Worthless and Weak

First day of work, ack it was boring...

Good things happen in threes...

I got the only A in econometrics

I have a job interview tomorrow

what else shall befall me I wonder?

Amanda: "Oh my God!!!!... [Garrett] has a girlfriend"

Well, after 16 games, we're 10% of the way through the season, so I thought I might look at where we're going.

At 10 and 6, the Sox are on pace to win 100 games, a good mark. They've allowed merely 62 runs, only the ChiSox have allowed fewer in the AL. In addition, the sox have scored 87 runs, only the Blue Jays, Yankees, and Tigers have scored more (but not much more).

Looking at the players individually, Varitek Ramirez, and Nixon all look very good, Ortiz isn't bad (although he's been in a bit of a slump lately) Millar is acceptable but not too good, and Damon, Mueller, Bellhorn and Renteria have done quite poorly so far.

Obviously, Varitek won't end the year hitting .367, and Renteria won't end the year hitting .210 either.

In terms of backups, Payton and Mirabelli have been nice, Vazquez has been horrible.

The starting pitching has been wonderful, while the bullpen suspect so far. After a pair of rough starts, Wells has dominated for his past two games, shutting down the powerful Baltimore and better than you think Devil Ray lineups, bringing his ERA down to 3.51. Arroyo has been fine, and Wakefild has been awesome, with a Gibsonesque 1.37 ERA. The only dissapointment so far has been Schilling, with a 6.75 ERA, but thats only over two starts. All told, the starting 5 have pitched exactly 100 innings with a 3.15 ERA, which, if that held up, would be incredible. Of course, it won't, but its still awesome.


As for the bullpen, its been the weak spot on the Sox so far. 39.1 innings pitched so far with a 4.81, including dissapointing contributions so far from Matt Mantei and John Halama, and Foulke has yet to find his groove. Only Timlin and Myers (in limited playing time) have been impressive so far.

Of course, the flip side to that is that so few of the relievers have any number of innings pitched to make any sort of conclusion yet, but as a whole, they haven't been too good.

Of course, with everything here, its waaaaaayy to early to draw any conclusion about anything. Manny will continue hitting, not like he has this season so far, but like he's been hitting these past several seasons. Varitek could have a breakout season, but he probably won't hit above .330.
Looking at MLB so far, the biggest suprises are...

The demise of the Yankees
The Dodgers are at 12-2
Florida has scored 33 more runs than it allowed
The Chicago White Sox are at 12-4

The Yankees probably aren't really any more dead now than they were at the start of the season. They'll most likely come back.

The Dodgers are for real, and will finish first in their division.

Florida is too early to call (of course, everything is way too early to call, but whatever)

And the Chicago Sox will fall back to earth, although they might compete for the division.

For the awards, if the season ended today...

AL MVP Brian Roberts (Varitek as a runner up)
AL Cy Young Rich Harden (Wakefield runner up)

NL MVP Jeff Kent (Edgardo Alfonzo runner up)
NL Cy Young Josh Beckett(Roger Clemens runner up)

Do Ten Five rights affect baseball salary? I tackle the question in a lengthy econometric analysis.

The Ten Five Conundrum

Introduction

In 1969, St. Louis center fielder Curt Flood, who had played twelve straight years for St. Louis, was traded to the Philadelphia Phillies. However, he refused his assignment, and sued for the right to free agency. He eventually lost his court battle, but in 1972 the league granted all players who had played for ten years, the previous five with the same team, the right to veto any trade of that player. This is known as Ten Five rights, and is the first major right the players won from the owners. However, in subsequent years, players with six years of service time received the right to unrestricted free agency. Since players who are given Ten Five rights cannot refuse those rights, and as theory states that players who would bargain for these rights would receive less monetary compensation for their service, players who qualify for Ten Five rights should earn less than those who don’t. This is important because if true, it would mean that the Ten Five rights hold virtually no purpose, as although they were important when first granted, they now actually hurt the players the attempt to help, and should thus the Ten Five rights should be removed from the collective bargaining agreement.

