0.0257 | 1.284 | 0.613 | −0.006 | −0.003 | |
0.0477 | 0.0633 | −3.908 | −0.536 | 0.003 | 0.011 | |
−0.0009 | −0.0012 | 0.411*** | 0.367*** | 0.0005* | 0.001* | |
−0.0023 | −0.0016 | 0.003* | −0.134** | −0.0002 | 0.0001 | |
0.0003 | −0.00009 | −0.097 | 0.114 | −0.0003 | 0.0009 | |
−0.0022 | −0.0018 | 0.123* | 0.100* | 0.0002* | 0.001* | |
0.0015 | 0.00002 | 0.160* | −0.139* | 0.0002 | −0.0008 | |
0.2164 | 0.6844 | −30.264 | −10.683 | 0.143* | 0.159** | |
0.5015 | 0.1020 | 18.491* | −5.654** | 0.0278 | −0.045 | |
1.4482 | 1.4942** | 73.890*** | 24.375* | 0.328*** | 0.225*** | |
−0.2415 | −0.2485 | −16.792 | −0.796 | 0.176** | 0.183** | |
0.2558 | 0.5972 | −73.140** | 1.712* | −0.291* | −0.208** | |
0.0003* | 0.0037 | −0.154 | 0.568** | −0.0002 | 0.004 | |
−0.0082*** | −0.043* | 0.299 | −0.602* | 0.0001 | 0.001 | |
0.0064** | 0.0094* | −0.248 | 0.199 | −0.00009 | 0.0004 | |
−0.0007 | −0.0218 | −0.111 | 0.337 | −0.0002 | −0.003 | |
−0.0048** | −0.0042* | 0.118 | −0.175 | 0.0001 | −0.002 | |
−0.0045* | −0.0154 | −0.058 | 0.145 | 0.00030 | −0.00008 | |
0.0044 | 0.0367* | 0.088 | 0.217 | 0.0001 | 0.002 | |
0.00006 | −0.0362 | 0.020 | 0.089 | 0.00001 | 0.00009 | |
−0.0009 | 0.0522 | −0.141 | 0.105 | −0.005** | −0.001 | |
−0.0003 | −0.0343 | 0.006 | 0.079 | 0.00004 | 0.0002 | |
R Squared | 0.303 | 0.3122 | 0.345 | 0.340 | 0.320 | 0.317 |
Adjusted R2 | 0.206 | 0.2161 | 0.253 | 0.248 | 0.225 | 0.221 |
Durbin–Watson | 2.003 | 2.0003 | 2.038 | 1.998 | 2.004 | 1.996 |
Note(s): Robust standard errors are reported in square bracket and significant level
*p-value < 0.1; **p-value < 0.05; ***p-value < 0.01
Notes
Google search statistics (2019), https://99firms.com/blog/google-search-statistics
The World Bank in Africa (2019), https://www.worldbank.org/en/region/afr/overview
Africa Internet Users Statistics (2019), https://www.internetworldstats.com/stats1.htm.
Nigerian Communications Commission (2019), https://www.ncc.gov.ng
References
Adachi, Y., Masuda, M. and Takeda, F. (2017), “Google search intensity and its relationship to the returns and liquidity of Japanese startup stocks”, Pacific Basin Finance Journal, Vol. 46, pp. 243-257.
Alizadeh, S., Brandt, W.M. and Diebold, X.F. (2002), “Range-based estimation of stochastic volatility models”, Journal of Finance, Vol. 57, pp. 1047-1091.
Antweiler, W. and Frank, M.Z. (2004), “Is all that talks just noise? The information content of internet stock message boards”, The Journal of Finance, Vol. 59, pp. 1259-1294.
Aouadi, A., Arouri, M. and Teulon, F. (2013), “Investor attention and stock market activity: evidence from France”, Economic Modelling, Vol. 35, pp. 674-681.
Audrino, F., Sigrist, F. and Ballinari, D. (2019), “The impact of sentiment and attention measures on stock market volatility”, International Journal of Forecasting, doi: 10.1016/j.ijforecast.2019.05.010.
Bank, M., Larch, M. and Peter, G. (2011), “Google search volume and its influence on liquidity and returns of German stocks”, Financial Markets and Portfolio Management, Vol. 253, pp. 239-264.
Batista, G. and Monard, M. (2003), “An analysis of four missing data treatment methods for supervised learning”, Applied Artificial Intelligence, Vol. 17 No. 5-6, pp. 519-533.
Bijl, L., Kringhaug, G., Molnár, P. and Sandvik, E. (2016), “Google searches and stock returns”, International Review of Financial Analysis, Vol. 45, pp. 150-156.
Brandt, W.M. and Kinlay, J. (2005), Estimating Historical Volatility, available at: http://www.investment-analytics.com.
Cergol, B. and Omladič, M. (2015), “What can Wikipedia and Google tell us about stock prices under different market regimes?”, Ars Mathematica Contemporanea, Vol. 9 No. 2, pp. 301-320.
Chesney, M., Crameri, R. and Mancini, L. (2015), “Detecting abnormal trading activities in option markets”, Journal of Empirical Finance, Vol. 33, pp. 263-275.
