The informational role of thin options markets: Empirical evidence from the Spanish case

García Martín, C. José; Herrero Piqueras, Begoña; Ibañez Escribano, Ana María Article The informational role of thin options markets: Empirical evidence from the Spanish case Estudios de Economía Provided in Cooperation with: Department of Economics, University of Chile Suggested Citation: García Martín, C. José; Herrero Piqueras, Begoña; Ibañez Escribano, Ana María (2016) : The informational role of thin options markets: Empirical evidence from the Spanish case, Estudios de Economía, ISSN 0718-5286, Universidad de Chile, Departamento de Economía, Santiago de Chile, Vol. 43, Iss. 2, pp. 233-263 This Version is available at: https://hdl.handle.net/10419/194246 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/ The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 233Estudios de Economía. Vol.43 - Nº2, Diciembre 2016. Págs. 233-263 The informational role of thin options markets: Empirical evidence from the Spanish case * 1 El papel informativo de los mercados de opciones estrechos: Evidencia empírica del caso español C. José García Martín** Begoña Herrero Piqueras*** Ana María Ibáñez Escribano**** Abstract This study investigates the informational role of thin options markets, specifically the Spanish options market. Firstly, we examine the effect of options markets by analysing stock market reaction to earnings news, conditional on the availability of options markets. Secondly, we examine options trading activity before the release of earnings news (including the announcement period). The results show that the impact on prices before the earnings release is significantly bigger when options trading is available. Moreover, the dissemination of earnings news is associated with significant unusual activity in the options market due to informed trading, especially when the earnings surprise is highly good. Key words: options market, thin market, informed trading, price discovery process, earnings announcement. JEL Classification: G12, G13, G14. Resumen El trabajo analiza el papel informativo de los mercados de opciones para el caso de escasa negociación, centrándose en el mercado español. Estudiamos el efecto de la llegada de nueva información relevante, como es el anuncio de beneficios, en el mercado de contado bajo la presencia de opciones sobre dichas acciones. Además, examinamos la actividad negociadora en el mercado de opciones ante dicho suceso. Los resultados muestran que el impacto de * The researchers wish to acknowledge Ángel Pardo from the University of Valencia for helpful comments and suggestions and the valuable comments and suggestions made by the anonymous referees. We thank the Cátedra Finanzas Internacionales-Banco Santander for their financial support. All remaining mistakes are our own. ** University of Valencia. Email: [email protected]. *** University of Valencia. Email: [email protected]. **** University of Valencia. Email: ana.m.ibanez @uv.es. Estudios de Economía, Vol.43 - Nº2234 la información es superior cuando existen opciones cotizadas sobre dichas acciones. Adicionalmente, se observa una actividad negociadora anormal en el mercado de opciones debido a la negociación informada, principalmente cuando la noticia es muy buena. Palabras clave: Mercado de opciones, mercado poco líquido, negociación informada, proceso de formación de precios, anuncio de beneficios. Clasificación JEL: G12, G13, G14. 1. Introduction This study analyses the informational role of options markets. Although traditionally options are seen as redundant assets, nowadays a recurring topic in financial literature is the role that options markets play in the financial system. Specifically, informed traders might prefer trading in the options market due to the advantages that this market offers in terms of trading costs (Black, 1975; Manaster and Rendleman, 1982), higher leverage (Back, 1993; Biais and Hillion, 1994), absence of short sales restrictions, and a greater range of trading strategies by combining positions in options and in the underlying stock. Ultimately, it is the presence of these informed agents in the options market that leads one to hypothesize the improvement of stock market efficiency. For example, as shown in the previous literature, the availability of the options market contributes to overall stock price efficiency in several ways: the stock price adjustment is faster (Jennings and Stark, 1986), market price reactions to earnings announcements are smaller (Skinner, 1990; Ho, 1993), and post-announcement drifts are less pronounced (Botosan and Skinner, 1993) in stocks with listed options. The purpose of our study is to analyse the stock market reaction to the arrival of new information conditional on the availability of options trading, in particular the release of earnings news. The fact that the earnings announcements have a clear impact on the value of a firm1 makes this an optimal event to analyse whether options markets enhance the informational efficiency of stock markets (Amin and Lee, 1997; Mendenhall and Fehrs, 1999; Donders etal. 2000; Roll etal. 2010; Billings and Jennings, 2011; Johnson and So, 2012; and Hu, 2014, among others). The first aim of our study is to test whether the formation of the price in the stock market is more efficient with the existence of listed options and if there is a migration of informed agents in favour of the options market. To this end, we study the stock market behaviour in terms of price formation, trading activity and liquidity around the announcement of quarterly and annual earnings in a sample of companies that present such information at two different times, that is, before and after the options listing. We expect that the response to the earnings 1 One must highlight previous literature (Beaver, 1968; Ball and Brown, 1968; Beaver etal., 1979; Atiase, 1985, 1987; Pope and Inyangete, 1992; Opong, 1995; Elsharkawy and Garrod, 1996; Booth etal., 1997; Gajewski and Quéré, 2001; among others, and for the Spanish market: Arcas and Rees, 1999; Sanabria, 2005; and Garcia etal. 2008, 2010). The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 235 announcement of prices of optioned stocks is larger and more complete than similar non-optioned stocks because of their higher informational efficiency and also a reduction in the informed trading on the stock market. Moreover, we analyse the behaviour of the trading activity in the options market around the release of the