Literature Review

There have been a lot of studies done about baseball salaries and what affects them. While none of them tackle the issue of Ten Five rights, many of them are very important, and can help illuminate the question of Ten Five rights. Burger and Walters, in Market Size, Pay, and Performance show that the size of a market significantly affects the value of a player to a particular team. In addition, it also shows that the contending status of a particular team greatly affects the value of a player to a team. (ie, if the team reaches the point where it contends for the playoffs, it will value players significantly more than if it didn’t). These conclusions show that a variable for city size should be included in the model.

The most similar study to this is was done by Mathew Clayton and David Yermack of New York University. In Major League Baseball Player Contracts: An Investigation of the Empirical Properties of Real Options, Clayton and Yermack look at the effect of player and team options on contracts. It finds that players who bargain for player options (options to extend a contract that can be executed by the player) in their contracts make a significantly less amount of money per year than players who don’t have these options. Likewise, a team option (an option to extend a player’s contract that can be executed by a team) will significantly increase the pay of the player. In this paper, instead of simply using salary per year data, they look at the net present value of the contract signed by a player. This allows them to take into account various irregularities, players who have deferred payments on their salary, and other options. As such, they explain much more variation in player compensation than they would have been able to otherwise.

Krautman and Oppenheimer, in Contract Length and the Return to Performance in Major League Baseball, looks at the affect of contract length in player salaries. They theorize that players with longer contracts should make less money per year than those with shorter contracts, as players with longer contracts are insuring against injury or catastrophic decline, and thus should pay an insurance “premium.” However, upon finishing his research, they find that players with longer contracts actually make more money than those with shorter ones. This, they theorize, is because a longer contract is actually a form of compensation instead of a form of insurance.

Bodvarsson and Pettman, in Racial wage discrimination in major league baseball: do free agency and league size matter?, looked to see if there was discrimination against non-white baseball players. Using a regression for the 1992 and 1993 baseball seasons, they found two interesting things. First, that baseball did discriminate against non-white ballplayers, but only for those that didn’t have free agency status (since this study looks only at free agency eligible players, a variable on race was not included). Also, they found that when the league expanded in 1993, previous effects of discrimination were eliminated, thus more teams (and thus more competition) leads to less discrimination in major league baseball.

Description of Model

The model explains the dependant variable, salary, with five independent variables, three of which are quantitative, one of which is qualitative, and a series of six qualitative variables to describe the position of the baseball player. The dependant variable is the natural log of salary (lnSALARY), which simply measures the salary of the baseball player in 2002. The four quantitative independent variables are 2001 win shares (WINSHARES), which is a measure of player performance, age (AGE), years of MLB service time (SERVICE) and city size (CITY). The first qualitative variable (TENFIVE) represents if the player has 10/5 status. Finally, six dummy variables are included to describe the player’s position, (CATCHER, OUTFIELD, SHORTSTOP, 1B, 2B, and 3B).

Because a team will pay more for better players, WINSHARES is expected to have a significant positive effect on player salaries. Because younger players are more likely to improve than older players, and older players are more likely to get injured or fall in performance, the player's salary should decline as age increases. On the other hand, teams may value veteran leadership and experience, thus increasing salary as the the player gets older. Therefore, just considering age could lead to a bias, as information on age includes information on playing time. To correct this, AGE (which is expected to have a negative sign) and SERVICE (which is expected to have a positive sign) are both included.

The position variables, (CATCHER, OUTFIELD, SHORTSTOP, 1B, 2B, and 3B), are expected to effect the salary of players based on demand and supply for each position, and while it is expected that some positions will make more than others, at this point in time it is unknown which positions those are. Thus, there are no expected signs on any of these variables.

A team in a large city will benefit more from each additional win than a team in a smaller one, thus a team in a larger market will be demand a particular player at a higher level than a team in a smaller market. Thus, the sign on CITY is expected to be positive.

Finally, as there are additional costs of signing a player who has Ten Five rights, (a loss of flexibility for the team) players who have Ten Five rights are expected to earn less than those who don't.