Chung, K.H. and Chuwonganant, C. (2018), “Market volatility and stock returns: the role of liquidity provider”, Journal of Financial Markets, Vol. 37, pp. 17-34.
Da, Z., Engelberg, J. and Gao, P. (2011), “In search of attention”, The Journal of Finance, Vol. 66 No. 5, pp. 1461-1499.
Dimpfl, T. and Jank, S. (2016), “Can internet search queries help to predict stock market volatility?”, European Financial Management, Vol. 22 No. 2, pp. 171-192.
Dimpfl, T. and Kleiman, V. (2019), “Investor pessimism and the German stock market: exploring google search queries”, German Economic Review, Vol. 20 No. 1, pp. 1-28.
Fama, E.F. and French, K.R. (1993), “Common risk factors in the returns on stocks and bonds”, Journal of Financial Economics, Vol. 33 No. 1, pp. 3-56.
Fama, E.F. and French, K.R. (2015), “A five-factor asset pricing model”, Journal of Financial Economics, Vol. 116 No. 1, pp. 1-22.
Fang, L. and Peress, J. (2009), “Media coverage and the cross-section of stock returns”, The Journal of Finance, Vol. 645, pp. 2023-2052.
Fehle, F., Tsyplakov, S. and Zdorovtsov, V. (2005), “Can companies influence investor behavior through advertising? super bowl commercials and stock returns”, European Financial Management, Vol. 11, pp. 625-647.
Garman, M.B. and Klass, M.J. (1980), “On the estimation of security price volatilities from historical data”, Journal of Business, Vol. 53 No. 1, pp. 67-78.
Hervé, F., Zouaoui, M. and Belvaux, B. (2019), “Noise traders and smart money: evidence from online searches”, Economic Modelling, Vol. 83, pp. 141-149.
Huang, M.Y., Rojas, R.R. and Convery, P.D. (2019), “Forecasting stock market movements using google trend searches”, Empirical Economics, doi: 10.1007/s00181-019-01725-1.
Kim, Y.H. and Meschke, F. (2011), “CEO interviews on CNBC”, Working Paper, available at: http://ssrn.com/abstract=1745085.
Kim, N., Lucivjanska, K., Molnar, P. and Villa, R. (2019), “Google searches and stock market activities: evidence from Norway”, Finance Research Letters, Vol. 28, pp. 208-220.
Moat, H.S., Curme, C., Avakian, A., Kenett, D.Y., Stanley, H.E. and Preis, T. (2013), “Quantifying wikipedia usage patterns before stock market moves”, Scientific Reports, Vol. 3, pp. 1801, doi: 10.1038/srep01801.
Parkinson, M. (1980), “The extreme value method for estimating the variance of the rate of return”, Journal of Business, Vol. 53, pp. 61-68.
Preis, T., Moat, H.S. and Stanley, H.E. (2013), “Quantifying trading behavior in financial markets using Google Trends”, Scientific Reports, Vol. 3, pp. 1684.
Rogers, L.C.G. and Satchell, S.E. (1991), “Estimating variance from high, low and closing prices”, Annals of Applied Probability, Vol. 1, pp. 504-512.
Takeda, F. and Wakao, T. (2014), “Google search intensity and its relationship with returns and trading volume of Japanese stocks”, Pacific-Basin Finance Journal, Vol. 27, pp. 1-18.
Takeda, F. and Yamazaki, H. (2006), “Stock price reactions to public TV programs on listed Japanese companies”, Economics Bulletin, Vol. 137, pp. 1-7.
Yang, D. and Zhang, Q. (2000), “Drift independent volatility estimation based on high, low, open and close prices”, Journal of Business, Vol. 73, pp. 477-491.
Yoshida, M., Arase, Y., Tsunoda, T. and Yamamoto, M. (2015), “Wikipedia page view reflects web search trend”, Proceedings of the 2015 ACM Web Science Conference, November, California, CA, Article No. 65, pp. 1-2, doi: 10.1145/2786451.2786495.
Zagoruyko, S. and Komodakis, N. (2019), “Paying more attention to attention: improving the performance of convolutional neural networks via attention transfer”, 5th International Conference on Learning Representations, Iclr 2017 - Conference Track Proceedings, New Orleans, Louisiana.
Zhang, J., Djajadikerta, H.G. and Zhang, Z. (2018), “Does sustainability engagement affect stock return volatility? Evidence from the Chinese financial market”, Sustainability, Vol. 10 No. 10, pp. 336.
Further reading
Joseph, K., Wintoki, M.B. and Zhang, Z. (2011), “Forecasting abnormal stock returns and trading volume using investor sentiment: evidence from online search”, International Journal of Forecasting, Vol. 27 No. 4, pp. 1116-1127.
Khelladi, I. and Boutinot, A. (2017), “The role of wikipedia on corporate E-reputation: evidence from French companies”, International Studies of Management and Organization, Vol. 47 No. 1, pp. 23-41.
Acknowledgements
The author sincerely thanks the editor and the anonymous reviewer(s) for their very helpful comments and insightful suggestions. For any other error(s), the usual disclaimer applies.Conflicts of interest: The author declares that to the best of the knowledge, there is no conflict of interest. Funding: The author received no funding for this research.