earnings in order to detect if there is any activity related to the information released. Due to the presence of informed agents, abnormal trading activity on the options market is expected. Finally, we investigate if informed agents anticipate earnings announcements and take advantage of this knowledge by trading on the options market. For this purpose, we consider the relation between the cumulative abnormal returns prior to the release of the new information and the previous trading activity on the options market. We expect that options traders execute orders in the right direction for the upcoming earnings surprise. The results of the analysis of the stock markets show that the impact on prices is significantly bigger when there are options listed. Additionally, the trading activity shows that the number of orders is significantly larger and the average size of the transactions shows a significant decrease. In light of these results, we can hold that there is an increase in the trading activity of uninformed agents. With respect to the options market, the results show significant abnormal trading activity whose sign depends on the analysed variable and the degree of moneyness. Finally, we observe that in those announcements where the impact is the greatest, there is more trading in options prior to an earnings release, which is reflected in a greater number of open positions. Our results suggest that there are informed agents that trade actively in the options markets in anticipation of earnings news, mainly when the earnings surprise is good. This behaviour improves the efficiency of the market. To the best of our knowledge, the first look at trading activity on the Spanish options market around the release of economic news is Blasco etal. (2010), who analyse the relationship between informed trading volume in the options market and variations in underlying index prices around the arrival of economic news. We expand this evidence for the Spanish market using firm specific news, earnings announcements, stock options and two additional measures of informed trading.2 Moreover, we contribute to the literature in different ways. Firstly, our study analyses the relevance of accounting information in the options market in order to determine whether thin markets, such as the Spanish options market, play a similar role to that observed in other markets with higher trading activity. Secondly, we contribute by considering the behaviour of the open interest in options markets before earnings announcements from the point of view of informed trading. Although trading volume in the options market may be informative about the price discovery process in financial markets, volume by itself does not give any indication of the direction of the transactions. In this regard, changes in open interest may provide additional evidence with respect to the informational role of options. Finally, we control for stock volume with the relative trading activity in options and stock markets, measures developed by Roll etal. (2010). As well, we bring new evidence employing panel data 2 A preliminary version of this paper was presented at the XXVII AEDEM Annual Meeting, Huelva 2013. Estudios de Economía, Vol.43 - Nº2236 methodology focused on the Spanish market. The panel data model provides estimators that are more efficient than linear regression models and allow one to control for individual unobserved heterogeneity (one of the major problems in non-experimental research). The remainder of the paper is organized as follows. Section 2 provides the background motivation and a literature review and sets forth the tested hypotheses. In section 3 we describe our data sets. The methodology and results for the analysis of the stock market reaction to earnings news conditioned on the availability of options markets are detailed in section 4. Section 5 presents the results of the regression analysis of the determinants of options trading variables. Section 6 analyses the relation of options trading variables to cumulative abnormal returns. Section 7 concludes. 2. Hypotheses and previous literature In general, if the markets are incomplete, one supported hypothesis is that the options market improves equity market efficiency (Black, 1975; Mayhew etal., 1995). The arguments in favour of this hypothesis are based on, firstly, a higher level of information around stocks with options and, secondly, that certain characteristics of the options market make it more attractive to informed traders (lower transaction costs, absence of short sales restrictions, and greater financial leverage). The above-mentioned hypothesis would not be supported if the markets were complete because in that case options are redundant securities (Black and Scholes, 1973; Merton, 1973). The role of the options market in the underlying stock price discovery has received considerable attention in previous studies. Specifically, the effect of the options market on stock market efficiency around earnings announcements has been extensively studied. The main findings are consistent with a significantly different price discovery between non-optioned and optioned stocks, with a key finding being that the options market increases the informational efficiency of the stock market motivated by informed trading. In this regard, Jennings and Starks (1986) study the effects of options trading on the behaviour of underlying stock prices around earnings releases. The authors find that the non-optioned firms require substantially more time to adjust the price to the new information than optioned ones. Furthermore, Mendenhall and Fehrs (1999) measure the impact of options listing on the magnitude of the absolute stock price movement at the time of earnings announcements. They argue that options trading increases the speed of price adjustment to earnings before the release. Their results indicate a larger and more complete earnings announcement response for optioned firms as a result of informed traders. The idea that informed traders prefer to trade in options markets is also supported by Skinner (1990) who studies the stock price response to earnings when the stock is listed in the options market. He finds that market price reactions to earnings announcements are smaller after options listing. The same result is obtained by Ho (1993), who finds a smaller earnings response for optioned firms on the announcement date because stock prices of optioned firms anticipate earnings information. Similar results are