The determinants of betting volume for sports in North America: evidence of sports betting as consumption in the NBA and NHL.

Paul, Rodney J. ; Weinbach, Andrew P.

Introduction

The analysis of sports gambling markets, either as simple financial markets or as a study of the behavior of its participants, has been limited due to the lack of information on key sportsbook data such as actual percentages of wagers on each side of the proposition and overall betting volume. Recently, however, more detailed sports betting market data have started to become available. Betting percentage data, noting the percentage of dollars bet on each side of a wagering proposition (favorites and underdogs; overs and unders), has become available on-line through the individual sportsbook, http://www.sportsbook.com, and as an aggregate of three on-line sportsbooks through a service offered by Sports Insights (http://www.sportsinsights.com).

Data from these sources have been used to support alternatives to the balanced book assumption commonly used as a benchmark in analysis of betting markets, as suggested in early models of sportsbook behavior such as Pankoff (1968), Zuber et. al. (1985), and Sauer et. al. (1988). Betting percentages were shown to increase with each point of the pointspread in "sides" (wagering on game outcomes against a pointspread) markets and with each point of the totals in "over/under" markets in both the NFL (Paul & Weinbach, 2008a) and in the NBA (Paul & Weinbach, 2008b). These studies provide evidence of systematic imbalances and shed some light on the preferences of bettors for big favorites and road favorites (the best teams) and for the over (scoring). Such imbalances are consistent with the notion that recreational bettors, rather than informed traders, may be the dominant group in these markets.

Although historical betting volume data for earlier seasons is not available for purchase from Sports Insights, a form of this data is available in the daily premium service offered by http://www.sportsinsights.com. At each moment of the day and after the close of betting (available for approximately 24 hours after the game is played), Sports Insights lists information on the "Number of Bets" for each game offered on the sports betting rotation. The Number of Bets represents the total number of bets made on a particular game from three on-line sportsbooks (BetUS.com, CaribSports.com, and SportBet.com), with combined betting volume approaching 10,000 bets per game for NBA and 5,000 bets per NHL game. Although this figure does not represent the dollar volume on the game, (1) as wagers likely vary in size, the number of bets on a game allows us to analyze the relationship between game characteristics and betting activity. The data were collected based on a systematic schedule to build a complete and consistent dataset for the 2008-09 regular season. (2) We wish to note these data contain a single season for the NBA and NHL, which we believe will shed insight on how bettors behave, but may still be too short a time frame to capture all of the intricacies of bettor behavior. This limitation of relatively short time frames studied was noted in relation to market efficiency studies (Osborne, 2001) and in attendance studies for English Soccer (Buraimo & Simmons, 2008).

Given the availability of betting volume data for large, international sportsbooks, a study of the factors which contribute to differences in betting volume across different games is possible. If bettors view wagers on sports contests as investments and use careful analysis of the matchups to identify those wagers with the greatest expected value, it may be difficult to identify patterns of betting behavior, since easily identifiable patterns are expected to attract entrants and drive out profit opportunities. If, however, these betting markets are influenced by utility maximizing agents who consider wagering to be a consumption activity, the game characteristics associated with increased betting activity are likely to be recognizable and similar to those characteristics associated with increased attendance or higher television ratings.

Our rationale for this point of view dates back to Samuelson (1952), who commented that basing the value of a gamble solely on the monetary prize is likely incorrect, indicating "When I go to a casino, I go not alone for the dollar prizes, but also for the pleasures of gaming--for the soft lights and the sweet music." In the case of sports wagering, bettors might find certain propositions and games more desirable than others. If all, most, or even some bettors wager partially for the purpose of consumption, taking a financial stake in the game to make watching the game and following certain teams more exciting, betting volume should vary with factors fans find enjoyable.

It is important to note here that the goal of this paper is not to develop a new model of gambling behavior. Conlisk (1993) develops a "Small Gamble Theorem" suggesting that individuals obtain utility from gambling on small bets, and this utility is not inconsistent with observed risk-aversion (such as the purchase of fire insurance) when the stakes are high. Terrell and Farmer (1996) provide a formal model of informed bettors and noise traders simultaneously existing in greyhound racing, which could be amended in relation to sports betting markets. A formal model of utility for sports bettors may be a productive avenue for research, and the authors suggest anyone interested in this topic start with the comprehensive review of the gambling literature by Sauer (1998), but the goal of this research is empirical and applied in nature. If consumption plays a significant or even dominant role in the actions of these sports bettors, identifying the factors that sports bettors find attractive is potentially very useful information.

