1. Regression Continued: Functional Form LIR 832

2. Topics for the Evening 1. Qualitative Variables 2. Non-linear Estimation

3. Functional Form Not all relations among variables are linear: Our basic linear model: y=  0 +  1 X 1 +  2 X 2 +…+  k X k + e

4. Functional Form Q: Given that we are using OLS, can we mimic these non-linear forms? A: We have a small bag of tricks which we can use with OLS.

5. Functional Form

8. A first point about functional form: You must have an intercept. Consider the following case: We estimate a model and test the intercept to determine if it is significantly different than zero. We are not able to reject the null in a hypothesis test and we decide to re-estimate the model without an intercept. What is really going on? Return to our basic model: y=  0 +  1 X 1 +  2 X 2 +…+  k X k + e What are we doing when we remove the intercept? y=  +  1 X 1 +  2 X 2 +…+  k X k + e

9. Functional Form

11. /* Regression without an intercept */ Regression Analysis: weekearn versus years ed The regression equation is weekearn = 57.3 years ed 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Noconstant years ed 57.3005 0.1541 371.96 0.000 S = 534.450

12. Functional Form /* Regression with an intercept */ Regression Analysis: weekearn versus years ed The regression equation is weekearn = - 485 + 87.5 years ed 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Constant -484.57 18.18 -26.65 0.000 years ed 87.492 1.143 76.54 0.000 S = 530.510 R-Sq = 11.0% R-Sq(adj) = 11.0%

13. Functional Form Consequences of forcing through zero: Unless the intercept is really zero, we are going to bias both the intercept and the slope coefficients. Remember that we calculate the intercept so that the line passes through the point of means: Assures that the Σε = 0 If we impose 0 as the intercept, the line may not pass through the point of means and the sum of the errors may not equal zero. Biases the coefficients and leads to incorrect estimates of the standard errors of the βs. Never suppress the intercept, even if your theory suggests that it is not necessary.

14. Functional Form /* What About Those Residuals? */ Descriptive Statistics: RESI1, RESI2 Variable N N* Mean SE Mean StDev Minimum Q1 Median RESI1 47576 7582 -8.67 2.45 534.38 -1180.31 -359.12 -122.21 RESI2 47576 7582 0.00 2.43 530.50 -1329.77 -340.32 -107.62 Variable Q3 Maximum RESI1 218.59 2311.61 RESI2 237.69 2494.26

15. Functional Form Returning to the issue of non-linearity… In our basic model:  =  Y/  X = change in Y for a one-unit change in X Consider the effect of Education on base salary…

16. Functional Form Descriptive Statistics: years ed, Exp Variable N N* Mean SE Mean StDev Minimum Q1 Median Q3Maximum years ed 55158 0 15.734 0.00941 2.211 1.000 14.000 16.000 18.000 21.000 Exp 55107 51 21.644 0.0496 11.640 0.0000 13.000 22.000 30.000 76.000 Regression Analysis: weekearn versus years ed The regression equation is weekearn = - 485 + 87.5 years ed 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Constant -484.57 18.18 -26.65 0.000 years ed 87.492 1.143 76.54 0.000 S = 530.510 R-Sq = 11.0% R-Sq(adj) = 11.0%

17. Functional Form Now create a graph in MINITAB: Work in a new worksheet: Create values for years of education 0 - 21 Use the calculator to create the predicted weekly earnings. Use the scatterplot graphing function:

18. Functional Form Every year of education increases earnings by $87.49!

19. Functional Form Q: How do we estimate non-linear relations? A: We can use log transforms of variables to measure relations between variables as percentages rather than units. What is a log? What is a log transform? Take any number, let’s take 10. Then calculate b such that 10 = 2.71828 b. Then b is the log of 10. In this case b = 2.302585. You can do this on your calculator, in a spreadsheet, or in MINITAB.

20. Functional Form As your text shows: ln(100) = 4.605 100 = 2.71828 b ln(1000) = 6.9081000 = 2.71828 b ln(10,000) = 9.21010,000 = 2.71828 b ln(1,000,000) = 13.8161,000,000 = 2.71828 b We typically do not write 2.71828, rather we substitute e the natural base (there are also base 10 logs). So… 10 = e 2.302585 Some nice properties of log functions: ln(X*Y) = ln(X) + ln(Y) ln(X 2 ) = 2*ln(X)

21. Functional Form This property made it possible to manipulate very large numbers very easily and provides the foundation for slide rules and many modern computer calculations. Consider: 1,212,345*375,282 A real mess to do by hand Now consider the following transformation of this problem: ln(1,212,345*375,282) =ln(1,212,345) + ln(375,282) =14.008067 + 12.83543 = 26.8435 = 2.71828 26.8435 = antilog(26.8435) = 45,484,956.5078803

22. Functional Form The Shell presentation has an equation associated with an upward curve of: Earnings = 62988x 0.2676 Or… y=  0 X  1 We cannot estimate this in its current form using regression, but think about taking the log of each side: ln(y) = ln(  0 X  1 ) ln(y) = ln(  0 )+ln(X  1 ) ln(y) = ln(  0 )+  1 ln(X) So, if we take the log of each side, we get a linear equation that we can estimate!

