I have been reading through Feyrer and Sacerdote's paper on the Stimulus (

available here), and there were a couple points. It seems to me that their state level results that are very dependent on a couple very small states, namely North Dakota and Vermont. It appears that North Dakota is particularly problematic because

60 percent the job growth that occurred during the time period that you studied was simply due to jobs in the oil industry boom (and presumably there was some secondary job creation from that expansion), jobs which would seem difficult to relate to the Stimulus. In addition, you weight your county level regressions by population, but if you do the same thing for the state level regressions, the results seem to go away. I have used slightly different months, but I believe that the pattern is the same. Sorry that this is hard to read, but the coefficients and t-statistics are in bold.

. reg changeemppop122010to022009 recoveryfundsawardedpercapita if state~="DC"

Source | SS df MS Number of obs = 50

-------------+------------------------------ F( 1, 48) = 5.94

Model | 5.2409e-10 1 5.2409e-10 Prob > F = 0.0186

Residual | 4.2381e-09 48 8.8295e-11 R-squared = 0.1101

-------------+------------------------------ Adj R-squared = 0.0915

Total | 4.7622e-09 49 9.7188e-11 Root MSE = 9.4e-06

------------------------------------------------------------------------------

c~op12100209 |

Coef. Std. Err.

t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

recoveryfu~a |

7.69e-09 3.16e-09

2.44 0.019 1.34e-09 1.40e-08

_cons |

-.0000224 3.48e-06

-6.44 0.000 -.0000294 -.0000154

------------------------------------------------------------------------------

. reg changeemppop122010to022009 recoveryfundsawardedpercapita if state~="DC" & state~="North Dakota" & state~="Vermont"

Source | SS df MS Number of obs = 48

-------------+------------------------------ F( 1, 46) = 1.65

Model | 1.0780e-10 1 1.0780e-10 Prob > F = 0.2060

Residual | 3.0137e-09 46 6.5516e-11 R-squared = 0.0345

-------------+------------------------------ Adj R-squared = 0.0135

Total | 3.1215e-09 47 6.6415e-11 Root MSE = 8.1e-06

------------------------------------------------------------------------------

c~op12100209 |

Coef. Std. Err.

t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

recoveryfu~a |

3.71e-09 2.89e-09

1.28 0.206 -2.11e-09 9.52e-09

_cons |

-.0000193 3.09e-06

-6.25 0.000 -.0000256 -.0000131

------------------------------------------------------------------------------

. reg changeemppop122010to022009 recoveryfundsawardedpercapita [aweight= pop2010] if state~="DC"

(sum of wgt is 3.0918e+08)

Source | SS df MS Number of obs = 50

-------------+------------------------------ F( 1, 48) = 0.54

Model | 2.5924e-11 1 2.5924e-11 Prob > F = 0.4649

Residual | 2.2924e-09 48 4.7759e-11 R-squared = 0.0112

-------------+------------------------------ Adj R-squared = -0.0094

Total | 2.3183e-09 49 4.7313e-11 Root MSE = 6.9e-06

------------------------------------------------------------------------------

c~op12100209 |

Coef. Std. Err.

t P>|t| [95% Conf. Interval]

-------------+----------------------------------------------------------------

recoveryfu~a |

3.29e-09 4.47e-09

0.74 0.465 -5.69e-09 1.23e-08

_cons |

-.0000182 3.97e-06

-4.58 0.000 -.0000262 -.0000102

------------------------------------------------------------------------------

Labels: book, Research, stimulus