With this in mind, we construct a simple regression model with sports betting volume aggregated across three major on-line sportsbooks as the dependent variable and test factors such as television coverage, uncertainty of outcome (measured by the pointspread), scoring (measured by the posted total), and quality of teams (win/loss record coupled with the pointspread), for their influence on the number of bets on each sporting event. If these factors are not significant determinants, then the role of consumption in sports gambling may be unimportant or too inconsistent to be useful. However, if these factors are significant determinants of betting activity, these independent variables can help further our understanding of the actions of sportsbooks and the behavior of bettors.

Betting Volume Regression for the NBA and the NHL

The data set from Sports Insights includes betting information for all games played during the NBA and NHL seasons. When considering the availability of data on variables we wished to study, we ultimately were able to compile a full data set of volume of bets, game information, and television information for the NBA and NHL, for the 2008-09 regular season. Summary statistics for the non-binary variables used in this study appear in Appendix I of the paper. The regression model used for the NBA is noted in equation 1 below. The regression model used for the NHL is noted in equation 2. The variables are described in the paragraphs following the equation.

Betting Volume [NBA.sub.i] = [[beta].sub.o] + [[beta].sub.1] (Road Favorite [Dummy.sub.i]) + [[beta].sub.2] ([Pointspread.sub.i]) + [[beta].sub.3]([Pointspread.sub.i.sup.2]) + [[beta].sub.4]([Total.sub.i]) + [[beta].sub.5](Sum of Win [Pct.sub.i]) + [summation] [[beta].sub.i](TV Network Dummies) + [summation][[beta].sub.i](Day of Week Dummies) + [summation][[beta].sub.i](Monthly Dummies) + [[epsilon].sub.i] (1)

Betting Volume [NHL.sub.i] = [[beta].sub.o] + [[beta].sub.1](Road Favorite [Dummy.sub.i]) + [[beta].sub.2]([Odds.sub.i]) + [[beta].sub.3]([Odds.sub.i.sup.2]) + [[beta].sub.4]([Total.sub.i]) + [[beta].sub.5](Sum of Win [Pct.sub.i]) + [summation][[beta].sub.i](TV Network Dummies) + [summation][[beta].sub.i](Day of Week Dummies) + [summation][[beta].sub.i](Monthly Dummies) + [[epsilon].sub.i] (2)

The dependent variable in the regressions is betting volume in terms of the number of bets on each game. To model this variable we have considered classes of variables which we consider to be important to fans. If consumption is important to wagering activity, fan-favorite characteristics such as television coverage, quality teams, timing of the game, etc. would likely have a significant effect on betting volume. Fan sentiment has been shown to be important in the betting market for Spanish soccer (Forrest & Simmons, 2008) and may prove to be quite important in other sports as well.

We attempted to model each league in the same general fashion, allowing for the differences in the sports (or in the case of the NHL--a difference in the betting market itself--odds as opposed to pointspreads) where necessary. We believe that fans of team sports, such as basketball and hockey, are similar in terms of their preferences for high-quality and exciting games. Therefore, we believe fans enjoy uncertainty of outcome, the best teams, and the opportunity to watch the game on television. Although we believe fans are similar across these sports, ultimately, this is an empirical question which can be tested through the available data. Gambling-related data was gathered from Sports Insights itself, while data on television came from the websites of the leagues (NBA.com, NHL.com).

The regression model specification for each sport includes an intercept and has other variables listed by appropriate category. The first category includes the gambling data available on the game. First, there is a dummy variable indicating whether a team is a road favorite. Earlier research suggests that strategies of wagering on home underdogs (against road favorites) could outperform the hypothesized returns from fair bets, and even earn positive returns (Levitt, 2004; Gray & Gray, 1997; Golec & Tamarkin, 1991). In addition, higher betting percentages on favorites were found for road favorites (Paul & Weinbach, 2008b). The inclusion of a dummy for road favorites will allow us to determine if games with road favorites attract more wagers than games with home favorites.