23. Functional Form Consider the following equation: (single log equation) ln(weekearn) =  0 +  1 *YearsEd + e The interpretation of the coefficient on years of education is now the % change in base salary for a 1 year change in Education. How to do this in MINITAB: Calculate the log of weekly earnings Estimate the regression as…

24. Functional Form Regression Analysis: ln week earn versus years ed The regression equation is ln week earn = 4.87 + 0.109 years ed 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Constant 4.86646 0.02382 204.33 0.000 years ed 0.108980 0.001497 72.78 0.000 S = 0.694967 R-Sq = 10.0% R-Sq(adj) = 10.0% Analysis of Variance Source DF SS MS F P Regression 1 2558.4 2558.4 5297.03 0.000 Residual Error 47574 22977.3 0.5 Total 47575 25535.6

25. Functional Form Now we find that an additional year of education results in a 10.98% increase in salary. Interpretation is different from linear model r 2 is different between linear and log model. Linear: r 2 =11.0% Log:r 2 = 10.0% Does this mean the fit of the log model is worse than the linear model? No, cannot compare the two because you have transformed the equation. Fundamentally altered the variance of the dependent variable.

26. Functional Form Descriptive Statistics: weekearn, ln week earn Variable N N* Mean SE Mean StDev Minimum Q1 Median weekearn 47576 7582 894.53 2.58 562.22 0.01 519.00 769.23 ln week earn 47576 7582 6.5843 0.00336 0.7326 -4.6052 6.2519 6.6454 Variable Q3 Maximum weekearn 1153.00 2884.61 ln week earn 7.0501 7.967 What Does the Log Model Look Like? -- How to create a prediction in MINITAB & graph: Use regression equation to create estimated log wage from years of education data Exponentiate the predicted value using the MINITAB calculator Graph predicted wage against years of education

27. Functional Form

28. What is the equation underlying this model? Model of growth (such as compound interest)…

29. Functional Form Now lets try another approach, taking the log of both sides (double log equation): The interpretation of the coefficient on JEP is now the % change in base salary for a 1 % change in JEP. Note that this is an elasticity (which you will discuss in 809 in talking about supply and demand – the elasticity of labor demand with respect to the wage is the % change in the demand for labor for a 1% change in the wage).

30. Functional Form Regression Analysis: ln week earn versus ln ed The regression equation is ln week earn = 2.13 + 1.62 ln ed 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Constant 2.12844 0.06203 34.32 0.000 ln ed 1.62142 0.02254 71.93 0.000 S = 0.695775 R-Sq = 9.8% R-Sq(adj) = 9.8%

31. Functional Form

33. What is going on graphically? What are we really doing?

34. Functional Form

36. Q: How do we choose? A: Prior work and theory Is it sensible to measure as a linear model, or does one of these non-linear forms make better sense? Example: Thinking of the relationship between education and wages: wage = β 0 + β 1 *Years_of_Education ln(wage) = β 0 + β 1 *Years_of_Education ln(wage) = β 0 + β 1 *ln(Years_of_Education)

37. Functional Form What does prior work indicate? We typically use a log wage equation rather than a wage equation because… Turns out the error term is normally distributed in a log wage equation. More readily compared across models as it is not dependent on the scaling of the variable. Comparing the effect of education in percentage terms frees us from the effect of inflation and alternative currencies.

38. Functional Form A more general non-linear form (The Polynomial Form) Problem: Do we really believe that you get an additional 0.723% in weekly earnings for each year you get older. Hardly makes it worth getting older.

39. Functional Form Regression Analysis: ln(wkern) versus age, gender, edattain The regression equation is ln(wkern) = 2.41 + 0.00723 age - 0.368 gender + 0.105 edattain 47576 cases used 7582 cases contain missing values Predictor Coef SE Coef T P Constant 2.41075 0.06470 37.26 0.000 age 0.0072344 0.0002669 27.11 0.000 gender -0.368278 0.006115 -60.22 0.000 edattain 0.105032 0.001491 70.45 0.000 S = 0.6626 R-Sq = 18.2% R-Sq(adj) = 18.2% This model remains linear in ln(weekly earnings), each unit increase in age causes earnings to rise by 0.7%.

