ECON2206 (z3389988, z3389750, z3389677)

ECONOMETRICS PROJECT
Pilar Carnicero Mazn (3389988) Javier Muruzbal Huarte (3389750) Rosala Rodrguez Bermdez (3389677)

Introductory Econometrics (ECON2206)

THE UNIVERSITY OF NEW SOUTH WALES

ECON2206 (z3389988, z3389750, z3389677)

TABLE OF CONTENTS

1. INTRODUCTION ...... 3 2. DATA ... 5 3. MODEL .... 8 4. CONCLUSIONS ...... 11 5. REFERENCES .... 12 6. APPENDIX ..... 13

ECON2206 (z3389988, z3389750, z3389677)

1. INTRODUCTION
Is there marriage premium for professional athletes? Researchers have noticed that married men earn considerably more per hour worked than men who are not currently married. There are several approaches in the literature that provide this statement some empirical support. Before the description and explanation of our own econometric theoretical model, it is necessary to discuss some of these results and their implications in order to clarify our ideas regarding the topic. There is widespread agreement that cross-sectional marriage differentials for men are considerable. However, what makes researchers differ is the reason of this wage premium, the source of this gap. The first one is based in productivity differentials. Some researches states that marriage becomes people more productive (Becker 1981, 1985; Kenny 1983; Greenhalgh 1980). The second one relates this differences to employer favoritism (Hill 1979; Bartlett and Callahan 1984). Lastly, a third hypothesis is due to selection into marriage on the basis of wages or personal characteristics that are valued in labor markets (Becker 1981; Nakosteen and Zimmer 1987; Keeley 1977). Basing their models on different hypothesis and variables, we have focused in the conclusions presented by Koremnan and Neurark (1991), Reed and Harfor (1989) and Cornaglia and Feldman (2010) since they provide good approaches in the explanation of the marriage premium. In their paper, Reed and Harfor present a different and extreme explanation. They state that marital status and higher productivity are unrelated so marriage premium simply is due to labor market substitutions by married workers resulting in higher wages. Both authors agree that substituting wages for less attractive job characteristics, married workers may also consume less leisure-on-the-job, working harder than single workers. In our model, we will also consider productivity as an important part of our analysis. Numerous authors have followed Koremnan and Neurark approach since it shown new evidence in this field. They conclude that, marriage premiums seem to ascend slowly, resulting from quicker wage growth for married men as compared to never-married men, than from an intercept change associated with any particular marital status. They affirm that the marriage premium appears due to the location of married workers in higher paying job grades within the company, rather than to married workers receiving higher pay than single workers within the

ECON2206 (z3389988, z3389750, z3389677)

same job grades. Despite the interest of all these contributions, our model will be simpler as our data are already given. A different view is given by Cornaglia and Feldman in their recent approach The Marriage Premium Revisited: The Case of Professional Baseball. This paper has been really useful for us; despite it refers to baseball players, it has helped us to understand what can determine marriage premium. Besides, it is simpler and more suitable for our data and our level of knowledge. Does marriage premium exist? Although there do not appear to be differences in productivity between married and unmarried players, Cornaglia and Feldman investigate the uncertainty of marriage premium assuming in the model that impacts salary can be due to: Positive causal effect on productivity since one spouse engages in particular actions that influence the productivity of the other spouse. Employers to discriminate in favor of married men as they have usually more stability. They create a model in order to explain and give evidence of their hypothesis. Based on it, these researchers confirm the existence of a marriage premium. However, they find no robust statistically significant effect on productivity when using a variety of productivity measures. Hence, employers may discriminate in favor of married players because they lead to greater team success that is not necessarily captured by the productivity measures. It could also be interesting to make a brief reference to reverse causality: higher levels of wage imply more opportunities to get married. Korenman (1988) is the only paper that attempts some sort of formal test, however he could find no evidence for it. Once we have introduced the topic by referencing to the most important literature, we can confirm its importance and interest. It is a fascinating field of investigation, as the male marital pay premium accounts for about one-third of estimated gender-based wage discrimination in the United States (e.g., Neumark 1988). We could learn important concepts of econometrics through its application to a sector that is interesting to most people. See if the marriage affects the wage involves several fields such as econometrics or statistics but also sociology, which always brings a note of realism to the issue. Easy data collection process on various productivity measures and the public nature of their salary makes it easier to check whether these differences are related the to salary.