The pointspread (favorite odds in the case of the NHL) is included in the regression to account for the possibility that bettors prefer to wager on games with bigger favorites. The pointspread (odds) is the closing pointspread for BetUS, which handles the largest volume of the three sportsbooks. The percentage of bets was shown to increase with greater pointspreads on favorites for the NFL and NBA (Paul & Weinbach, 2008a, 2008b). In the NHL, a reverse favorite-longshot bias has been found in previous studies (Woodland & Woodland, 2001; Gandar et al., 2004) in that underdogs win more than implied by efficiency. If games involving higher pointspreads or odds(typically representing one good team playing in the game--who may be highly recognizable to fans and the betting public) attract a greater volume of bets, this variable should have a positive and significant effect on the volume of bets on the game. For the NBA, there is evidence fans prefer their home team to be favorites as Rascher and Solmes (2007) found that attendance is maximized when the home team is favored by 67% over the visiting team. To allow for the possibility that bettors like good teams, but still value uncertainty of outcome, a square of the pointspread (odds in the NHL) is also included in the regression model. If the squared pointspread term is found to be negative, the result would suggest that bettors, like fans, prefer uncertainty of outcome in sporting contests.

The total is also included in the regression to account for the possibility that bettors prefer to wager on games that are expected to be high scoring. Higher totals have been shown to increase television ratings for Monday Night Football (Paul & Weinbach, 2007) and increase the percentage of totals wagers on the over (Paul & Weinbach, 2008a, 2008b). Strategies of wagering on the under at the highest totals have been shown to generate positive profits over long samples in the NFL and NCAA football (Paul & Weinbach, 2002, 2003) and other sports. Although it is very likely the size of the "sides" (pointspreads and odds) market dominates the size of the totals (over/under) market, it is still possible, if more bettors wager on games with higher totals in the NBA and NHL, this variable will have a positive and significant effect on the volume of bets.

Through the inclusion of the pointspread (and the pointspread squared), some measure of team quality is captured. A small pointspread, however, cannot distinguish a matchup of two high-quality teams from a matchup of two low-quality teams. Therefore, to distinguish the quality of teams, the sum of the teams' win percentages, in the NBA model, is also included in the model. The higher the sum of the win percentages, the more fan interest there is expected in the game. Therefore, if bettors are also fans and gambling is a form of consumption, the sum of the win percentages should have a positive effect on the betting volume. For the NHL, we use a similar figure, but since the NHL used a points-based system to determine standings (two points for a win, one point for an overtime loss or shootout loss), we calculated the "win percentage" for the NHL teams by dividing the number of points earned divided by two times the number of games played (maximum number of points possible).

It should be noted here, this variable is the sum of the win percentages of the teams. While the individual win percentages of the teams or the difference in the win percentages would be highly correlated with the pointspread, which would lead to likely endogeneity problems in the regression. The sum of the win percentage is not highly correlated (3) with the pointspread, as small pointspreads are likely to exist with small, mid-range, or high sums of win percentages of the teams. Therefore, the inclusion of the sum of the win percentage helps to distinguish between a matchup of two high-quality teams compared to two low-quality teams. Alternative regression model specifications are presented in the regression results below, including a model with the pointspread and not the sum of win percentages and models with the sum of win percentages without the pointspread and/or total.

Early in the season, win percentage has limited accuracy as a measure of quality due to the limited number of observations. Therefore, for the first eight games (approximately 10%) of the NBA and NHL seasons, we included the previous year's sum of win percentage as this independent variable. After the eighth game, we use the sum of the win percentages in the current year, allowing for bettor memory of the previous season, and then replacing and updating for current performance as the season progresses.

Game timing is likely another important aspect in determining sports betting volume. Since games are played on different days of the week, dummy variables for different days of the week were included in the model. Also included in this category of independent variable are special holiday events (Christmas for the NBA), to determine to what extent holiday games are bet in these sports.