40. Functional Form It would be more reasonable to believe we will get a relationship which looks like: Why?

41. Functional Form How do we mimic this? Consider estimating the following linear regression: Notice that age enters twice, first as a linear term and then as a square. What does this model look like with real data?

42. Functional Form Regression Analysis: ln(wkern) versus age, age2, gender, edattain The regression equation is ln(wkern) = 0.927 + 0.104 age - 0.00113 age2 - 0.376 gender + 0.0948 edattain 47576 cases used 7582 cases contain missing values Predictor Coef SE Coef T P Constant 0.92706 0.06640 13.96 0.000 age 0.103919 0.001547 67.17 0.000 age2 -0.00112565 0.00001776 -63.37 0.000 gender -0.376012 0.005874 -64.01 0.000 edattain 0.094822 0.001441 65.82 0.000 S = 0.6363 R-Sq = 24.6% R-Sq(adj) = 24.6%

43. Functional Form Note that we now have two coefficients on Age: Age.103919 Age 2 -0.00112565 We know that the first term indicates that for each additional year our weekly earnings rise by 10.39%. But how do we chart out the second term. so that we have the full effect of age on earnings?

44. Functional Form

45. The effect of an additional year on earnings (formula for a polynomial model): If our model is: y =  0 +  1 X +  2 X 2 + …. Then  Y/  X =  1 +2*  2 *X First issue, look at the prediction of ln weekly earnings based on age (leave all other variables at their mean).

46. Functional Form

48. What about the ‘marginal effect’ of age? What is the effect on income of getting an additional year older? Obviously varies with how old you are. Things are pretty good when you are young Two ways of obtaining this: 1. Calculate the difference in the total effect of age for any two years. Age221.741 Age211.686 Diff0.055 or + 5.5%

49. Functional Form 2. Alternatively, use the polynomial formula:

50. Functional Form What is the increase in earnings at age 21?.103919 -.0022513*21 =0.056642 What about age 25?.103919 -.0022513*25 =0.0476365 What about age 50? (Class work) Note that the effect of an additional year of education is no longer constant, it depends on how old you are.

51. Functional Form

52. The gains to aging are greatest when you are youngest: They decline steadily as you age. By age fifty your earnings are falling as you get older (oops!). A couple points about polynomial and functional forms: Polynomial forms have the strength of letting the data tell you if the relationship is linear or not. If it is, the coefficient on X 2 will be 0 or very close to it. You cannot compare r 2 across log and non-log forms because it changes the dependent variable and the sum of squares. You can between linear and non-linear forms.

53. Recap on Functional Form Not all relationships are linear Regression allows us to estimate non- linear models and to let the data tell us whether we should be using a non-linear form Single and double log transforms Polynomial form

54. MultiCollinearity Issue: What happens when two variables contain the same, or almost the same information? Condition is called multicollinearity

55. Perfect MultiCollinearity Is Not a Problem Try putting both a Male and Female dummy variable in a wage equation

56. Base Regression: Earnings=F(age, Education) Regression Analysis: weekearn versus years ed, age The regression equation is weekearn = - 707 + 83.5 years ed + 6.87 age Predictor Coef SE Coef T P Constant -706.63 19.24 -36.73 0.000 years ed 83.463 1.137 73.38 0.000 age 6.8717 0.2118 32.45 0.000 S = 524.739 R-Sq = 12.9% R-Sq(adj) = 12.9%

57. Now Put Male & Female Into Model Regression Analysis: weekearn versus years ed, age, Male, Female * Female is highly correlated with other X variables * Female has been removed from the equation.

58. The Regression The regression equation is weekearn = - 720 + 76.4 years ed + 6.29 age + 319 Male Predictor Coef SE Coef T P Constant -720.28 18.35 -39.25 0.000 years ed 76.432 1.089 70.16 0.000 age 6.2874 0.2021 31.11 0.000 Male 318.522 4.625 68.87 0.000 S = 500.391 R-Sq = 20.8% R-Sq(adj) = 20.8%

59. Male & Female Contain the Same Information Correlations: Male, Female Pearson correlation of Male and Female = -1.000 P-Value = *