obtained by Botosan and Skinner The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 237 (1993) who conclude that the post earnings announcement price drift is lower for optioned firms. On the other hand, several studies have investigated the impact of stock options listings on several aspects of the market quality of the underlying stocks. Kumar etal. (1998) suggest that options listings improve the market quality of the underlying stocks. They observe an improvement in the liquidity –a decrease in the spread and an increase in the quoted depth–, and an increase in trading volume, trading frequency and transaction size after options listing. Fedenia and Grammatikos (1992) study the effect of options listing on the bid-ask spread of the underlying stocks and they report that the bid-ask spread declines for highly liquid stocks, but increases for illiquid stocks. Earnings announcements convey relevant price information and their timing is largely predictable. This fact can motivate informed traders to trade on private information and take advantage of it. Under the hypothesis that informed traders prefer to trade in the options market for the advantages that this market offers, and in light of previous literature, we consider the follow hypotheses: H1. After options listing, a larger proportion of the total stock price adjustment associated with earnings release takes place before the announcement. H2. We expect a substitution effect in informed trading activity in favour of the options market. H3. We expect an improvement in stock market liquidity because of the new allocation of information in the financial markets after options listing. The improvement in stock market efficiency as a result of the presence of options means, as noted above, that informed traders take positions in options markets ahead of the release of the information about corporate events, and that there is more information circulating in optioned firms. If the reason for the improvement in efficiency is the presence of informed agents taking advantage of their information, we expect an abnormal volume reaction to these information releases in the options market. In this regard, some empirical studies have sought to explain the interaction of the options market and equity returns surrounding public events. Cao etal. (2005) relate option volume to returns of underlying stocks and show evidence of informed traders in the options market around takeover announcements. Arnold etal. (2006) find abnormal trading activity in the options market around tender offer announcements, and this abnormal trading volume precedes abnormal stock volume. In addition, Amin and Lee (1997), looking at earnings announcements, find that options traders initiate a greater proportion of long (short) positions immediately before good (bad) earnings news. Mendenhall and Fehrs (1999) find that options trading allows for increases in the speed of adjustment of prices to earnings. Roll etal. (2010) consider the relative trading volume in options and stocks and find that it is higher around earnings announcements, suggesting increased trading in the options markets. Moreover, they observe that part of the options trading prior to the release is informed. Truong etal. (2012) conclude that the options market absorbs the information conveyed by the earnings announcement and that this news has a profound impact on option value. For the Spanish market, Blasco etal. (2010) analyse the price impact on the underlying stock index of informed trading in the options market in the presence of eco- Estudios de Economía, Vol.43 - Nº2238 nomic news, concluding that potential informed trading is channelled through out-of-the-money options. Many papers analyse both options and equity trading volume as determinants of equity prices. The impact of open interest has been a less developed topic. Open interest measures options positions created through past trading that are not yet liquidated. Intuitively, open interest is a better measure of the beliefs of investors about future stock returns than trading volume. Schachter (1988) documents a significant decline in options open interest prior to quarterly earnings announcements. Specifically, this effect is most pronounced in those options with a short time to maturity and more sensitivity to volatility. On the contrary, Donders etal. (2000) find an increase in the open interest during the days before the earnings announcements. After the news dissemination, traders cancel part of their options position, thereby reducing open interest to normal levels. Moreover, previous literature suggests that informed trading depends on the degree of option moneyness, that is, at the money (ATM), in the money (ITM) or out of the money (OTM). Chakravarty etal. (2004) and Chen etal. (2005), among others, argue that this is an important issue because options with different degrees of moneyness have different levels of liquidity and different degrees of leverage. Theory suggests that informed agents would prefer OTM options due to their greater leverage, while investor trading on volatility will tend to concentrate on ATM options, as they provide camouflage for their intensions, while commissions tend to be lowest for ITM options. The relative importance of these factors is an empirical question that has not yet been resolved. Blasco etal. (2010) for the Spanish market find that informed investors trade in OTM and ATM options. Meanwhile, Truong etal. (2012) and Billings and Jennings (2011) argue that ATM options will attract the most trading interest due to their high sensitivity to changes in implied volatility. Other authors such as De Jong etal. (2006) argue that informed agents use ITM options to increase their trading profits because they are more sensitive to underlying equity price changes than other options. In the light of this literature, we formulate the following hypothesis: H4. Trading activity in options markets experiences abnormal variations before and on the day of the publication of the earnings announcements. Following previous studies, we will contrast this hypothesis by considering the relationship between the option’s strike price and the underlying stock’s closing price to check whether informed trading exists and, if so, where it is found. Thus we will split options trading volume into ATM, OTM and ITM options and make the analysis for each of the three groups. Finally, as suggested by Roll etal. (2010), if the options trading activity around