Description of Data

The model will attempt to explain variations in player salary, thus the dependent variable will be the natural logarithm of a player's salary. This data was obtained from the Society of American Baseball Research’s committee on business of baseball, and can be found online at http://businessofbaseball.com/data.htm.

To measure performance, I used Bill James’s Win Shares, a statistic designed to measure all aspects of performance. It uses data on hitting, base-running, pitching, and fielding. Its major advantage is that it allows some confidence when comparing hitters to pitchers, and there are few other statistics which can make that comparison. While I used the 2002 salary, I decided to use 2001 Win Shares, as salary is primarily based on past performance. The data can be found in the book “Win Shares” by Bill James and Jim Henzler.

For age, I simply used the age the player was on July 1st of 2002. For position I intend to use 6 dummy variables, one each for 1st baseman, 2nd baseman, 3rd baseman, shortstop, outfielder and catcher. I am using one variable for outfielder instead of three (for three outfield positions) for two reasons. First, because the outfield positions require overlapping skills, thus outfielders frequently play more than one outfield position in any given year, and secondly because the methodology James uses to calculate win shares treats all outfield positions equally. When a player plays more than one position, I use the position he played most.
For Ten Five rights, I simply use a dummy variable. Data on player age, service time, position, and contract status can be found at http://www.baseball-reference.com/. For city size, I used the size of the metropolitan area the team play in, which can be found at http://www.citypopulation.de/.

Finally, it should be noted that I am using only players who have qualified for free agency, (have either 6 years of service time starting in 2002 or have been granted free agency by other means) as “reserved” players are not paid market prices for their services.

Description of Results

The results of the regression are:

Si = 17.4 + .066 Wi - .141 Ai + .146 SEi-.389Ci +-.261 Fi - .823 SBi - .267 Ti – .425 SSi - .323 Oi + .295 TFi + .016 SIi

Where:

Si = the natural log of the salary of the ith ballplayer
Wi = the number of win shares the ith player earned in 2001
Ai = the age of the ith ballplayer
SEi = years of service of the ith ballplayer
Ci = 1 if ith player was a catcher, else 0
Fi = 1 if ith player was a first baseman, else 0
SBi = 1 if ith player was a second baseman, else 0
Ti = 1 if ith player was a third baseman, else 0
SSi = 1 if ith player was a shortstop, else 0
Oi = 1 if ith player was an outfielder, else 0
TFi = 1 if ith player had Ten Five rights, else 0
SIi = size of the city the ith player plays in

Every coefficient had the expected sign, except for the ten five rights, which has a positive sign instead of a negative one. The coefficients for Win Shares, Age, Second Baseman, Outfield and Years of service were all significant at the 1% level, while those for Catcher Shortstop and city were significant at the 5% level. The coefficient for First baseman and Ten Five rights were significant at the 10% level, while the coefficient for third baseman was not significant. An F-Test shows that the coefficients on the position dummy variables are significant as a group.

The coefficient for W shows that an increase in performance equal to one win share will result in an increase in salary of 6.63 %. The coefficient for A means that a one year increase in age will decrease salary by 14.1 %. One additional year of service time, however, will increase salary by 14.6%. The coefficients for position show that a catcher will make 38.6% less than a pitcher, a first baseman 26.4% less, a second baseman 82.4% less, a third baseman 26 % less, a shortstop 42.4% less, and an outfielder 32.4 % less than a pitcher would, provided equal age, service time, and performance. In addition, for every million people living in a city a ballplayer lives in, his salary will increase by 1.6%. Finally, if a player has Ten Five rights, he will make 29.25 % more than one who doesn't.

The adjusted R squared measure shows that the model explains 46% of the variation in the dependent variable. Furthermore, using an F-Test for the whole model, we can see that the model is significant at the 1% level.

