Hello,
I have downloaded data from the Barro-Lee educational attainment data set and merged this with data from the World Bank on GDP per capita, GDP growth and Gross Capital Formation.
I am currently running a fixed effects model testing the effects of human capital accumulation (education) on GDP growth, using an augmented Solow model. My education variables are no schooling (the percentage of the population aged over 25 who have not completed any schooling), primary education (the percentage of the population aged over 25 who have completed primary education), secondary education (the percentage of the population aged over 25 who have completed secondary education) and tertiary education (the percentage of the population aged over 25 who have completed tertiary education). As well as these variables I also have gross capital formation, population and lagged GDP. I have taken logs of GDP per capita and gross capital formation. The time period is 1965-2010 delta 5 years.
When I run the following command:
areg gdpgrowth laggdppc NoSchool PrimaryTotal SecondaryTotal TertiaryTotal laggcf Population1000s i.Year, cluster(Country) absorb(Country)
I get statistically significant results, but both the coefficients and standard errors for education appear to be very similar. Is this due to my education variables being correlated?
Ideally what I am trying to discover is whether different levels of human capital accumulation increase GDP growth over my sample period.
Linear regression, absorbing indicators Number of obs = 924
Absorbed variable: Country No. of categories = 136
F( 15, 135) = 61.44
Prob > F = 0.0000
R-squared = 0.6223
Adj R-squared = 0.5490
Root MSE = 0.2632
(Std. Err. adjusted for 136 clusters in Country)
---------------------------------------------------------------------------------
| Robust
gdpgrowth | Coef. Std. Err. t P>|t| [95% Conf. Interval]
----------------+----------------------------------------------------------------
laggdppc | -.5783381 .0724007 -7.99 0.000 -.7215245 -.4351517
NoSchool | .2758305 .0528499 5.22 0.000 .1713096 .3803514
PrimaryTotal | .2712908 .0524151 5.18 0.000 .16763 .3749517
SecondaryTotal | .2769055 .0524878 5.28 0.000 .1731008 .3807102
TertiaryTotal | .2843725 .0520174 5.47 0.000 .1814981 .3872469
laggcf | .0464763 .0473452 0.98 0.328 -.0471579 .1401104
Population1000s | 1.73e-06 4.41e-07 3.91 0.000 8.53e-07 2.60e-06
|
Year |
1975 | .4948126 .0469035 10.55 0.000 .4020519 .5875733
1980 | .703068 .0689545 10.20 0.000 .5666973 .8394387
1985 | .3157352 .0817349 3.86 0.000 .1540888 .4773816
1990 | .6829648 .0842277 8.11 0.000 .5163883 .8495413
1995 | .6718469 .1026378 6.55 0.000 .4688609 .8748329
2000 | .558993 .1170705 4.77 0.000 .3274637 .7905224
2005 | .9363778 .1263443 7.41 0.000 .6865075 1.186248
2010 | 1.123591 .1467162 7.66 0.000 .833432 1.413751
|
_cons | -24.63292 5.343615 -4.61 0.000 -35.20095 -14.0649
---------------------------------------------------------------------------------
Related Posts with Multicollinearity in a panel regression
Possible to put interaction terms of X and indictors of Y quintile in a regression?Dear Statalist users, I am wondering if I would estimate a regression like this: Y = X + X * Y_Q2 …
create a loop to convert string variables in numeric variablesHello, I am Salvatore, happy to join the Stata Forum community. I am a new user who recently started…
Matching without replacement from a file of pairs for case-control and other applicationsShort version: I’m seeking a solution to how to do 1:m matching of cases and controls without replac…
how to replace/recode variables that are a multiple of some numberHi I have a variable "level" that ranges from 1 to 125 I would like to recode it into a variable "pi…
Shortcut for generating new variables based on many existingI have groups of variables contactoutcome_con_<y>_<x> and contactmethod_con_<y>_&l…
Subscribe to:
Post Comments (Atom)
0 Response to Multicollinearity in a panel regression
Post a Comment