If sports betting is heavily influenced by consumption, television coverage is likely to be very important to betting volume. Therefore, dummies for televised coverage on different networks were included as independent variables. The more popular and available the network is to the betting public, the more likely betting volume is to increase when the network televises a game. Television dummies were included for all forms of national broadcasts (either over-air networks or cable/satellite) for each league. The inclusion of regional dummy variables for television coverage was considered, but nearly all games have some form of regional coverage when they are not nationally broadcast (and subject to restrictions on local television broadcasts), resulting in very little, if any, variation across games throughout the sample.

The regression results are shown in the tables. For each independent variable, the coefficient is presented along with the associated t-statistic in parentheses. Due to the presence of heteroscedasticity in the initial regressions we ran, we used White's Heteroskedasticity consistent standard errors and covariances. The results using this method are shown in the tables.

There are similarities that are immediately evident across the two sports that support the notion that sports gambling is mainly a consumption-oriented activity. It is important to remember that if sports betting was purely an investment activity, it is unlikely there would be systematic patterns of investment based on easily observable fan-oriented consumption-related variables.

The first factor which is easily identifiable across sports is that the quality of teams in the game affects the number of bets on a game. In the NBA and NHL, the sum of the win percentages of the teams has a large positive effect on the number of bets on a game. The sum of win percentages variable is significant at the 1% level across the different model specifications for each sport. Bettors appear to enjoy wagering on contests between the best teams.

The second factor which is readily apparent across sports is that bettors, like fans, appear to prefer uncertainty of outcome (pre-game measure) in sporting contests. In the NBA betting market, which uses pointspreads as the betting mechanism, the pointspread was found to have a positive effect on the number of bets, but the pointspread squared variable was found to have a negative and significant effect. NBA fans appear to enjoy wagering on good teams (with slightly higher pointspreads), but the more lopsided a game appears, the fewer bets a game attracts. In the NHL, the odds and odds squared were found to have a negative effect on the number of bets. Fans appear to not prefer wagering on lopsided games. This could be due to the additional monetary outlay which would be necessary to wager (to win a single dollar) to bet on the presumed better team in the NHL (the favorite) or it could be that fans do not prefer to watch and wager on perceived lopsided contests. Bettors again appear to be much like fans in that they enjoy wagering on games which are expected to be close and entertaining.

Another common factor determining betting volume for both the NBA and the NHL is bettor attrition throughout the season. The monthly dummies reveal that betting volume tends to be highest at the beginning of the season. With the initial euphoria of the beginning of a sports season, where hope springs eternal for each team, the volume of bets tends to be at its peak (for the regular season). As the season continues, the number of bets significantly drops over the course of the season. This could represent a drop in interest over the course of the season due to some teams being eliminated from the playoff race, but it is more likely this phenomenon is due to bettors losing their initial deposits and not refunding their accounts. Given the data from http://www.sportsinsights.com is from on-line sportsbooks, money must be funded to accounts up front. In illegal betting markets, this is not the case as bettors generally wager on credit and accounts are settled weekly. Therefore, illegal betting markets may not see as precipitous a drop in betting activity. An alternative explanation for these results is that bettors believe that their best opportunity in winning bets lies early in the season, before sportsbooks fully understand the relative ability of teams. Therefore, some bettors may only wager early in the season, then stop placing wagers later in the season when they believe the pointspreads and odds set by the sportsbook fully reflect all available information.

The NBA results reveal that fans appear to prefer wagering on road favorites as opposed to home favorites. In studies of sports wagering market efficiency, road favorites have been found to be overbet and simple wagering strategies involving betting on home underdogs have been shown to be profitable. This result illustrates a potential reason for the rejection of market efficiency found in these betting markets. This likely stems from a misevaluation of the true home field advantage by bettors. Bettors appear to need to lay fewer points on good teams (who field a strong enough team to justify being a road favorite) and bettors tend to bet more on these games and, specifically, on these road favorites. For the odds-based betting market in the NHL, the road favorite dummy variable was not found to have a significant effect.