ECON2206 (z3389988, z3389750, z3389677)

2. DATA
The main database we use comes from the National Basketball Association (NBA), which contains information of the professional basketball players that are currently playing in the American National League of Basketball. The data set is NBASAL.DTA, and gives detailed information about the points scored, rebounds, assists, playing time, and demographic information such as race, number of children, age and wage of the players. In order to develop a model that can take into consideration the elements we think we are the most relevant in order to see possible relationships. First of all, we took experience and age, as they are always related with the wage. Employers take into account the previous experience and the age of their employees to set the price. Studies have proven that the more years a person has stayed in a company, the more they earn (Lazear, E., 1976). The fact that the player as been an allstar and the number of points and minutes played on average in each game are important facts that must be taken into account also (Cornaglia, F. & Feldman, N.E., 2010). On the other hand, we rejected numerous variables. We thought that the fact that the players were black or they are married with black people (variables black and marrblack) does not really matter in our case. It is true than in a lot of companies, it is true that being black is a relatively low salary compared to other workers. to be talking about the NBA, the American Basketball League, this fact is not a disadvantage. a large percentage of players are black, and therefore do not suffer discrimination (Scully, G.W., 1973). In the same way, the position of each player on the basketball team (variables guard, forward and central) were not considered to develop our model either. We don't think they are relevant, as they not pose significant effect on the salary for us. Other variables are more relevant for our study. This is the model we developed: lwage = 0 + 1marr + 2exper + 3coll + 4age + 5points + 6draft + 7allstar + 8children + 9minutes + 10expersq Where, - lwage is the variable that measures the annual salary in thousands of $. - The variable marr means the marital status of the players, if they are married or not.

ECON2206 (z3389988, z3389750, z3389677)

- The variable exper means the years that the subject has played as professional player. - The variable coll are the years played in college. - The variable age measures the age in years. - The variable points measures the points scored per game. - Draft means draft number. The NBA Draft is a happening that takes place every year in which the teams chooses players who want to join the league. These players are usually amateur college basketball players. - Allstar means if she has ever been an all star. - The children variable means if they have any kids, no matter how many, just if they have or not. - Minutes measures the average of minutes played per year. - Expersq is the square of the years as professional player.

When analysing variable wage, we observe that there is a decreasing tendency. The way to eliminate this tendency is by using the log of the variable. In our case, we can not specify if there is seasonality or not because the data are per year (we only have annual data).

In order to get some more information, we made some basic descriptive statistics using STATA. We got to the following results: Almost 45% of the NBA players are married (.4423792 mean). The wages vary between 150 and 5749 thousand $, i.e. There is a wide range of salaries

among the professional players. - The oldest player in the NBA is 41 years old and the youngest is 21 years, with an average age of 27.40 (=27.39405).

ECON2206 (z3389988, z3389750, z3389677)

Although they are players who didn't play at college, the average number of years in college The average number of minutes played per year is 1682.193. The average number of points scored per year is 10.21041.

is 3.72 (=3. 717472).

- 11.52% of the NBA players have been all stars (.1152416 mean). - 34.57% of the NBA players have children (.3457249 mean).

Variable marr wage exper age coll minutes points allstar children expersq

Obs 269 269 269 269 269 269 269 269 269 269

Mean .4423792 1423.828 5.118959 27.39405 3.717472 1682.193 10.21041 .1152416 .3457249 37.72119

Std. Dev. . 4975945 999.7741 3.400062 3.391292 .7544096 893.3278 5.900667 .3199085 .4764905 46.53702

Min 0 150 1 21 0 33 1.2 0 0 1

Max 1 5740 18 41 4 3533 29.8 1 1 324

ECON2206 (z3389988, z3389750, z3389677)

