|
from equation (3). We find that:
<br />AUM = Aut',Rent +DUtM'Payoff +AUy"NC (13)
<br />+QUtM,MC + AUMt,Miles + QUtM,Amenity + AUIV1t,Logit.
<br />That is, the average utility difference between the treatment group and the control arises from
<br />differences in average rent paid AutM,Rent in transfers received from landlords AUt f'PayoffI in
<br />accumulated neighborhood capital AU; f'Ne, in fixed costs AUtu'Mc in variable moving costs
<br />�UM,Miles in neighborhood amenity values AutM,Araenity and in idiosyncratic valuations
<br />Q UM,Logit Suppressing the dependence of j and T on x, we can formally write these terms
<br />as:
<br />AUt"at — E )7 ryM exp (Rt (.i, d, Th)) 1pt (x, Bt -1) - pt (x, Bt -1)/
<br />dt-i x
<br />AUt ayoff = At (x, dt-1,M) (pi (x, Bt -1) -pt (x, Bt -1))
<br />AUM'Nc = aMTa (Pt (x, Bt-].) - Pt (x, Bt -1))
<br />Bt_i x
<br />DUtvr,Mc - Y, E t00,M1 [x =� S] (pi (x, Bt -1) - Pt (x, Bt -1)J
<br />yt-1 X
<br />AUtj'V Mi•le.s = d 1 x S T x B_ c x B_
<br />t L� Pd,M yft- � j �pt ( 1) -pt (, t 1))
<br />et-, X
<br />M,Anenity C `
<br />DUt = Wjt �ptT (•x, Bt -1) — pt (x, Bt -1)J
<br />Ot-1 a
<br />We can measure each of these terms." We recover estimates of ym, AM, am, cpam, ,pd,M,
<br />and Wft from our structural model. We can then recover the other needed quantities from
<br />standard reduced form differences -in -differences analysis. For example,
<br />Eet_i E. exp (Rt (j, d, Th)) (pt (x, Bt -1) - pt (x, Bt -1)) is simply the average difference in
<br />rents paid between treatment and control in year t, Eot-1 EX Tn, (pt (x, Bt -1) - pt (x, Bt -1))
<br />17Since we measure rents as monthly rents/3000, we multiply by 36,000 to convert to an annual rent
<br />number.
<br />36
<br />
|