Another factor which affects the number of bets on an NBA game is television coverage. Television coverage on ABC, ESPN, and TNT were all shown to have large positive and significant effects on the number of bets. ESPN2, which is not as widely available on cable television as ESPN or TNT and does not show NBA games as regularly as the other networks, had a negative and significant effect. The availability of close substitutes for bettors on ESPN, such as college basketball or college football, could account for the negative impact on volume seen for ESPN2 games. In the NHL, Canadian TV is shown to have a positive, but insignificant, effect on the number of bets. These results generally support the notion that bettors are consumers of sports, as opposed to investors. Bettors appear to enjoy wagering on games which they can watch, therefore, watching sports and sports wagering appear to be strong complements.

The results for the NBA also revealed that bettors appear to wager in greater numbers on games with higher totals (over/under bets). A significant increase in the number of bets was seen in games with higher totals. Although totals wagering is generally small compared to sides wagering (betting on a team to cover the pointspread), the number of bets in the NBA appeared significantly affected by the magnitude of the total. For the NBA, bettors, like fans, appear to enjoy games where there is likely to be more scoring during the contest.

One other interesting point is the effects of holidays, such as Christmas. Christmas is typically thought of as a family oriented holiday. Games on Christmas, however, do not lead to a decrease in betting volume, but rather leads to a large positive and significant effect on the number of bets on games played on this holiday. Christmas games led to 16,000+ more wagers on NBA games. As fans incorporate these sports into their holiday rituals, betting volume on these contests also increase.

Conclusions

Bettor behavior appears closely tied to fan behavior. Using betting volume data, bettors seem to prefer the same qualities in games which appeal to fans when choosing their consumption of sports (in viewing or attending). Bettors appear to prefer games between good, evenly matched teams that appear on the major television networks. These results are exactly what would be expected from fans of sports and bettors appear to mimic their preferences.

Overall, betting on the NBA and NHL appear to be much more about consumption than investment. Sports wagering appears to be a complement to watching sporting events. If wagering on sports was pure investment, such as the investment of money in the stock or bond market, we would not expect to see such large differences in betting volume across games based on typical consumption activity. Bettors as investors would likely search for prices (pointspreads and totals), where value was offered, resulting in a non-systematic relationship with fan-oriented variables. This is found to simply not be the case.

Bettors, like fans, appear to enjoy seeing the best teams. They prefer games televised on major networks. Bettors also appear to enjoy uncertainty of outcome, as betting volume was shown to decrease with higher pointspreads or higher odds on the favorite. Road favorites, which have previously been shown to be overbet and have been the focus of profitable betting strategies, tend to attract more bets overall than home favorites as fans tend to underestimate the home field advantage in the NBA. In addition, early season games appear to be the most popular games to bet in the regular season and the Christmas holiday attracted many bettors in the NBA.

Given the results that bettor behavior is similar to fan behavior, it is still possible that a small group of bettors may exist who are truly investors. It is possible that some small portion of the betting market may be using complicated strategies and advanced systems of money management. What the results of this paper show, however, is that these bettors are not the norm. Investigations of these bettors and how they behave are interesting to consider, if these bettors truly exist, but the normal rank and file of the betting masses do not appear to emulate these fine-tuned wealth maximizing agents outlined in financial texts. They simply appear to be utility maximizers, making consumption decisions where betting is a complement to watching a game and following a sports season.

Although tying bettors to investors may be helpful to financial and economic theory in understanding some elements of investor theory and testing the efficient markets hypothesis, realizing the importance of consumption when it comes to the sports gambling decision also helps to further understand behavioral theories. Understanding the actions of the majority of bettors helps us to further grasp the results of Levitt (2004) and Paul and Weinbach (2007, 2008) concerning findings of an unbalanced book. It helps us to understand the results of Strumpf (2003) as we consider the implications as we move from a small local sportsbook, with many inherent advantages including the power to price discriminate, to a large open sportsbook and the prices they offer.

Likely the biggest advantage, however, of realizing the importance of consumption value in sports gambling and the link between fan behavior and bettor behavior is the possibilities of prices in these markets helping to explain fan behavior. Sports have become a rather large industry in the United States and in many other places around the world. The betting market as a prediction market serves to generate ex-ante prices which can help us to better understand consumer behavior. Biases which may be apparent in betting markets due to gamblers behaving as consumers rather than investors are a benefit, not a curse. Information gathered through wagering markets, in terms of prices, volume, and bettor preferences seen through betting imbalances, are all useful in being able to help estimate fan demand before a game is played. It is the hope of the authors that this will eventually be a major component of the study of the economics and finance of sports wagering markets, rather than simply a mechanism to study market efficiency.