60. What If Several Variables Contain the Same Information Regression Analysis: weekearn versus age, years ed, Female, NE, MW, S, W * W is highly correlated with other X variables * W has been removed from the equation. The regression equation is weekearn = - 392 + 6.25 age + 75.9 years ed - 318 Female + 47.7 NE - 18.2 MW - 20.3 S 47576 cases used, 7582 cases contain missing values Predictor Coef SE Coef T P Constant -392.10 19.21 -20.42 0.000 age 6.2532 0.2019 30.98 0.000 years ed 75.895 1.089 69.67 0.000 Female -318.406 4.619 -68.93 0.000 NE 47.658 6.768 7.04 0.000 MW -18.155 6.594 -2.75 0.006 S -20.323 6.317 -3.22 0.001 S = 499.701 R-Sq = 21.0% R-Sq(adj) = 21.0%

61. What Are the Regional Dummies Correlated With? Descriptive Statistics: NE, MW, S, W Variable N N* Mean SE Mean StDev Minimum Q1 Median NE 55158 0 0.22310 0.00177 0.41633 0.00000 0.00000 MW 55158 0 0.23873 0.00182 0.42631 0.00000 0.00000 S 55158 0 0.29211 0.00194 0.45474 0.00000 0.00000 W 55158 0 0.24606 0.00183 0.43072 0.00000 0.00000

62. Imperfect MultiCollinearity Two or more variables contain similar but not identical information

63. Log Wage Regression Source | SS df MS Number of obs = 156130 -------------+------------------------------ F( 11,156118) = 4227.42 Model | 11630.4798 11 1057.31635 Prob > F = 0.0000 Residual | 39046.5066156118.250108934 R-squared = 0.2295 -------------+------------------------------ Adj R-squared = 0.2294 Total | 50676.9864156129.324584071 Root MSE =.50011 ------------------------------------------------------------------------------ lnwage3 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age |.0712402.0005528 128.87 0.000.0701567.0723237 age2 | -.0007535 6.58e-06 -114.54 0.000 -.0007664 -.0007406 female | -.1999096.0025452 -78.54 0.000 -.2048982 -.1949211 married |.0947973.0028481 33.28 0.000.089215.1003796 black | -.1314511.0043814 -30.00 0.000 -.1400385 -.1228637 other | -.0063689.0057833 -1.10 0.271 -.0177041.0049663 NE |.0328108.0038223 8.58 0.000.0253191.0403024 Midwest |.007487.0036482 2.05 0.040.0003367.0146373 South | -.0204817.0035696 -5.74 0.000 -.027478 -.0134854 city1mil |.1440377.0026054 55.28 0.000.1389312.1491443 union2 |.1358151.0037783 35.95 0.000.1284097.1432205 _cons |.9784856.0107005 91.44 0.000.9575129.999458

64. Switch CBC for Union Source | SS df MS Number of obs = 156130 -------------+------------------------------ F( 11,156118) = 4242.43 Model | 11662.2696 11 1060.20633 Prob > F = 0.0000 Residual | 39014.7168156118.249905307 R-squared = 0.2301 -------------+------------------------------ Adj R-squared = 0.2301 Total | 50676.9864156129.324584071 Root MSE =.49991 ------------------------------------------------------------------------------ lnwage3 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age |.0710808.0005528 128.59 0.000.0699974.0721642 age2 | -.000752 6.58e-06 -114.34 0.000 -.0007649 -.0007391 female | -.2003086.0025431 -78.77 0.000 -.205293 -.1953242 married |.0946468.002847 33.24 0.000.0890668.1002269 black | -.1321203.0043799 -30.17 0.000 -.1407048 -.1235358 other | -.0061873.005781 -1.07 0.284 -.0175179.0051434 NE |.033546.0038197 8.78 0.000.0260595.0410324 Midwest |.0079032.0036465 2.17 0.030.000756.0150503 South | -.0200437.003568 -5.62 0.000 -.0270369 -.0130504 city1mil |.1442921.0026043 55.41 0.000.1391878.1493965 cbc2 |.1363582.0036181 37.69 0.000.1292668.1434495 _cons |.9799436.0106968 91.61 0.000.9589782 1.000909 ------------------------------------------------------------------------------

65. Use Union & CBC Source | SS df MS Number of obs = 156130 -------------+------------------------------ F( 12,156117) = 3889.14 Model | 11662.8996 12 971.908303 Prob > F = 0.0000 Residual | 39014.0867156117.249902872 R-squared = 0.2301 -------------+------------------------------ Adj R-squared = 0.2301 Total | 50676.9864156129.324584071 Root MSE =.4999 ------------------------------------------------------------------------------ lnwage3 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age |.0710741.0005528 128.58 0.000.0699907.0721575 age2 | -.0007519 6.58e-06 -114.32 0.000 -.0007648 -.000739 female | -.2001837.0025443 -78.68 0.000 -.2051704 -.1951969 married |.0946413.002847 33.24 0.000.0890612.1002213 black | -.1321795.00438 -30.18 0.000 -.1407643 -.1235947 other | -.0061938.005781 -1.07 0.284 -.0175244.0051367 NE |.0333811.0038211 8.74 0.000.0258919.0408703 Midwest |.0078341.0036468 2.15 0.032.0006864.0149817 South | -.0199589.0035684 -5.59 0.000 -.0269529 -.0129649 city1mil |.1442482.0026044 55.39 0.000.1391436.1493528 union2 |.0175444.0110493 1.59 0.112 -.0041121.0392008 cbc2 |.1205632.0105851 11.39 0.000.0998166.1413098 _cons |.9800641.010697 91.62 0.000.9590982 1.00103