3. MODEL
After observing data provided, we thought which variables could be relevant for our analysis. Hence, se started with a basic econometric model that included these variables. lwage = 0 + 1marr + 2exper + 3coll + 4age + 5points + 6draft + 7allstar + 8children + 9minutes + expersq

lwage marr exper coll age points draft allstar children minutes expersq _cons

Coef. -.0700351 .1379216 -.0709909 .0278891 .0608719 -.0122365 -.2521796 .0359167 .0000686 -.0069926 5.637006

Std. Err. .0837678 .0484446 .0533948 .0357233 .0134554 .0021765 .1430369 .083163 .0000784 .0026695 .8267092

t -0.84 2.85 -1.33 0.78 4.52 -5.62 -1.76 0.43 0.87 -2.62 6.82

P>|t| 0.404 0.005 0.185 0.436 0.000 0.000 0.079 0.666 0.383 0.009 0.000

[95% Conf. Interval] -.2350892 .0424675 -.1761987 -.0424993 .0343596 -.0165251 -.5340161 -.1279458 -.000086 -.0122525 4.008077 .095019 .2333757 .034217 .0982775 .0873841 -.0079478 .029657 .1997791 .0002231 -.0017327 7.265935

Looking at the confidence interval we can see that some of the elements chosen are irrelevant explanatory variables that could have negative effects on the other variables of the model, such as years played in college, age, if the person was an ever all star, if he has children and minutes per season played. The final model would be: lwage = 0 + 1marr + 2exper + 3points + 4draft + 5expersq

lwage marr exper points draft expersq _cons

Coef. -.0426041 .1666236 .0611561 -.0117598 -.0071634 6.071818

Std. Err. .0800065 .0367069 .0070094 .0021244 .0025991 .1250561

t -0.53 4.54 8.72 -5.54 -2.76 48.55

P>|t| 0.595 0.000 0.000 0.000 0.006 0.000

[95% Conf. Interval] -.2002292 .0943053 .0473465 -.0159451 -.0122841 5.825439 .115021 .2389419 .0749656 -.0075744 -.0020428 6.318198

lwage =

6.071818 -.0426041marr + .1666236exper + .0611561points -.0117598draft -

.0071634expersq

ECON2206 (z3389988, z3389750, z3389677)

Analysing the results we can see that wage is not affected by marriage because: - 0 is included in the confidence interval. - the p value is greater than 0.05. - t in absolute value is lower than 1.96. We use 1.96 because its the reference value for the ttest at a 95% significant level for degrees of freedom. All the other variables analysed have an effect on wage. The variable experience is relevant for the analysis because its confidence interval at a significance level of 95% is from 0.094 to 0.24, so 0 is not included and thus we reject the null hypothesis (exper = 0). The players salary increases by 16.66% for every year of experience. Points per game have a positive impact on wage because the confidence interval at a 95% significance level goes from 0.047 to 0.075, and since 0 is not included in it, we reject the null hypothesis (points = 0). Salary increases 6.12% for every point scored per game, creating a system based on incentives. Draft number is relevant because the confidence interval at a .95% significance level goes from -0.016 to -0.008, and since 0 is not included in the interval we reject the null hypothesis (draft = 0) (Scully, G.W., 1973). For every unit increased in draft number, salary decreases 1.18%. Thus, dgraft number has a negative impact on wage: the larger the draft number, the lower the salary. The reason is that the later a player is drafter (that is, the larger the draft number, the worse the player is). Correlation matrix of coefficients of regress model:
e(V) marr exper points draft expersq _cons marr 1.0000 -0.1608 -0.1715 -0.3033 0.0669 0.1840 exper 1.0000 -0.2167 -0.0989 -0.9697 -0.5868 points draft expersq _cons

1.0000 0.5303 0.1677 -0.5364

1.0000 0.1120 -0.5299

1.0000 0.5481

1.0000

Looking at the coefficients of the correlation matrix we can see that there is homocedasticity, because if there were heterokedasticity, all values around de diagonal would be 0. For a model to be correctly estimated it has to be homocedastic, because the errors presented in all observations have a constant variance. If it presented heterokedasticity we would have to do a logarithmic transformation to be able to continue working with it. To further support our argument, we conduct a Cook-Weisberg test for

heteroskedasticity.