Appendix I: Summary Statistics for Non-Binary Variables

Available Games with full data (volume, pointspread,
total, etc.)--NBA: 1170; NHL:1227

 NBA

 Volume Pointspread Total Sum of Win Pct.

Mean 9993.8 6.25 199.5 1.00
Median 9230 6.00 197.5 1.00
St. Deviation 4514.4 3.66 12.3 0.26

 NHL

 Volume Favorite Odds Total Sum of Win Pct.
 (positive value)

Mean 5008.7 161.6 5.55 1.12
Median 4293 147 5.50 1.11
St. Deviation 3170.9 50.7 0.27 0.14

References

Buraimo, B., & Simmons, R. (2008). Competitive balance and attendance in Major League Baseball: An empirical test of the uncertainty of outcome hypothesis. International Journal of Sport Finance, 3(3), 146-155.

Conlisk, J. (1993). The utility of gambling. Journal of Risk and Uncertainty, 6, 255-275.

Forrest, D., & Simmons, R. (2008). Sentiment in the betting market on Spanish football. Applied Economics, 40(1), 119-126.

Golec, J., & Tamarkin, M. (1991). The degree of inefficiency in the football betting market: Statistical tests. Journal of Financial Economics, 30, 311-323.

Gray, P., & Gray S. (1997). Testing market efficiency: Evidence from the NFL sports betting market. Journal of Finance, 52, 1725-1737.

Levitt, S. (2004). Why are gambling markets organized so differently from financial markets? The Economic Journal, 114, 223-246.

Osborne, E. (2001). Efficient markets? Don't bet on it. Journal of Sports Economics, 2(1), 50-61.

Pankoff, L. (1968). Market efficiency and football betting. Journal of Business, 41, 203-214.

Paul, R., & Weinbach, A. (2002). Market inefficiency and a profitable betting rule: Evidence from totals on professional football. Journal of Sports Economics, 3, 256-263.

Paul, R., Weinbach, A., & Weinbach, C. (2003). Fair bets and profitability in college football gambling. Journal of Economics and Finance, 27(2), 236-242.

Paul, R., & Weinbach, A. (2007). The uncertainty of outcome and scoring effects on Nielsen ratings for Monday Night Football. Journal of Economics and Business, 59(3), 199-211.

Paul, R., & Weinbach, A. (2008a). Does Sportsbook.com set pointspreads to maximize profits? Tests of the Levitt Model of Sportsbook Behavior. Journal of Prediction Markets, 1(3), 209-218.

Paul, R., & Weinbach, A. (2008b). Price setting in the NBA gambling market: Tests of the Levitt Model of Sportsbook Behavior. International Journal of Sports Finance, 3(3), 2-18.

Rascher, D., & Solmes, J. (2007). Do fans want close contests? A test of the uncertainty of outcome hypothesis in the National Basketball Association. International Journal of Sport Finance, 2(3), 130-141.

Samuelson, P. (1952). Probability, utility, and the independence axiom. Econometrica, 20, 670-678. Sauer, R., Brajer, V., Ferris, S., & Marr, M. (1988). Hold your bets: Another look at the efficiency of the gambling market for National Football League games. Journal of Political Economy, 96, 206-213.

Sauer, R. (1998). The economics of wagering markets. Journal of Economic Literature, 36, 2021-2064.

Strumpf, K. S. (2003). Illegal sports bookmakers. Unpublished manuscript.

Zuber, R., Gandar, J., & Bowers, B. (1985). Beating the spread: Testing the efficiency of the gambling market for National Football League games. Journal of Political Economy, 93, 800-806.

Endnotes

(1) Although the data available from http://www.sportsinsights.com does not contain total dollars bet, betting percentage data on college football available from http://www.sportsbook.com (which bases percentages on dollars bet) and http://www.sportsinsights.com (which bases percentages on number of bets) have been shown to be similar (Paul & Weinbach, 2008). Given that the percentages are similar between the two sources, we believe it is likely volume would be similar also, although we cannot prove this as http://www.sportsbook.com does not provide volume with its data.

