More Economic Fun: Alcohol Consumption per Capita versus GDP Per Capita by US State

by robekulick

This is a follow up to an analysis of alcohol consumption by state versus GDP per capita:

https://robekulick.wordpress.com/2011/08/02/some-random-economic-fun/

After seeing this, my very clever girlfriend suggested I should do the same thing for the states in the US. So here it is!

First, the results are surprisingly similar to those for the country level analysis.

As a starting point the raw correlation between per capital alcohol consumption and state GDP per capita was 0.5047. Correlation is a measure that can take on values from -1 to 1, so 0.5047 is a fairly strong positive correlation.

I also did two simple univariate regression analyses, one of alcohol consumption versus per capita gdp and one taking the natural logs of each variable. Using natural logs allows us to interpret the data in percentage terms.

Regression One

. reg alcohol gdp

      Source |       SS       df       MS              Number of obs =      51
-------------+------------------------------           F(  1,    49) =   16.75
       Model |   3.3941235     1   3.3941235           Prob > F      =  0.0002
    Residual |  9.93109218    49  .202675351           R-squared     =  0.2547
-------------+------------------------------           Adj R-squared =  0.2395
       Total |  13.3252157    50  .266504314           Root MSE      =  .45019

------------------------------------------------------------------------------
     alcohol |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         gdp |    .000015   3.67e-06     4.09   0.000     7.65e-06    .0000224
       _cons |   1.784646   .1727301    10.33   0.000     1.437532     2.13176
------------------------------------------------------------------------------

Regression Two

. reg lnalcohol lngdp

      Source |       SS       df       MS              Number of obs =      51
-------------+------------------------------           F(  1,    49) =   14.83
       Model |  .464823212     1  .464823212           Prob > F      =  0.0003
    Residual |  1.53556643    49   .03133809           R-squared     =  0.2324
-------------+------------------------------           Adj R-squared =  0.2167
       Total |  2.00038964    50  .040007793           Root MSE      =  .17703

------------------------------------------------------------------------------
   lnalcohol |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       lngdp |   .3686876   .0957306     3.85   0.000     .1763097    .5610655
       _cons |  -3.051485   1.019307    -2.99   0.004    -5.099859    -1.00311

For those of you unfamiliar with reading STATA output (which would include 99% of normal people), the first regression has a t-statistic of 4.09 indicating that the positive connection between alcohol and per capita gdp is not simply attributable to random chance. The second regression suggests that overall a 1% increase in per capita gdp for a state is associated with a 0.369% increase in per capita alcohol consumption. Of course, one can’t assess causality from these regressions because one would have to control for a number of additional variables that effect alcohol consumption. But the connection is pretty interesting, especially considering its similarity with the country data.

Casual observation of the data suggested to me that cold states seem to drink a lot. I might update this analysis soon with a temperature variable. I also think it would be interesting to add a variable for blue states versus red states. I know you all can’t wait (at least I’m excited about it)!


					
Advertisements