To study the effect of poverty, we have collected the data of 58 cities in California.

Consider the following equation povrate = 1+2urb + 3famsize + Lunemp + 5highschl + mcollege + Lmedinc + where povrate=poverty rate, urb=urban population percentage, famsize=family size, unemp=unemployment rate, highschl=the percentage of people with high school diploma who are 25 or older, college=the percentage of people with University or higher education who are 25 or older, medinc=family's medium income.

Table 1 presents the regression results:

Table 1

VariableCoefficientStd. Errort-StatisticProb.

16.81757

8.502563

1.977941

0.0534

urb

-0.018735

0.014757

-1.269543

0.2100

famsize

6.091761

1.881073

3.238450

0.0021

unemp

-0.011796

o. 119457

-0.098749

0.9217

highschl

-0.118552

0.068100

-1.740840

0.0877

college

0.171106

0.098165

1.743033

0.0874

medinc

-0.535992

0.070354-7.618501

0.0000

R-squared

Adjusted R-squared

Durbin-Watson stat

0.836184

0.816911

1.903818

Akaike info criterion

Schwarz criterion

Hannan-Quinn criter.

4.003044

4.251718

4.099907

Table 2

urb famsize unemp highschl college medinc

urb

1.000000

0.350264

o. 109609

0.210622

-0.358034

-0.084427

famsize

0.350264

1.000000

0.484816

-0.507556

-0.299949

-0.034907

unemp

0.109609

0.484816

1.000000

-0.108688

-0.756632

-0.714417

highschl

0.210622

-0.507556

-0.108688

1.000000 -0.357567

-0.280333

college

-0.358034

-0.299949

-0.756632

-0.357567

1.000000

0.847734

medinc

-0.084427

-0.034907

-0.714417

-0.280333

0.847734

1.000000

a)Reading Table 1, do you think there exist multicollinearity in the data? If so, why? (10 points)

b)To test the possibility of multieollinearity, we have calculated pair-wise correlations, and the results are shown in Table 2. If we were to eliminate one variable from the data, which one do you think should be taken out, and why (you may combine the results from table

1&2)? (10 points)