Laserfiche WebLink
TECHNICAL APPENDIX <br />To examine the relationship between <br />market-rate housing construction and displacement <br />of low-income households we developed a simple <br />econometric model to estimate the probability of a <br />low-income Bay Area neighborhood experiencing <br />displacement. <br />Data. We use data on Bay Area census tracts <br />(small subdivisions of a county typically containing <br />around 4,000 people) maintained by researchers <br />with the University of California (UC) Berkeley <br />Urban Displacement Project. This dataset included <br />information on census tract demographics, housing <br />characteristics, and housing construction levels. We <br />focus on data for the period 2000 to 2013. <br />Defining Displacement. Researchers have <br />not developed a single definition of displacement. <br />Different studies use different measures. For our <br />analysis, we use a straightforward yet imperfect <br />definition of displacement which is similar to <br />the definition used by UC Berkeley researchers. <br />Specifically, we define a census tract as having <br />experienced displacement if (1) its overall <br />population increased and its population of <br />low-income households <br />decreased or (2) its overall <br />population decreased and <br />its low-income population <br />declined faster than the <br />overall population. <br />Our Model. We <br />use probit regression <br />analysis to evaluate how <br />various factors affected <br />the likelihood of a <br />census tract experiencing <br />displacement between 2000 and 2013. This type <br />of model allows us to hold constant various <br />economic and demographic factors and isolate <br />the impact of increased market-rate construction <br />on the likelihood of displacement. The results <br />of our regression are show in Figure A1. <br />Coefficient estimates from probit regressions are <br />not easily interpreted. While the fact that the <br />coefficient for market-rate housing construction <br />is statistically significant and negative suggests <br />that more construction reduces the likelihood <br />of displacement, the magnitude of this effect <br />is not immediately clear. To better understand <br />these results, we used the model to compare the <br />probability that an average census tract would <br />experience displacement when its market-rate <br />construction was low (0 units), average (136 units), <br />and high (243 units). As shown in Figure A2 (see <br />next page), with low construction levels, a census <br />tract’s probability of experiencing displacement was <br />47 percent, compared to 34 percent with average <br />construction levels, and 26 percent with high <br />construction levels. <br />Figure A1 <br />Regression Results <br />Dependent Variable: Did Displacement Occur (Yes=1 and No=0)? <br />Independent Variable Coefficient Standard Error <br />Number of market-rate housing units built -0.002370.00043 <br />Share of population that is low income 1.740750.54137 <br />Share of population that is nonwhite -0.612130.29151 <br />Share of adults over 25 with a college <br />degree <br />1.900540.38599 <br />Population density -0.000010.00000 <br />Share of housing built before 1950 1.165060.22569 <br />Constant -1.458860.33420 <br /> www.lao.ca.gov Legislative Analyst’s Office 13 <br />AN LAO BRIEF 8.A - Page 21