INFORMATIVE PRIOR
DIFFUSE PRIOR
|
NORTH |
CENTER |
SOUTH |
NORTH |
CENTER |
SOUTH | |
|
σ###KL |
0.14255 |
0.05044 |
0.03586 |
-0.12257 |
-0.24246 |
-0.32245 |
|
st. err. |
0.11011 |
0.08743 |
0.10769 |
0.35188 |
0.30326 |
0.23794 |
|
σ⅛⅛⅛KK |
-0.27686 |
-0.10627 |
-0.06685 |
0.23805 |
0.51019 |
0.60109 |
|
st. err. |
0.21385 |
0.18419 |
0.20075 |
0.44355 |
0.63890 |
0.44355 |
|
σ≡zr^ |
-0.07340 |
-0.02394 |
-0.01924 |
0.06311 |
0.11494 |
0.17298 |
|
st. err. |
0.05670 |
0.04150 |
0.05777 |
0.18118 |
0.14394 |
0.12764 |
|
ε≡KΓ- |
0.09410 |
0.03420 |
0.02334 |
0.08091 |
0.16422 |
0.20987 |
|
st. err. |
0.07268 |
0.05929 |
0.07009 |
0.23228 |
0.20565 |
0.15487 |
|
ε≡LK- |
0.04845 |
0.01624 |
0.01252 |
0.04166 |
0.07795 |
0.11258 |
|
st. err. |
0.03743 |
0.02814 |
0.03760 |
0.11960 |
0.09761 |
0.08308 |
|
£###KK |
-0.09410 |
-0.03420 |
-0.02334 |
0.08091 |
0.16422 |
0.20987 |
|
st. err. |
0.07273 |
0.05929 |
0.07009 |
0.23228 |
0.20565 |
0.15487 |
|
ε≡zr^ |
-0.04845 |
-0.01624 |
-0.01252 |
0.04166 |
0.07795 |
0.11258 |
|
st. err. |
0.03743 |
0.02814 |
0.03760 |
0.11960 |
0.09761 |
0.08308 |
TABLE 2 - Elasticities conditional on 1989 labor share
It’s striking to note that all the expected values for price and substitution
elasticities with a diffuse prior have the "wrong" sign. Expected own elasticities are
positive and cross elasticities are negative, against neoclassical theory. However, using
the usual sampling theory language, we should immediately add that we cannot reject
the hypothesis that they are not significantly different from zero. It’s not clear how
useful is such a statement, but, in our framework, we can easily figure out the
probability that elasticities are greater (or lower) than zero. In the same fashion we can
calculate posterior odds (or Bayes Factor, if we assign equal prior probabilities to both
hypotheses) in favor of the (marginal) regularity condition, as discussed above. At any
rate Bayes Factor with a symmetric distribution are always against regularity (less than
one) since expected values have the "wrong" sign. Therefore they are not presented here,
and visual inspection of such distributions (the dotted ones) depicted in Fig. B - H is
12