(2) Playoff games are likely to be quite different from regular season games in terms of attendance, TV ratings, and gambling activity. Therefore, we confined the analysis of this study to regular season games.

(3) In a simple regression with the NBA pointspread as the dependent variable and an intercept and the sum of win percentage as independent variables, the R-squared of this relationship was found to be 0.0028 and the t-stat of the sum of win percentage variable was 1.80 (significant at 10%). Compared to a similar regression with the difference of win percentage as an independent variable, the R-squared was found to be 0.2712 and the t-stat of the difference in win percentage variable had an associated t-statistic of 20.85 (significant at 1%). Similar results were found in relation to odds in the NHL.

Rodney J. Paul (1) and Andrew P. Weinbach (2)

(1) St. Bonaventure University

(2) Coastal Carolina University

Rodney J. Paul is a professor of economics in the School of Business. His research interests include sport economics, efficient markets, and time-series macroeconomics.

Andrew P. Weinbach is an assistant professor of economics in the Wall College of Business. His research interests include sports betting markets, fan attendance, and the economics of television ratings.

Table 1: NBA Betting Volume Regression--2008-09 Season

Dependent Variable: Number of Bets on Game

 I: II:
 Dep Var: Dep Var:
 Volume Volume (No
 Win Pct)

Constant -2532.786 6184.946 ***
 (-1.3547) (3.4029)
Road Favorite 1049.035 *** 1051.968 ***
Dummy (4.8270) (4.5158)
Pointspread 862.4169 *** 934.8673 ***
 (9.0978) (9.2180)
Pointspread (2) -46.26045 *** -48.8271 ***
 (-7.2637) (-7.2103)
Total 25.3324 *** 3.6214
 (3.0484) (0.4254)
Sum of Win 5244.878 ***
Percentage (12.8093)
Christmas 16340.94 *** 16757.80 ***
 (5.9614) (5.3276)
ABC 4748.817 *** 6815.462 ***
 (4.6699) (6.0669)
ESPN 3748.251 *** 5076.405 ***
 (7.7376) (9.7516)
TNT 4334.969 *** 5596.021 ***
 (4.1156) (4.6107)
ESPN2 -3389.846 *** -4920.021 ***
 (-8.2423) (-11.7142)
Sunday 2376.909 *** 2541.426 ***
 (5.9451) (5.9982)
Monday 3020.789 *** 3005.304 ***
 (9.5689) (8.9249)
Tuesday 2175.288 *** 2475.360 ***
 (6.8199) (7.2372)
Thursday 2370.708 *** 2733.125 **
 (2.5159) (2.5591)
Friday 2293.424 *** 2303.605 ***
 (8.0608) (7.6363)
Saturday 575.8243 *** 490.0281 *
 (2.0583) (1.6659)
November -2605.387 *** -2365.229 ***
 (-4.3063) (-3.7263)
December -2631.866 *** -2365.229 ***
 (-4.4096) (-3.8153)
January -1356.821 ** -997.764
 (-2.2088) (-1.5415)
February -864.2571 -489.8215
 (-1.3750) (-0.7415)
March -2664.741 *** -2267.150 ***
 (-4.4713) (-3.6420)
April -3446.656 *** -3117.507 ***
 (-5.3823) (-4.6381)
R-Squared 0.4852 0.4114

 III: IV:
 Dep Var: Dep Var: Volume
 Volume (No (No Pointspread
 Pointspread) or Total)

Constant -625.115 5264.650 **
 (-0.3287) (7.4087)
Road Favorite 724.4578 *** 680.6146 ***
Dummy (3.2371) (3.0373)
Pointspread

Pointspread (2)