the earnings release is because of informed agents, we should find a relationship between trading activity in the options market and post-announcements returns, indicating whether informed options traders are acting in the right direction. H5. We expect a negative (positive) and significant relation between options trading activity and the post-announcements returns when the earnings release has a negative (positive) impact on prices. The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 239 3. Sample description We intend to study the price formation process around the quarterly and annual earnings announcements released by those Spanish firms listed on the Spanish electronic stock market (hereinafter SIBE) that have optioned stocks. The contracts on stocks in the Spanish options market are on the largest and most liquid companies of the Spanish stock market. These aspects are of great importance when studying the price formation process around the arrival of new information since these companies are the most informative for investors. Appendix 1 exhibits the firms with Spanish stock options listed and the number of contracts listed and traded for the period of study. Since its inception in 1993, the number of stocks on which options are traded in the Spanish Futures and Options Exchange (MEFF) has increased from four in 1993 (BBVA, Endesa, Repsol y Telefónica) to the 42 that are currently listed. Table1 shows some characteristics of the Spanish stock options market for the sample period. Obviously, the growth in the number of contracts outstanding each year has been progressive, although this trend is not observed in the number of contracts that were traded at least one day. These features, together with the behaviour observed in the trading volume and the number of days, allow us to state that the Spanish stock options market is thin. The study extends from January 2004 to May 2012 and we have used the data described below. · We have daily stock market data for all the firms that have optioned stocks. This data set has been obtained from Sociedad de Bolsas. Specifically, this data set includes closing prices, dividends, capital increases and changes in nominal value, the number of shares listed, the number of transactions and volume (number of shares) traded and the daily average spread. The data set includes the closing prices from the Madrid Stock Exchange Index –hereinafter IGBM– for the period analysed. · We have obtained from the Bank of Spain the daily return on Letras del Tesoro (Spanish Treasury Bill) for the same period. · We prepared a database consisting of the date of the announcement of quarterly and annual earnings released by Spanish firms on SIBE for the period from the first quarter 2004 to the last quarter 2011. This data set was obtained from the Spanish Security Exchange Commission (hereinafter CNMV) and by consulting the financial press. · Daily information on stock options is provided by MEFF, the Spanish Financial Futures Market, a subsidiary of Bolsas y Mercados Financieros (BME), the company which runs Spain’s securities markets. Specifically, we have used daily data of all contracts traded in this market, expiration date, strike price, the volume traded (number of contracts), delta and implicit volatility. For an earnings announcement to be included in the final sample, we imposed several conditions: · We have selected those firms listed in the SIBE for which there are optioned stocks and for which stock market data was available in the period that comprises 50 days before the event day and 5 days after it. Estudios de Economía, Vol.43 - Nº2240 TABLE 1 SUMMARY STATISTICS FOR THE SPANISH OPTIONS MARKET The table exhibits for the period 2004-2012 some characteristics of the Spanish stock options market: the number of contracts outstanding each year, the number of contracts that were traded at least one day and, for them, the daily average mean and standard deviation for the trading volume, the open interest and the number of days traded. CALL OPTIONS 2004 2005 2006 2007 2008 2009 2010 2011 2012 Number of contracts 5,873 5,018 4,373 6,846 21,932 38,162 24,260 39,921 47,642 Number of contracts traded at least one day 1,899 1,422 1,254 1,692 2,016 4,381 4,634 3,465 4,061 Trading volume Mean 24.62 44.59 49.87 39.35 51.84 91.31 47.63 47.78 47.22 St.D. 69.48 179.24 154.30 147.59 220.47 570.58 208.94 199.19 227.90 Open interest Mean 900.60 1581.46 1963.58 1575.63 1981.73 2399.84 1831.80 2,034.50 2,093.39 St.D. 2,624.89 5,829.46 6,259.09 5,058.89 6,182.66 8,689.13 7,579.93 7,842.83 8,884.56 Days traded Mean 13.60 11.07 11.19 10.19 8.52 7.33 8.34 6.56 1.05 St.D. 16.63 14.66 16.67 15.68 11.43 11.68 11.77 9.51 3.66 PUT OPTIONS 2004 2005 2006 2007 2008 2009 2010 2011 2012 Number of contracts 5,873 5,018 4,373 6,846 21,920 38,173 24,274 39,952 47,598 Number of contracts traded at least one day 2,042 1,575 1,321 1,778 2,097 4,930 5,067 3,704 4,370 Trading volume Mean 19.04 27.72 38.03 40.91 54.26 81.99 49.44 52.94 53.58 St.D. 54.00 82.38 116.42 153.69 191.49 564.74 215.08 218.89 336.02 Open interest Mean 727.90 1,249.56 2,017.90 1,915.98 2,394.95 2,415.40 1,881.37 2,224.75 2,216.91 St.D. 2,415.39 3,399.98 6,746.95 8,053.13 9,621.37 9,974.04 7,495.86 7,848.68 9,552.03 Days traded Mean 12.62 8.41 8.62 9.00 9.70 7.24 8.85 7.44 1.04 St.D. 15.42 11.94 13.17 14.49 12.42 12.47 13.89 12.44 3.91 The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 247 trading activity as optioned stocks catch their attention, because the introduction of options trading implies more intensive information collection around earnings announcements. To conclude, the availability of options markets does not reflect improvements in liquidity around earnings announcements. 5. Reaction of Trading Activity in Stock Options to the Release of Earnings Announcements 5.1. Objective This section examines the trading activity in the stock options market before the release of quarterly and annual earnings announcements (including the announcement period). Trading volume has been used to determine whether an event has informational content and its study applies not only to trades on common stocks but applies equally well to stock options whose prices are determined by the value of the firm. To address this issue, we have analysed the trading volume in the options markets both in absolute terms and, following Roll etal. (2010), in relative terms related to trading volume in their underlying assets to control for stock volume. Although trading volume in the options market may be informative about the price discovery process in financial markets, the volume by itself does not indicate the direction of the transaction. In this regard, changes in open interest may provide additional evidence with respect to the informational role of options. Open interest measures the total number of options contracts that are currently open, in other words, contracts that have been traded but not yet liquidated by either an offsetting trade or an exercise or assignment. 