Thursday, February 1, 2018

Consumers balk when cheaters are revealed: more evidence from steroid suspensions in Major League Baseball

TLDR: There exists just a small body of evidence that supports the notion that sports fans do not like their athletes using performance-enhancing drugs (PEDs).  For example, I co-authored the first study that proved a systematic negative effect of a PED suspension on revenue in Major League Baseball (MLB) teams. This effect was found to be about a $500K loss in ticket revenue.

I have now also calculated the foregone revenue in concessions and loss of television viewership due to PED suspensions in the MLB.  I find that the average cost for a suspension to be $1.5 million.

In 2015, I co-authored a paper entitled "Do Fans Care About Compliance to Doping Regulations in Sports?" where we calculated the cost of a performance-enhancing drug (PED) suspension to Major League Baseball (MLB) and its teams. We found that immediately after a PED suspension was announced, the guilty player's team could expect to lose 8% of their ticket sales. However, fans forgive and forget and we found the negative effect of the suspension begins to wane shortly there after. In total, each suspension cost the team around $500K.

Since publishing the paper, I have pondered a similar issue of suspensions and live television viewing. I have since turned my focus to the television ratings of MLB games. Television ratings are the number of households watching a given program as a share of all households with a television, as estimated by the organization Nielsen. When we multiple the rating of the local baseball game by the number of television households in the area we get the number of viewers, which I call the television audience. I was able to collect ratings for locally broadcast games in the USA from 2006 to 2012, approximately the same period of the aforementioned study (sorry, no Toronto Blue Jays and no nationally broadcast games).

Although I expect PED suspensions to have a similar effect on television audiences as they do for ticket sales, there are two slight nuances for which we may see a different effect.
  1. Every game is broadcast locally while tickets are purchased in the local market for only home games.
  2. Every television viewer has the option of turning off the game while ticket sales are non-refundable (attendance is reported as the number of tickets sold, not the number of bums in the seats).
Because of the two reasons above, the impact to television may (a) be more immediate, and; (b) impact more games.

I use my measure of television audience to run a simple regression of log audience. Each game gets two observations - one for the broadcast of the home team and one for the visiting team. I control for team, the opponent, the day of the week, month, and year the game was played, and the probability the local team will win (here local team indicates the team for which the broadcast is affiliated with).

I also include a dummy that takes the value of one if the local team has a player serving a PED suspension during the game. There were 36 PED suspensions between 2005 and 2012. I consider 21 of these suspensions, excluding the 15 others due to various reasons (see page 6 here for a better explanation why). In the end, there are 611 broadcasts (or 2% of the observations) where the local team had a suspended player.

I also include a discrete count variable which indicates the time measured in days since the suspension was announced. This measure of elapsed time since announcement is always zero if the suspension dummy is zero. The estimated effect of the elapsed time gives us insight into how the stigma of the suspension may decay over time.

An illustrative depiction of the estimated effect of a PED suspension is below - the combined effect of the initial announcement and the elapsed time. In this graph, the television audience of the typical local team appears to be as expected prior to any suspension announcement. After a suspension has been announced (the dashed red line) the team can expect an immediate 6.7% decline in television audience. As time since the suspension announcement passes, fans slowly begin to return to watching their team as the model would customarily expect. It is not until 33 days after the suspension announcement that the effect can no longer be distinguished from zero.

Due to various reason, the cost of a PED suspension announcement is very wide-spread. Regardless of who carries the burden, I calculate three avenues where we can observe the cost of a suspension - ticket sales, concessions, and viewership. I have included an explanation below for those who are interested in how I arrive at my numbers.

With little surprise, the largest sum of foregone revenue from a hypothetical PED suspension would come from a suspension from the Yankees, Dodgers, Red Sox, Cubs, Phillies, or Angles. These are teams with large television markets and often have expensive tickets.

As I describe below, each part of the above figure are at least a lower-bound estimate of the true PED suspension effect. Additionally, there are many different avenues that may have direct or indirect impact from the consumer reaction to PED suspensions, such as merchandise. While we know some sort of dollar figure for the short-run effect of PED suspensions, we are still missing a large piece of the literature describing the long-run impacts of PED suspensions in sports.

The calculations:

Foregone Ticket Sales:

Revenue sharing means that each baseball club is only guaranteed to keep around ⅔ of their ticket revenue - the remainder is put towards a communal pot to be divvied up among the league's least financially successful clubs. There are clearly spillover effects when one team takes a large hit to ticket revenue, yet regardless of which team or teams lose out, we can calculate the total value of the foregone ticket sales of team i as follows:

The alpha's come from Model 4 of Cisyk & Courty (2015) and I multiply the effect by 1/2 because the attendance is only affected if the suspended player's team is at home (which happens half of the time). Little t refers to the number of days since the announcement and big T refers to the last day the effect is significant.

As mentioned above, this calculation of foregone ticket sales is a lower-bound estimate as season ticket holders or fans who purchased well in advance of the game are always counted in the official attendance figures despite the fact that they may not have truly attended.

Foregone Concessions:

Using some historical financial data, I estimate the total value of foregone concessions to be approximately ⅓ of foregone ticket sales as follows:

Because the foregone ticket sales are said to be a lower-bound estimate, the foregone concessions must also be a lower-bound estimate of the true loss in concessions.

Foregone Television Audience:

Despite the fact that less audience members are tuning in to the broadcasts, a baseball team with a PED suspension is would appear to be no worse off (at least in the short-run). Instead, Regional Sports Networks (RSNs) are locked into lengthy contracts with the local team to have the (often exclusive) rights to broadcast each game. RSNs then recoup their expensive fixed-cost contracts by selling ad space to numerous advertisers.

The cost of a PED suspension is a bit trickier here; for example SportsNet New York bought the exclusive rights to broadcast the games of the New York Mets for 46 million per year. The Mets have around 46 million viewers per year, which would suggest SportsNet New York must think they are able to sell the advertising space for a sum total of at least 1 dollar per viewer (called the cost per impression or CPI). This also suggests the collective revenue from the parties running their advertising which is attributable to the ad must sum to at least 1 dollar per viewer across all advertisers (called the revenue per impression or RPI).

Therefore, the lower-bound cost of the foregone television audience, as seen through either (a) a loss in average ad-based revenue for all advertisers; (b) a loss in the CPI for the RSN, or (c) a mix of (a) and (b), can be calculated as follows:

where, absent any suspension, advertisement j shown to viewer k during team i's broadcast generates at least as much revenue as it costs, such that:

Similarly, the beta's come from my model I run on the natural logarithm of attendance. Again, little t refers to the number of days since the announcement and big T refers to the last day the effect is significant.

No comments:

Post a Comment