Investors’ attention: does it impact the Nigerian stock market activities?

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.

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.

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 $GSV$ and dividing their difference with the standard deviation of the previous year. Then we compute the abnormal $GSV$ (hereafter referred to as AGSV).

where; $Xt$ represents search volume of company $i$ in week $t$⁠;

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.

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;

1. 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

2. We check and compare firms name as well as their associated ticker, stated in Wikipedia pages (https://en.wikipedia.org).

3. We exclude firm’s names and tickers that are found to be ambiguous or associated with searches other than firm’s financial information.

4. 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.

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 1 Table 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.

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;

where; $Rt$ is the raw log return, $Pt$ is the adjusted stock price for week t, P_t-1 is the adjusted stock price from the previous week.

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;

where;

$Rt$ is the total return of stock $t,Rm,t$ is the market return at time t,

$$$Rf,t$ is the risk-free rate at time t, $Rm,t−Rf,t$ is the risk premium at time t,

$SMBt$ is the difference between small and big stock returns based on market capitalization;

$HML$ is the difference between high and low book-to-market stock;

$RMWt$ is the profitability factor; $CMA$ is the investment factor

$β1,2,3,4,5$ are the estimated coefficients.

where;

$returni,d$ is the stock i’s return in day $d$ of week $w$⁠.

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 $ATVol$ and the $AGSV$⁠, the values of the daily trading volume are first converted to the weekly trading volume $TVolt$⁠. Thus;

where; |S_t| is the numbers of trading days in each week. Thus; the values of the total volume $TVolt$ are standardized as shown in Eqn (1)

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 $(hight)$ and low $(lowt)$ values of daily prices to estimate the variance, see Eqn (6). Monte Carlo simulation and empirical data results by Brandt and Kinlay (2005), document that the Parkinson (1980) volatility estimator is highly more superior to the Garman and Klass (1980); Rogers and Satchell (1991); Alizahdeh et al. (2002) and Yang and Zhang (2000) when empirical data derived from the S&P 500 Index were used. We calculate the variance thus;

where;

C_t = Closing price on day t; O_t = Opening price on day t;

H_t = High price on day t; L_t = Low price on day t;

$ct=lnCt−lnOt$ , the normalized close price;

$ot=lnOt−lnCt−1$ , the normalized open price;

$ut=lnHt−lnOt$ , the normalized high price;

$dt=lnLt−lnOt$ , the normalized low price;

n = number of daily periods.

The weekly volatility as a square root of average daily variance is calculated thus;

Table 1 presents present two parts; the panel A which is the summary statistics and Panel B which contains correlation analysis of $ARet;$ $AGSV;$ $AWiki$⁠, $ATVol$ and $Volatility$ of the NGSE-30 index between August 3, 2015 and August 2, 2019. Panel A shows the descriptive statistics of the study variables. The values of St. Dev. and mean for AGSV and Awiki appears to follow the normal distribution. Volatility is shown to be low, indicating that the historical stock prices are low with less investment risk. While ARET and Volatility have lighter tails than the normal distribution, AGSV, AWiki and ATVOL excess kurtosis appears to show sign of heaviness at the tail than the normal distribution. Almost all the variables except volatility are skewed to the left. The $AGSV$⁠, $AWiki$ and $trading Volume$ are calculated using the Bijl et al. (2016) formula. Whilst volatility is calculated using Parkinson (1980) range-based volatility estimator, stock return is calculated from the 5-factors asset pricing model. The upper right triangular matrix side of panel B report the correlation matrix analysis of our data. Our data shows low correlation values of both negative and positive.

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 $AGSV$ and $AWiki$ are proxies for investors’ attention, while the stock return, trading volume and volatility are use as the dependent variables as well as control variables alternating in the model. We regress $ARet$⁠, $ATVol$ and $Volatility$ against the $AGSV$ and $AWiki$ plus the control variables in order to control the market effect. The lag variables of the dependent variables are included in our model in order to control the autoregressive pattern.

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 $AGSV$ and $AWiki$ explains and predict the future of NGSE-30 $ARet$⁠, respectively. While the former equation investigates the contemporary descriptive relationship between stock $ARet$ and investors’ attention, the later examines the predictive relationship between stock $ARet$ and investors’ search query. The first lagged value of $ARet$ controls the autoregressive in both models.

where; $ARet, AGSV, AWiki, ATVol,Volatility, βi$ are the abnormal return, abnormal GSV, Wikipedia page views, trading volume, volatility and the regression coefficients, respectively.

Similarly, the trading volume is regressed on of $AGSV$ and $AWiki$ and the control variables (returns and volatility). The estimation results from Eqn (9) and (9′) examines if there is the existence of a relationship between the trading volume and the investors’ attention. In other word, it examines whether the lagged values of the independent variables can predict the future values of trading volume, or whether the investors’ search intensity can explain the variations associated with the trading volume. The first lagged of the trading volume is added to the explanatory and predictive models to control the autoregressive effects.

where $ATVol, ARet, AGSV, AWiki, Volatility, βi$ are the trading volume, abnormal return, abnormal GSV, Wikipedia page views, volatility and the regression coefficients, respectively.