5.2. Sample, data and methodology To illustrate how investors react to news events, we take a sample of 514 announcements from the first quarter of 2006 to the last quarter of 2011. In order to maintain the sample homogeneity taking only American options, we have reduced the analysis period from the first analysis because American options began trading in Spain in 2006. Moreover, we restrict our analysis to the options series with a time to maturity between 70 and 10 calendar days during the complete event period and focus our attention on three variables to estimate the trading activity: · Options trading volume (O) in number of contracts. We calculate the total daily number of contracts traded for each stock by adding the contracts traded across all options listed on the stock. · The options/stocks trading volume ratio (O/S). This is the ratio for a given day between the trading options volume for each firm (calculated by multiplying the total contracts traded in each option by 100 shares and aggregating across all options listed on the stock) and the corresponding trading volume (number of shares) on the stock market of a given firm. · The open interest (OI), measured as the daily variation of the number of contracts that remain open, aggregated across all options listed on the stock. Estudios de Economía, Vol.43 - Nº2248 Besides the information released, there are other factors that may explain part of the surge in trading activity that follows an accounting earnings announcement. Market participants may be trading based on volatility or may have placed hedges ahead of the earnings announcements, which might be unnecessary after the release, leading to additional trading activity. To pick up these effects, we also include other variables for which we have available data, namely implied volatility (V), delta (∆) and firm size (S) , measured as the capitalization of the firm, which could explain the activity in the options market.7 Table5 presents summary statistics for the dependent and explanatory variables. A daily cross-sectional mean is computed for each trading day in the sample and then mean, median, standard deviation, maximum and minimum are computed from the daily means across all the trading days in the sample. The daily means and medians are close, showing that they are not outliers. The O/S indicates that on average the activity in the stock market is bigger than the activity in the options market. The regression model we propose to explain the behaviour of options market trading activity is presented in the expression [3]: [3] ln TAit ( ) =C+ β 1Dit + β 2Vit + β 3∆it + β 4ln Sit ( ) +eit where TA refers to the measures of trading activity above mentioned and D is a dummy variable that takes a value of 1 in the days of the window -5 to 1 with respect to the earnings announcement and 0 otherwise. We try to analyse whether in the five days before an earnings announcement, the announcement day and the day after there is additional informed options trading volume. Clearly, if the earnings announcement has informational content, we expect a positive relationship between trading activity and the dummy variable. Finally, to reduce the influence of possible outliers, we use the natural logarithm of the 7 Implied volatility (V) and delta (∆) have been taken from the information supplied by the Spanish Options Market (MEFF). TABLE 5 DESCRIPTIVE STATISTICS OF INDEPENDENT AND DEPENDENT VARIABLES The table shows some statistics for size (S) measured as the capitalization of the firm, the intrinsic volatility (V), the delta (∆) and options trading activity variables, i.e. the options trading volume in number of contracts (O), the options/stocks trading volume ratio in number of shares (O/S) and the open interest (OI). A daily cross-sectional mean is computed for each trading day in the sample, and then the mean, median, standard deviation, maximum and minimum are computed from the daily means. Size Volatility Delta Ln(O/S) Ln(O) Ln(OI) Mean 9.5396 0.3328 0.4403 –11.0500 4.4570 12.4382 Median 9.5513 0.3095 0.4349 –6.3753 4.3820 12.4388 SD 0.3890 0.1012 0.0636 0.7369 0.7743 0.0392 Maximum 10.9392 0.7729 0.7182 –4.3342 7.3837 12.5020 Minimum 8.6993 0.1797 0.2598 –9.2964 2.1171 11.0573 The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 249 TABLE 6 AVERAGE CORRELATIONS OF INDEPENDENT AND DEPENDENT VARIABLES The table shows the average correlations of explanatory and dependent variables. Correlations are computed for each trading day in the sample among all dependent variables: the options trading activity variables, i.e. the options trading volume in number of contracts (O), the options/stocks trading volume ratio in number of shares (O/S) and the open interest (OI), and the explanatory variables, i.e. the size (S) measured as the capitalization of the firm, the intrinsic volatility (V), the delta (∆) and the earnings date (D). The correlations are then averaged across all trading days in the sample. Dummy Volatility Delta Ln(O/S) Ln(O) Ln(OI) Size –0.0040 –0.1491 0.0408 0.1318 0.5551 0.2171 Dummy 0.0288 –0.0201 0.0204 0.0118 0.0022 Volatility –0.0358 –0.0396 –0.0396 –0.0052 Delta –0.0143 0.1030 –0.0275 Ln(O/S) 0.6702 0.2738 Ln(O) 0.3558 trading volume variables as the dependent variables in the results presented. For convenience, we will often refer to the logged variables as simply option volume (O), options/stock trading volume ratio (O/S), open interest (OI) and size (S). Following the rule of the previous section, we have estimated the models using both the full sample of announcements (denoted aggregate sample) and the existing sample after removing all the announcements referring to financial companies (reduced sample). Table6 shows the average correlations of explanatory and dependent variables employed in the regression model. Correlations are computed for each trading day in the sample among all dependent variables. The correlations are then averaged across all trading days in the sample. The