ECON2206 (z3389988, z3389750, z3389677)

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity Ho: Constant variance Variables: marr exper points draft expersq chi2(5) = 41.60 Prob > chi2 = 0.0000 Because the probability is less than the critical value (41.60), we cannot reject the null hypothesis (Ho: Constant variance) and thus the model presents homocedasticity. R squared is a statistical term saying how good one term is at predicting another. If R squared is 1.0 then given the value of one term, you can perfectly predict the value of another term. If R squared is 0.0 then knowing one term does not help you know the other term at all. More generally, a higher value of R squared means that you can better predict one term from another. It represents the goodness of fit of a model. However, since the use of R squared has some drawbacks, we use adjusted R squared that adjusts for the number of explanatory variables. Our adjusted R squared is 0.4948. Yet another way to prove that the homocedasticity of the model is: W = R squared * number of observations = 0.5054 * 240 = 121.296 The chi squared value for 239 degrees of freedom at a 5% significance level is approximately 287.88. Since W is smaller, we cannot reject the null hypothesis of homocedasticity. One of the reasons why marriage could affect players wages is that it might have an influence on the persons productivity. We will use points per game as the measure of productivity, and we perform a regression analysis using the variables that are likely to influence a persons productivity: marriage, experience, years played in college and children. points = 10.15009 + .365394marr -.1967665exper -.0790488expersq -1.161226coll +

1.303023children

points marr exper expersq coll children _cons

Coef. .1992865 1.332364 -.0790488 -1.161226 1.303023 10.15009

Std. Err. .7535223 .3374112 .0241076 .4636924 .764609 1.861952

t 0.26 3.95 -3.28 -2.50 1.70 5.45

P>|t| 0.792 0.000 0.001 0.013 0.090 0.000

[95% Conf. Interval] -1.284418 .6679927 -.1265173 -2.074248 -.2025116 6.483856 1.682991 1.996735 -.0315804 -.248204 2.808557 13.81632

10

ECON2206 (z3389988, z3389750, z3389677)

4. CONCLUSION
We can see that the t-statistic and coefficient level for marriage are low, and it has a high statistical error. We can conclude that the effect of marriage on productivity is negligible. It is surprising to see that the longer a person has played basketball in college, the lower its productivity (there is a negative relationship). This could be due to the fact that the best basketball players never make it to college because they get selected to play professional basketball in high school. After analyzing the various studies conducted by academics in the field, we see how they incorporate more variables and assumptions of their models. Therefore, their results are more interesting and enriching than ours. If we had more knowledge about the subject and had taken more data and assumptions we could have made interesting contrasts and achieved amazing and unexpected results.

11

ECON2206 (z3389988, z3389750, z3389677)

5. REFERENCES
Cornaglia, F. & Feldman, N.E., 2010. The Marriage Premium Revisited: The Case of Professional Baseball, The London School of Economics. Lazear, E., 1976. "Age, Experience and Wage Growth.", The American Economic Review, vol.66, no. 4, pp.548-558. Reed, W.R. & Harford, K., 1989, "The marriage premium and compensating wage differentials", Journal of Population Economics, vol. 2, pp. 237-265 Sanders Korenman, K., & Neumark, D., 1991, "Does Marriage Really Make Men More Productive?", The Journal of Human Resources, vol. 26, no. 2, pp. 282-307 Scully, G.W., 1973, Economic Discrimination in Professional Sports, Law and Contemporary Problems, vol. 38, no. 1, Athletics, pp. 67-84.

12

ECON2206 (z3389988, z3389750, z3389677)

6. APPENDIX
sumarize marr exper coll age points draft allstar children minutes expersq histogram wage lwage regress lwage marr exper coll age points draft allstar children minutes expersq regress lwage marr exper points draft expersq vce, corr regress lwage marr exper points draft expersq hettest marr exper points draft expersq regress points marr exper expersq coll children

13