66. Consequences of MultiCollinearity Estimates remain unbiased Variances and Standard Errors Increase Computed t-scores fall Estimates will be very sensitive to specification Overall fit of the model (r-square) will be unaffected Predictions are also unaffected

67. What Is the Issue Where there is MultiCollinearity, we need to be careful about interpreting results Can be misleading about effect of variables

68. Detecting Collinearity High correlation between variables Issue: multiple variables are collectively collinear (region example) Variance Inflation Factor Regress each explanatory variable on all other explanatory variables Calculate

69. How Do We Calculate the VIF? Regression Analysis: age versus years ed, Female, NE, MW, S, W * W is highly correlated with other X variables * W has been removed from the equation. The regression equation is age = 35.8 + 0.480 years ed - 1.59 Female + 0.098 NE - 0.617 MW - 0.204 S Predictor Coef SE Coef T P Constant 35.7977 0.3712 96.43 0.000 years ed 0.47978 0.02241 21.41 0.000 Female -1.59360 0.09896 -16.10 0.000 NE 0.0979 0.1443 0.68 0.498 MW -0.6174 0.1416 -4.36 0.000 S -0.2044 0.1349 -1.52 0.130 S = 11.5764 R-Sq = 1.5% R-Sq(adj) = 1.5%

70. It’s a Different Story with Regional Variables Regression Analysis: NE versus age, years ed, Female, MW, S, W The regression equation is NE = 1.00 + 0.000000 age + 0.000000 years ed + 0.000000 Female - 1.00 MW - 1.00 S - 1.00 W Predictor Coef SE Coef T P Constant 1.00000 0.00000 * * age 0.00000000 0.00000000 * * years ed 0.00000000 0.00000000 * * Female 0.00000000 0.00000000 * * MW -1.00000 0.00000 * * S -1.00000 0.00000 * * W -1.00000 0.00000 * * S = 0 R-Sq = 100.0% R-Sq(adj) = 100.0%

71. CBC Has A High VIF. reg cbc2 age age2 female married black other NE Midwest South city1mil union2 Source | SS df MS Number of obs = 161792 -------------+------------------------------ F( 11,161780) =. Model | 18165.9762 11 1651.45238 Prob > F = 0.0000 Residual | 2301.31742161780.014224981 R-squared = 0.8876 -------------+------------------------------ Adj R-squared = 0.8876 Total | 20467.2936161791.126504525 Root MSE =.11927 ------------------------------------------------------------------------------ cbc2 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age |.0013903.0001288 10.80 0.000.0011379.0016426 age2 | -.0000133 1.53e-06 -8.72 0.000 -.0000163 -.0000103 female |.0025409.0005963 4.26 0.000.0013722.0037096 married |.0013089.0006676 1.96 0.050 4.52e-07.0026174 black |.0063441.001032 6.15 0.000.0043214.0083668 other | -.0016395.0013597 -1.21 0.228 -.0043046.0010255 NE | -.0043777.000895 -4.89 0.000 -.0061319 -.0026234 Midwest | -.0027157.0008563 -3.17 0.002 -.0043941 -.0010374 South | -.0041338.0008356 -4.95 0.000 -.0057716 -.0024961 city1mil | -.0018596.0006102 -3.05 0.002 -.0030555 -.0006636 union2 |.9811512.0008888 1103.92 0.000.9794092.9828932 _cons | -.013585.0025048 -5.42 0.000 -.0184943 -.0086757

72. What To Do About MultiCollinearity Do Nothing Get More Data We had 156,000 observations for the wage regressions Drop the Redundant Variable Care needed in interpretation

73. Compare Specification Issues OmittedExtraneousMultiCollinearity Added VariableRight signed & Large in Magnitude Coefficient close to zero Right or wrong signed SignificanceHighly SignificantNon-significantWeak or n.s. Other CoefChange signLittle ChangePossibly change sign SignificanceRemains singificantLittle ChangeBecomes weak or n.s. R-squareIncrease alotLittle change New SampleLittle Difference Unstable Estimates