correlations between the independent variables and the explanatory variables show modest values. The independent variable with the highest correlation with the dependent variables is the size of the firm. With respect to the correlation between the dependent variables, the values are low. In addition to using all the options contracts available every day, we also study separately OTM, ATM and ITM options. As Easley etal. (1998), Chakravarty etal. (2004) and Chen etal. (2005) among others, point out, the options with different degrees of moneyness have different characteristics that attract different agents. So we can consider that informed traders would prefer to trade OTM options since they are cheaper and have a higher degree of leverage (Roll etal., 2010). Investors trading on volatility would prefer ATM options because the spread tends to be lower and they offer high liquidity (Billings and Jennings, 2011) We compute an option’s moneyness by comparing the option’s strike price to the underlying asset price. Specifically, we calculate the price ratio for options as the strike price divided by the underlying asset price. We define ATM options as when the price ratio falls between 0.925 and 1.075, ITM options as when the price ratio is lower than 0.925 for call options and greater than 1.075 for put options, and OTM options as when the price ratio is greater than 1.075 for call options or lower than 0.925 for put options. The ratio values were chosen according to the existing literature (Chen etal. 2005; Billings and Jennings, 2011) Estudios de Economía, Vol.43 - Nº2250 The availability of information allows for the use of the double temporal and cross-sectional dimension of the sample through an econometric model of panel data. This provides us, compared to just cross-sectional databases, with a more informative dataset (with more variability, less collinearity and more degrees of freedom) that permits us to attain more efficient estimates in an econometric linear regression model and to control for individual unobserved heterogeneity (one of the major problems in non-experimental research). We use an unbalanced panel data set with the daily data from 33 firms and 514 earnings announcements from the first quarter of 2006 to the last quarter of 2011. 5.3. Results Table7 shows the coefficients obtained after applying the best estimate procedure. To this end, a Breusch-Pagan test was applied which showed that the ordinary least square estimation was not the appropriate model and, after that, a Hausman test was carried out to determine which model, the fixed or random effects, was more appropriate.8 The result for the variable object of our interest, the dummy variable, is positive and highly significant in the options trading volume and options/stock volume ratio regression. Moreover, this behaviour is repeated, except for ITM options in the relative measure. This implies an increase in options trading activity in the analysis period. Agents trade in the options market in anticipation of the earnings announcement to profit from their knowledge of the unanticipated earnings surprise. These results are in line with previous literature (Philbrick and Stephan, 1993; Roll etal., 2010, among others) and confirm our hypothesis that, with the arrival of new information, investors trade in the options market. Additionally, the reaction in the options market is larger than in the stock market for OTM and ATM options. Of special interest for our study is the open interest variable. The results show, as in Schachter (1988), a significant decrease for the full sample. Considering the degree of moneyness, this result is only supported for ITM options. However, for ATM and OTM options the results show a significant increase, as in Amin and Lee (1997) and Donders etal. (2000). This result indicates that trading in the options market in the days prior to the earnings announcement is an offsetting trade in ITM options, and that it is in ATM and OTM options where investors decide to trade to take advantage of their information by opening new contracts. For the reduced sample, this result is observed only for ATM options. 8 The results in Table7 show that panel data works more appropriate than pooled OLS. Specifically, the fixed effects model is the best model. The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 251 TABLE7 PANEL DATA REGRESSIONS OF TRADING ACTIVITY VARIABLES ON THEIR DETERMINANTS The table exhibits the results of the regressions of options trading activity variables on their proximate determinants for a sample of 514 earnings announcements for Spanish stocks with listed options from the first quarter of 2006 to the last quarter of 2011. The options trading activity variables are the options trading volume in number of contracts (Oit), the options/stocks trading volume ratio in number of shares ((O/S)it) and the open interest (OIit). The model regressions are the following: ln TAit ( ) =C+ β 1Dit + β 2Vit + β 3∆it + β 4ln Sit ( ) +eit where: TAit refers to measures of trading activity, Dit is a dummy variable that reflects the earnings date and takes a value of 1 on trading dates -5 to +1 relative to the announcement day and 0 otherwise, Vit is the intrinsic volatility, ∆it is the delta, and Sit is the size of the firm i, measured as the capitalization on day t. The results are presented for all the options contracts and separately for out of the money (OTM), at the money (ATM) and in the money (ITM) options contracts. ALL CONTRACTS ITM ATM OTM Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Aggregate sample C 0.5677 –11.8882*12.4353*0.8827 –7.7951*12.5284*2.2613 –9.8862*10.3490*1.3188 –10.9933*9.8238* Dummy (D) 0.3182*0.2052*–0.0068*0.2680*0.0956 –0.0202** 0.3270*0.2039*0.0057** 0.2337*0.1066*** 0.0071*** Volatility (V) 0.9478** 0.3415 0.0133** 2.1825*1.0805*–0.0087** –0.7353 –1.4613*0.0331*2.2231*1.7237*0.0945* Delta (∆) 0.4126** 0.0994 –0.0153 3.0447*3.0080*–0.0521 0.3465 –0.1864 –0.0161** –0.4504 –0.7693 0.0451*** Size (S) 0.3444 0.5631*0.0068 –0.0797 –0.3159 –0.0034 0.1839 0.3573** 0.0047 0.1749 0.3305** 0.0057 F Statistic Prob > F 8.44 0.0001 6.52 0.0006 7.79 0.0004 16.55 0.000 25.07 0.0000 8.62 0.0001 11.77 0.0000 22.08 0.0000 9.81 0.0000 44.19 0.0000 12.31 0.0000 5.48 0.0001 F test (all ui=0) Prob > F 168.99 0.0000 43.61 0.0000 29.79 0.0000 10.77 0.0000 12.06 0.0000 10.01 0.0000 100.21 0.0000 26.01 0.0000 4.65 0.0000 58.35 0.0000 17.35 0.0000 7.34 0.0000 Hausman Statistic Prob (Hausman) 16.01 0.0010 22.16 