Eqns (10) and (10′) investigate whether there is a contemporary and predictive relationship between $AGSV$ and $Wiki$ on one hand and volatility on the other hand. We regressed volatility on its lagged value including the lagged values of the control variables as shown in Eqn (10′). Thus;

where; $Volatility; ATVol; ARet; AGSV; AWiki, βi$ are volatility, trading volume, abnormal return, abnormal GSV, Wikipedia page views and the regression coefficients respectively.

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 $AGSV$ and $AWiki$⁠) in explaining and predicting abnormal returns $(ARet)$⁠, trading volume and volatility, we employ the multivariate regression models. Results shown in Tables 2–4 and Table A2, have small values of $R2$ and adjusted $R2$ but are with robust white heteroscedasticity-consistent (Eicker-White) standard errors.

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 $ARet$ time series, see Table 2. Although significant at 99% confidence level at the initial regression of the contemporaneous model, the $AGSV$ becomes insignificant (both in univariate and multivariate models) when regressed together with other variables both in descriptive and predictive models. However, changes in the numbers of $AWiki$ page view are highly significant and negatively explaining the variation in the stock returns $(ARet)$⁠. This implies that a higher stock return could lead to a smaller number of investors viewing wiki pages related to stock activities. This is likely the case of a worried investors who get very nervous about the stock market. Hence, they quickly seek information about the financial market before trying to dump their stock. This behavior increases stocks trading activities before crashes.

Results from Eqns (9) and (9′) shown in Table A1, investigate the impact of search queries with the abnormal trading volume $(ATVol)$ as the dependent variable. Similarly, the $AGSV$ can neither explain nor predict the $ATVol$ of firms listed on NGSE-30 index, as demonstrated in both the univariate and multivariate models. The $AWiki$ page view is positively insignificant in predicting the $(ATVol)$ both in the univariate and multivariate models, see; Columns (8) and (11). However, it is significant at 99% confidence level in the descriptive model when the explanatory multivariate model in column (6) is applied. Results also indicate that with 99% confidence level, most of the variation in $ATVol$ can be explained by the variable $(ATVol)$ using the current and previous weeks. Although the regression values of R squared and adjusted R² are small, the values of the $Durbin Watson$ test statistic indicates that variables are serial correlation free and the estimated coefficients results are Eicker-White robust. Finally, the estimation in Eqns (10) and (10′) shown in Table 4 reveals the drivers of movements in the volatility dynamics of NGSE-30 index. The estimated values of $AGSV$ in the two models (explanatory and predictive), with $Volatility$ as the dependent variable are insignificant. This suggests that $AGSV$ does not influence volatility dynamics of firms on NGSE-30 index. Nevertheless, changes in $AWiki$ page views negatively explain volatility, (see; column (6)) when other variables are taken into consideration.

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, $AGSV$ and $AWiki$ standardization by Da et al. (2011), excess stock return and volatility. The stock return, volume and volatility are also controlled in the predictive model of five lags for each variable, see  Appendix 1, Eqn (11)–(14). We denote the five most recent lags as; $∑i=15Li$⁠, where $Li$ is the lag operator. The multiple regression with robust white heteroscedasticity-consistent (Eicker-White) standard errors is the model estimation with results for the two methods shown in  Appendix 1, Table A2. We regress$ARet$ on $AGSV$ and $AWiki$ including the control variables. Our results indicate the following; The lag values and directions of $AWiki$ in lags 1, 2, 3, 4 and 5 have weak negative coefficients. For $AGSV$⁠, the directions and coefficients in lags 1, 2 and 5 are similar and very weak for both SM (A) and (B). $AGSV$ lag 1 for SM (A) is significant, but non-significant in SM (B). We also observed an increase in R squared and adjusted R squared for both SM, coupled with a better Durbin–Watson test statistic. This result is similar to our main results in Table 2, that the $AGSV$ neither predict the stock returns while the $AWiki$ predict returns negatively.

We also compare the results of trading volume for both SM (A) and (B), shown on  Appendix 1, Table A2. The $AGSV$ have weak coefficients with positive non-significant directions at lags 2, 3 and 5. Even so, the $AWiki$ coefficients for both SM are in different directions and non-significant. These results imply that both the $AGSV$ and $AWiki$ does not predict the values of the trading volume. Lastly, the regression estimation results in  Appendix 1, Table A2 also reports volatility results using the alternative model for both SM (A) and (B) in five lags. Except for $AGSV$ lag 1, other lags (2,3,4 and 5) have coefficients with the same directions but weak non-significant values. Similarly, the lags values of $AWiki$ in both SM are in different directions, non-significant and weak. This estimation result is similar to results in Table 4 in our main study. Above all, the estimation results on Tables 2–4 are consistent with results from both SM (A) and (B) as shown in  Appendix 1, Table A2. This implies that investors’ Google searches does not predict the stock returns, trading volumes and the volatility dynamics of NGSE-30 index. We also obtained results similar in Tables 2–4 for descriptive model for both SM (A) and (B), but for brevity, we choose to report results with the predictive model.

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.

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.

where; $L,$ denote the lag operator and $∑i=15Li$ indicates a lag operator with five most recent lags.