If two or more independent variables are highly correlated with each other, they have the potential to affect the hypothesis testing of each other. At a quick glance, the two independent variables which would probably be most effected with each other are age and service time, as for each individual player as age increases, service time also increases. However, this is not as big of a problem using cross-section data, as there are many players who entered the majors at a young age, and have several years of experience while other player don’t enter the majors until they are in their late 20’s or early 30’s. However, the two are still highly correlated, at a .791 level. Using a variance inflation factor test to see if they would skew the results, we find that the VIF for AGE is only 3.0, while the VIF for SERVICE is 3.1, which, combined with the fact that both are significant at the 1% level, there is no need to correct for multicollinearity.

The results of a white test show that the model contains heteroskedasticity, which can reduce the efficiency of the model, and can cause the t-values to become unreliable. A series of Park tests show that the variable WINSHARES is responsible for introducing heteroskedasticity into the model. In order to correct for heteroskedasticity, the dependant variable should be changed from the natural log of salary to the natural log of salary divided by win shares. However, this is impossible, as the variable win shares can be zero. Thus, to eliminate this, the new dependant variable is the natural log of salary divided by win shares + 1 or ln(salary/(win shares + 1)). In the new model, the variable for win shares has obviously been removed, leaving only AGE, SERVICE, CATCHER, FIRSTBASE, SECONDBASE, THIRDBASE, SHORTSTOP, OUTFIELD, TENFIVE and SIZE. The new model is:
Si = 14.953 - .095 Ai + .1020 SEi - .423 Ci - .242 Fi - .980 Si - .409 Ti - .429 SSi - .334 Oi + .366 TFi + .0117 SIi

Where:

Si = the natural log of the salary divided by win shares + 1of the ith ballplayer
Wi = the number of win shares the ith player earned in 2001
Ai = the age of the ith ballplayer
SEi = years of service of the ith ballplayer
Ci = 1 if ith player was a catcher, else 0
Fi = 1 if ith player was a first baseman, else 0
SBi = 1 if ith player was a second baseman, else 0
Ti = 1 if ith player was a third baseman, else 0
SSi = 1 if ith player was a shortstop, else 0
Oi = 1 if ith player was an outfielder, else 0
TFi = 1 if ith player had Ten Five rights, else 0
SIi = size of the city the ith player plays in

The interpretations of these coefficients are now on a per win share basis. Thus, for each additional win share, a catcher will receive 42.3% less money than a pitcher, for instance. While the magnitude of the coefficients have all changed, the signs have remained the same, and the significance of every coefficient remains fairly unchanged, except for city size, which is now only significant at the 10% level, third base, now significant at the 5% level, and first base, which is no longer significant.
The white test shows that the model is no longer heteroskedastic, thus the new form has solved the problems inherent in the old one. However, since none of the signs changed, and there was little change in the significance of each variable the old model is still probably mostly correct, especially about the variable in question, Ten Five Rights.

Conclusion

When looking at the effect of Ten Five rights on player salaries, the initial theory, that Ten Five rights will decrease player salary has not only been not proved, it has been entirely contradicted. Ten Five rights, in fact increase the salary of a player, whether measured in absolute terms or on a salary per win share basis. It is interesting that this, like Krautman's work on contract length, contradicts the initial theory. This might be due to a variety of factors. First, player's who have played with a particular team for a long time might be fan favorites, and would thus be worth more to teams than other players. This might be large enough to counteract the inflexibility that Ten Five rights bring teams. Krautman showed that as contract length goes up, so does a player's salary. Thus, if players with Ten Five rights generally have longer contracts, then the coefficient on Ten Five rights will be biased upwards. These two problems could be accounted for by including variables on contract length and the number of years with the same team in a regression. Until that is done, we can only assume that Ten Five rights have a positive effect on player salary.

That was the easiest ten page paper I've ever written.

I nearly cried during today's ring ceremony. And I really wish I could have seen it all, but alas, I had to go to class. Anyways, the sox beat the yankees, thus tying them, and they are ready to start rocking, as Schilling arives on Wedsnesday. Anyway, happy baseball!!

Could life get any better?

Its spring out. I have a girlfriend. The Red Sox are world champions. The Patriots are world champions. Hockey's been canceled.

You know its a weird weekend when you find yourself thinking "thank god the weekend is over, can't wait to get back to the week."