0.0002 35,52 0.0000 26.60 0.0000 11.01 0.0011 10.25 0.0001 21.55 0.0002 15.04 0.0046 25.24 0.0000 13.85 0.0018 874.21 0.0000 13.97 0.0010 R-squared 0.5017 0.0928 0.0240 0.1649 0.1330 0.0400 0.4314 0.0873 0.0374 0.3286 0.0652 0.0360 Estudios de Economía, Vol.43 - Nº2252 ALL CONTRACTS ITM ATM OTM Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Ln(O) Ln(O/S) Ln(OI) Observations 18,397 18,395 18,397 7,021 7,021 7,021 14,800 14,800 14,800 11,462 11,461 11,462 Groups 33 33 33 32 32 32 33 33 33 33 33 33 Reduced sample C 2.5594 –9.1587*12.4529*–3.9739 –16.4789*12.5484*3.7510 –7.6480*10.3524*2.2980 –9.7094*9.8044* Dummy (D) 0.2566*0.1787*–0.0099*0.2346*** 0.1297 –0.0308** 0.2942*0.2145*0.0030** 0.1341** 0.0409 0.0017*** Volatility (V) 0.4909 0.0711 0.0138 1.8099*1.5433*–0.0081 –1.1330** –1.8901*0.0407 1.8654*1.5029** 0.0763** Delta (∆) 0.3811*** 0.0456 –0.01715*3.0138*2.9469*–0.0581 0.5011*** –0.0897 –0.0163*** –0.7212 –0.8838 0.0499 Size (S) 0.1070 0.2967** –0.0011 0.4000*** 0.5713***–0.0047 –0.0036 0.1468 0.0042 0.0570 .02402 0.0084 F Statistic Prob > F 4.58 0.0060 4.67 0.0060 5.71 0.0001 7.78 0.0003 12.92 0.0000 5.14 0.0056 5.10 0.0034 43.52 0.0000 8.20 0.0002 16.74 0.0000 4.85 0.0044 5.12 0.0031 F test (all ui=0) Prob > F 169.57 0.0000 40.21 0.0000 52.32 0.0000 8.61 0.0000 13.49 0.0000 14.49 0.0000 104.97 0.0000 20.53 0.0000 5.68 0.0000 61.64 0.0000 6.66 0.0000 8.47 0.0000 Hausman Statistic Prob (Hausman) 22.34 0.0010 18.59 0.0072 34.67 0.0000 18.28 0.0082 47.65 0.0014 17.65 0.0014 22.28 0.0002 6.30 0.0029 29.77 0.0000 11.06 0.0025 14.00 0.0073 30.71 0.0000 R-squared 0.4608 0.0985 0.0240 0.1642 0.1382 0.058 0.4194 0.073 0.051 0.3047 0.0300 0.0547 Observations 13,788 13,786 13,786 4,297 4,297 4,297 10,792 10,792 10,792 8,244 8,244 8,244 Groups 28 28 28 27 27 27 28 28 28 28 28 28 *,**,*** Significantly different from zero al the 1%, 5% and 10% level, respectively. The command vce(cluster name) in STATA, which specifies that the standard errors allow for intragroup correlation, relaxing the usual requirement that the observations be independent, has been used. Table 7 (continuation) The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 253 6. Cumulative Abnormal Returns Around Earnings Announcements and Options Trading Activity 6.1. Objective As our results confirm that investors consider the options market as an alternative place to make their investments, this section assesses whether the increase in trading activity before earnings announcements is due to increased trading in options by informed agents attempting to profit from their views about the unanticipated earnings surprise. 6.2. Sample, data and methodology To address this issue we relate the value of post-CAR, calculated as the cumulative abnormal return on days zero through five after the announcement day, with the trading activity variables defined in the previous section. In the light of the previous results and following Roll etal. (2010), the relation could depend on the size of the pre-CAR because the activity of informed traders before the announcement could anticipate the incorporation of new information. Thus, larger pre-CAR would imply less informative post-CAR. To pick up this effect we include pre-CAR, defined as the cumulative abnormal return from five to two days before the announcement date, as an explanatory variable. Previous empirical literature (Atiase, 1985, 1987; Pope and Inyangete, 1992; García, etal., 2012, among others) shows that the level of previous information depends on the firm size, so for this reason we include firm size as an explanatory variable. Since earnings surprise can be either good news or bad news and since options volume is unavailable, we examine absolute preand post-CAR in 514 earnings announcements j by all firms i with listed options from the first quarter of 2006 to the last quarter of 2011. The regression has the following form: [6] Abs CAR0,5 ( ) ij =C+ β 1ln TA−5,−2 ( ) ij + β 2Abs CAR−5,−2 ( ) ij +ln S−5,−2 ( ) ij +uij The model relates the absolute value of the post-CAR to the options trading activity variables (TA), that is, options trading volume, the options/stock volume ratio or the open interest, all of them averaged over the pre-announcement window (days -5 to -2) and with the absolute value of pre-CAR and with size (S) measured as the market capitalization averaged over the pre-announcement window.9 As in the previous sections, we have estimated the models using both the aggregate sample and the reduced one. 9 Similar to the previous section, to reduce the influence of possible outliers we use the natural logarithm of the trading volume variables and size. Estudios de Economía, Vol.43 - Nº2254 6.3. Results The results, presented in Table8, show that options trading volume and open interest are positive and significant in the preannouncement period and that the options/stock trading volume ratio is not significant in the same period. This indicates that in those announcements where the impact is greater, there has been more trading in options prior to an earnings release, which is reflected in a greater number of open positions. For the reduced sample, only open interest is significant. This result confirms the significance of this variable as a proxy for the trading activity in the options market. The results support the notion that some of the agents that trade actively in the options markets prior to an earnings announcement are informed. Firm size has a negative and significant impact on prices, indicating, as expected, that the impact of the announcement is bigger on small firms for which there is less public information before the release and it is more difficult to anticipate the new information. The pre-CAR variable is positive and significant indicating that more impact on prices before the earnings announcement is associated with more impact on prices after the release. This result is not as expected and could be affected by the fact that the returns are in absolute terms regardless of the direction of the prices. To address this issue, we also run regressions through the aggregated and reduced sample of signed post-CAR on volume variables for two separate cases, one for 232 (175) announcements with positive post-CAR (good news) and one for 282 (228) with negative post-CAR (bad news). Moreover, to consider if this relation is affected by the impact of the earnings announcement on prices, we use the technique of quantile regression of Koenker and Bassett (1978).10 10 Quantile regression