06.01.2015 Views

413047-Underground-Commercial-Sex-Economy

413047-Underground-Commercial-Sex-Economy

413047-Underground-Commercial-Sex-Economy

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

in column 3 of the table below. The coefficient obtained for coordinate 9 is the value<br />

represents the estimated distance decay exponent of the attraction force.<br />

Table 3.6 Proxy for Pimp Population Size<br />

, which<br />

City<br />

Atlanta 6.254 1.192<br />

Dallas 6.079 1.108<br />

Denver 3.379 0.156<br />

Miami 8.049 1.785<br />

San Diego 5.811 0.975<br />

Seattle 5.463 0.805<br />

DC 5.861 1.000<br />

Kansas City N/A N/A<br />

By including the distance between a UCSE participant’s home city and other candidate cities in the form<br />

of the “force of attraction,” the Gravity Model yields a proxy for pimp population size which deftly<br />

sidesteps potential biases that may arise from assuming that the UCSE participants are blind to distance<br />

(as was indeed the case for the initial proxy ). In what follows, we use this as our proxy for pimp<br />

population size, and then combine with the mean weekly gross cash intake to produce<br />

. The computed final values of the sex proxy are given in the table<br />

below. Note that these proxy values are without units and have no inherent interpretation.<br />

Table 3.7 Final Values of Proxy<br />

Year Atlanta Dallas Denver Miami San Diego Seattle DC<br />

2003 30,835 12,877 6,117 39,141 16,090 5,433 16,700<br />

2007 39,128 13,320 4,852 31,666 10,852 14,489 11,588<br />

1.2.2. Linear Proxy<br />

Originally, we envisioned using a single variable as a proxy for cross-city and over-time drug market<br />

differences. In reading the relevant literature, however, we developed a list of potential proxies. Rather<br />

than choosing one, we used regressions to aggregate across different variables.<br />

Let be the value of the proxy variable for drug activity, measured in city at time . For this<br />

analysis, we used, , , and . That is, Kansas City is excluded.<br />

Let be the standard error of the estimate of the proxy for city at time . To aggregate across the 7<br />

different drug proxy variables, we relied on a simple linear regression. While this imposes a lot of<br />

structure on the proxy information, we believe that it eliminates some of the concerns about noisy<br />

measurement. Our plan was to use a single regression for each city, with all proxies modeled as a linear<br />

function of time. The problem is that the different proxy variables are in different units, so a linear<br />

regression cannot properly combine them.<br />

34

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!