Total 28.9157 ***
 (3.3455)
Sum of Win 5607.110 *** 5322.801 ***
Percentage (13.6369) (13.0026)
Christmas 15577.72 *** 15498.83 ***
 (6.4884) (6.3089)
ABC 4877.289 *** 4928.892 ***
 (4.9443) (5.0908)
ESPN 3568.662 *** 3564.100 ***
 (7.1839) (7.0995)
TNT 4255.118 *** 4276.141 ***
 (4.0595) (3.9983)
ESPN2 -2602.203 *** -2339.196 ***
 (-6.2982) (-5.7754)
Sunday 2346.424 *** 2400.5
 (5.6295) (5.7238)
Monday 3058.321 *** 3058.536 ***
 (9.2027) (6.9094)
Tuesday 2326.232 *** 2350.491 ***
 (6.9217) (6.9084)
Thursday 2359.394 ** 2429.798 ***
 (2.5284) (2.5918)
Friday 2363.572 *** 2385.587 ***
 (7.9035) (7.9223)
Saturday 627.1391 ** 631.4744 **
 (2.1601) (2.1730)
November -2868.314 *** -2908.216 ***
 (-4.6394) (-4.6856)
December -2925.314 *** -2821.686 ***
 (-4.7727) (-4.5871)
January -1600.176 ** -1524.098 **
 (-2.5416) (-2.4027)
February -1196.358 * -983.5722
 (-1.8492) (-1.5052)
March -2927.847 *** -2744.319 ***
 (-4.7887) (-4.4685)
April -3761.649 *** -3573.997 ***
 (-5.7736) (-5.4611)
R-Squared 0.4408 0.4352

Significance of the t-tests are noted by *-notation.
* denotes significance at 10%, ** denotes significance
at 5%, and *** denotes significance at 1%.

Table 2: NHL Betting Volume Regression--2008-09 Season
Dependent Variable: Number of Bets

Independent I II: III:
Variable No Odds No Sum
 or Total Win Pct

Constant -3963.470 * 1799.405 ** 181.5083
 (-1.7976) (2.2724) (0.0881)
Road Favorite 194.0674 -214.289 1.1643
 Dummy (0.9686) (-1.0603) (0.2445)
Favorite Odds -55.4272 *** -55.3296 ***
 (-6.3210) (-6.2475)
Favorite Odds (2) -0.0991 *** -0.0971 ***
 (-4.7273) (-4.5877)
Total -17.8723 -180.4613
 (-0.0532) (-0.5346)
Sum of Win 2917.975 *** 3250.507 ***
 Percentage (4.9514) (5.3518)
CBC 556.9274 376.344 683.352
 (0.9971) (0.6494) (1.2127)
Versus 15.4724 123.8918 186.0642
 (0.0353) (0.2729) (0.4226)
TSN 486.8879 551.3528 504.2171
 (1.1981) (1.3114) (1.2285)
NHL Network -829.5571 -651.4412 -906.851
 (-1.5355) (-1.1643) (-1.6627)
Sunday 582.9357 646.8376 569.5462
 (1.4451) (1.5448) (1.3980)
Monday 851.2148 * 841.1775 * 768.5086 *
 (1.9293) (1.8368) (1.7258)
Tuesday -93.7566 -103.6503 -165.3204
 (-0.2590) (-0.2763) (-0.4526)
Thursday -218.7171 -102.7107 -274.2408
 (-0.6056) (-0.2742) (-0.7522)
Friday 26.4807 214.8673 -32.5852
 (0.0682) (0.5340) (-0.0831)
Saturday -796.3435 ** -829.0555 ** -839.5745 **
 (-2.2581) (-2.2644) (-2.3579)
October 1268.685 *** 966.4053 *** 1271.154 ***
 (3.7717) (2.8044) (3.7416)
November 105.6591 -73.2741 118.1316
 (0.3370) (-0.2269) (-0.5814)
December -179.4548 -220.6076 -178.1276
 (-0.5867) (-0.6953) (-0.5766)
February -162.7685 -148.0965 -181.3471
 (-0.5270) (-0.4618) (-0.5814)
March -1106.466 *** -1131.286 *** -1113.350 ***
 (-3.6406) (-3.5878) (-3.6270)
April -1954.770 *** -1849.676 *** -1958.523 ***
 (-5.0562) (-4.6221) (-5.0157)
R-Squared 0.175 0.1202 0.1554

Significance of the t-tests are noted by *-notation.
* denotes significance at 10%, ** denotes significance at 5%,
and *** denotes significance at 1%.