considers some of the typical statistical problems of financial series, such as sensitivity to outliers and heteroskedasticity, among others. TABLE8 CUMULATIVE ABNORMAL RETURN ON OPTIONS TRADING ACTIVITY The table relates the absolute value of the post-CAR (from zero to five days after the earnings announcement) to the options trading activity (TA) variables (options volume, options/stock volume ratio and open interest) accumulated over the pre-announcement windows (days -5 to -2) and with the absolute value of pre-CAR (from five to two days prior to the announcement) and the size of the firm. Heteroskedasticity is accounted for through White robust standard errors. Abs CAR0,5 ( ) ij =C+ β 1ln TA−5,−2 ( ) ij + β 2Abs CAR−5,−2 ( ) ij +ln S−5,−2 ( ) ij +uij Aggregate sample Reduced sample C 0.0756*0.0705*0.0269 0.0917*0.0860*0.0687* Trading activity Ln(O-5,-2) 0.0012*** –0.0003 Ln((O/S)-5,-2) 0.0005 –0.0006 Ln(OI-5,-2) 0.0028*0.0019* Pre-CAR (CAR-5,-2) 0.4319*0.4395*0.4409*0.2878*0.2883*0.2865* Size (Ln(S) -5,-2) –0.0053*–0.0037** –0.0033*** –0.0055** –0.0059*–0.0062* *,**,*** Significantly different from zero at the 1%, 5% and 10% level, respectively. The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 255 TABLE9 SIGNED CUMULATIVE ABNORMAL RETURN ON OPTIONS TRADING ACTIVITY The table relates the value of the signed post-CAR (from zero to five days after the earnings announcement) to the options trading activity (TA) variables (option volume, option/stock volume ratio and open interest) accumulated over the pre-announcement windows (days -5 to -2) and to the value of pre-CAR (from five to two days prior to the announcement) and the size of the firm. The table shows quantile regression results in comparison with OLS results. CAR0,5 ( ) ij =C+ β 1ln TA−5,−2 ( ) ij + β 2CAR−5,−2 ( ) ij + β 3ln S−5,−2 ( ) ij +vij Aggregate sample OLS 0;25 Quantile regressiona0;50 Quantile regressiona0;75 Quantile regressiona O O/S OI O O/S OI O O/S OI O O/S OI PANEL A CAR0,5>0 N=232 C 0.0831*0.0761*0.0183 0.0423** 0.0500*0.0437 0.0672*0.0826*0.0039 0.1016*0.1142*0.0314 Ln(TA-5,-2) 0.0030** 0.0018 0.0025 –0.0001 0.0006 –0.0001 0.0021 0.0020 0.0021 0.0025*** 0.0031 0.0035* Ln(CAR-5,-2) –0.1980*–0.2079*–0.2123*0.0386 0.0520 0.0392 0.0034 0.0061 0.0269 –0.0665 –0.1062 –0.0823 Ln(S-5,-2) –0.0062** –0.0023 –0.0009 –0.0031 –0.0036** –0.0032 –0.0049 –0.0039 –0.0003 –0.0587 –0.0034 –0.0018 R-squared 0.0851 0.0726 0.0688 0.0088 0.0084 0.0091 0.0250 0.0058 0.0114 0.0703 0.0574 0.05138 PANEL B CAR0,5<0 N=282 C –0.1317*–0.1214*0.6684 –0.2040*–0.2304*0.5989 –0.0817*–0.0808*–0.1582 –0.0369*–0.0248 –0.0933 Ln(TA-5,-2) –0.0010 0.0001 –0.0618 –0.0028 –0.0035 –0.0605 –0.0001 0.0002 –0.0187 –0.0004 0.0013 0.0045 Ln(CAR-5,-2) 0.0352 0.0322 0.0357 0.1432 0.0322 0.0842 –0.0015 –0.0041 –00189 0.02951 0.0015 0.0217 Ln(S-5,-2) 0.0092*0.0076*0.0079*0.0161*0.0148*0.0115*0.0045 0.0026** 0.0044** 0.0020 0.0013 0.0015 R-squared 0.0379 0.0358 0.0377 0.0343 0.0295 0.0369 0.0348 0.0344 0.0359 0.0261 0.0186 0.0299 Estudios de Economía, Vol.43 - Nº2256 Reduced sample OLS 0;25 Quantile regressiona0;50 Quantile regressiona0;75 Quantile regressiona O O/S OI O O O/S OI O O O/S OI OI PANEL A CAR0,5>0 N=175 C 0.0863*0.0854*0.0607 0.0487** 0.0486** 0.0568** 0.0658*0.0478 0.0530 0.1035** 0.1015** 0.0577 Ln(TA-5,-2) –0.0001 –0.0001 0.0018 –0.0007 –0.0005 –0.0001 –0.0020 –0.0020 0.0012 –0.00084 –0.0004 0.0033** Ln(CAR-5,-2) 0.0690 0.0689 –0.0699 0.0641 0.0647 0.0632 0.0466 0.0449 0.0587 0.0794 0.0896 0.0798 Ln(S-5,-2) –0.0050*** –0.0051** –0.0049*–0.0033 –0.0040*** –0.0044** –0.0028** –0.0033 –0.0042 –0.0045 –0.0051 –0.0045 R-squared 0.0369 0.0369 0.0396 0.0353 0.0361 0.0362 0.0291 0.0264 0.0309 0.0357 0.0363 0.0381 PANEL B CAR0,5<0 N=228 C –0.1348*–0.1267*–0.2415 –0.1923*–0.1859*0.0353 –0.0763*–0.0748*0.0709 –0.0482*–0.0351** –0.3257 Ln(TA-5,-2) 0.0001 0.0007 0.0082 0.0001 0.0003 –0.0180 –0.0001 0.0004 –0.0114 0.0002 0.0009 0.0217 Ln(CAR-5,-2) –0.0165 –0.0179 –0.0167 –0.0696 –0.0751 –0.0685 –0.0456 –0.0330 –0.0432 0.0036 0.0047 0.0025 Ln(S-5,-2) 0.0092*0.0088*0.0093*0.0137*0.0134*0.0142*0.0042 0.0042 0.0041*** 0.0031 0.0025 0.0031*** R-squared 0.0440 0.0446 0.0440 0.0432 0.4347 0.0431 0.0389 0.0428 0.0391 0.0425 0.0393 0.0418 *,**,*** Significantly different from zero at the 1%, 5% and 10% level, respectively (a) The command qreg2 in STATA, which estimates quantile regression and reports standard errors and t-statistics that are asymptotically valid under heteroskedasticity or under heteroskedasticity and intra-cluster correlation, has been used. Table 9 (continuation) The informational role of thin options… / C. J. García, B. Herrero, A. M. Ibáñez 263 COMPANY NAME (TICKER) 2004 2005 2006 2007 2008 2009 2010 2011 2012 GAS NATURAL (GAS) 274 (105) 410 (219) 2822 (201) 8578 (527) 1688 (252) 1671 (185) 1994 (265) GRIFOLS (GRF) 678 (25) 1104 (151) 976 (207) 1600 (71) 2858 (150) IAG (IAG) 1506 (0) 1310 (9) IBERDROLA (IBE) 559 (194) 716 (153) 1130 (207) 2372 (289) 2632 (287) 3582 (559) 2068 (512) 2167 (384) 2895 (620) INDRA (IDR) 465 (129) 484 (147) 298 (134) 380 (200) 704 (137) 1376 (256) 878 (211) 1856 (144) 2246 (174) INDITEX (ITX) 537 (224) 384 (130) 452 (200) 396 (196) 944 (244) 1731 (442) 1891 (584) 3188 (420) 4743 (514) MAPFRE (MAP) 268 (11) 698 (31) 1304 (154) 1096 (159) 1560 (78) 1650 (45) ARCELORMITTAL (MTS) 1382 (86) 1044 (149) 2416 (179) 1734 (194) METROVACESA (MVC) 68 (1) NH HOTELES (NHH) 222 (5) 1098 (0) 2400 (1) 1340 (9) 3324 (34) 1676 (6) OBRASCÓN HUARTE (OHL) 1526 (3) 2014 (58) BANCO POPULAR (POP) 447 (200) 512 (164) 436 (137) 334 (181) 2988 (211) 4769 (460) 2288 (507) 2369 (403) 4525 (279) RED ELECTRICA (REE) 272 (142) 742 (145) 1268 (226) 1020 (232) 1870 (205) 1740 (195) REPSOL (REP) 583 (245) 908 (223) 874 (241) 780 (240) 2890 (310) 4209 (675) 1908 (589) 1917 (557) 2690 (750) BANCO SABADELL (SAB) 342 (170) 774 (150) 1860 (351) 1260 (209) 1654 (78) 2562 (40) BANCO SANTANDER (SAN) 841 (302) 744 (264) 934 (280) 854 (286) 6085 (565) 5780 (1084) 2486 (1087) 2585 (1011) 2645 (1062) SACYR VALLEHERMOSO (SCYR) 536 (60) 1228 (37) 1844 (4) 2028 (35) 3718 (42) 4163 (77) SOGECABLE (SGC) 657 (225) 464 (183) TELEFONICA (TEF) 595 (265) 1098 (303) 898 (297) 1004 (367) 2357 (392) 3757 (914) 2065 (915) 1999 (910) 2225 (1016) TELEFONICA MOVILES (TEM) 547 (151) 300 (80) MEDIASET (TL5) 214 (13) 943 (30) 1970 (231) 2152 (319) 3542 (149) 2444 (50) TPI (TPI) 630 (172) 310 (102) TÉCNICAS REUNIDAS (TRE) 1406 (6) 1768 (102) TERRA (TRR) 707 (135) 356 (22) UNION FENOSA (UNF) 501 (156) 332 (128)