If you move, I move: The social influence effect on residential mobility

Abstract

There are many theories that account for why households move between residential areas. In this paper, we advance on this by formulating a new mechanism whereby a household's probability of leaving a neighborhood is informed by the number of other households who have previously left that neighborhood. We call this mechanism: the social influence (SI) effect. By applying matching to Swedish register data for Stockholm County (1998-2017), and after adjusting for theoretically relevant confounders from the existing literature, we find that SI has a significant effect on neighborhood out-mobility. Furthermore, we find that the SI effect is moderated by the visibility with which others' behaviors is observed, measured as the number of previous out-movers, the distance to ego, and its salience in the social environment. Our study also discusses some ways in which SI might be entangled with other mechanisms, and outlines future directions from which studies of residential segregation dynamics might be approached.

Fig 1. The SI effect on residential out-mobility.

Difference in the probability of moving-out between natives exposed to out-movers and no in-movers (X = 1) and natives exposed to no change (neither in- nor out-movers) (X = 0). The estimates are the β coefficient from the LPM on the raw data (in white) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-value in parenthesis. The straight horizontal line indicates no effect.

Fig 2. The marginal effect of the number of previous out-movers on the strength of SI.

Difference between natives who are exposed only to 2, 3, or 4 or more previous out-movers (X = 1, left), or fewer than this (X = 0, right), respectively. The estimates are the β coefficient from the LPM on the raw data (in white) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-values in parentheses. The straight horizontal line indicates no effect. The upper bars indicate whether coefficients across models are statistically significantly different from one another. ‘***’ p-value<0.001, ‘**’ p-value<0.01, ‘*’ p-value<0.05, ‘NS’ p-value>0.05.

Fig 3. The effect of residential density on the strength of SI.

Effect heterogeneity between natives exposed to out-movers (X = 1) and natives exposed to no change (X = 0) in 100m x 100m residential areas with different numbers of inhabitants: (1) less or equal to 15; (2) between 16 and 30; (3) more than 30. The estimates are the β coefficient from the LPM on the raw data (in white) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-values in parentheses. The straight horizontal line indicates no effect. The upper bars indicate whether coefficients across models are statistically significantly different from one another. ‘***’ p-value<0.001, ‘**’ p-value<0.01, ‘*’ p-value<0.05, ‘NS’ p-value>0.05.

Fig 4. The marginal effect of distance on the strength of SI.

Difference between natives being exposed only to out-movers leaving from a point close to their own residential location (X = 1, upper) or from farther away (X = 0, lower), respectively for three different spatial levels: (1) 100m x 100m; (2) 200m x 200m; (3) 300m x 300m. The maximum area is bounded by the 400m x 400m square at all levels. The estimates are the β coefficient from the LPM on the raw data (in white) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-values in parentheses. The straight horizontal line indicates no effect. The upper bars indicate whether coefficients across models are statistically significantly different from one another. ‘***’ p-value<0.001, ‘**’ p-value<0.01, ‘*’ p-value<0.05, ‘NS’ p-value>0.05.

Fig 5. The SI effect across residential areas with varying ethnic compositions.

Effect heterogeneity between natives exposed to out-movers (X = 1) and natives exposed to no change (X = 0) in 100m x 100m residential areas across three types of ethnic composition: (1) only natives (left); (2) less than 10% non-natives (middle); and (3) more than 10% non-natives (right). The estimates are the β coefficient from the LPM on the raw data (in white) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-values in parentheses. The straight horizontal line indicates no effect. The upper bars indicate whether coefficients across models are statistically significantly different from one another. ‘***’ p-value<0.001, ‘**’ p-value<0.01, ‘*’ p-value<0.05, ‘NS’ p-value>0.05.

Fig 6. The SI effect and the strength of SI for Göteborg and Malmö.

(Top-left) Difference in the probability of moving-out between natives exposed to out-movers and no in-movers (X = 1) and natives exposed to no change (neither in- nor out-movers) (X = 0). (Bottom-left) Difference between natives who are exposed only to 2, 3, or 4 or more previous out-movers (X = 1, left), or fewer than this (X = 0, right), respectively. (Top-right) Effect heterogeneity between natives exposed to out-movers (X = 1) and natives exposed to no change (X = 0) in 100m x 100m residential areas with different numbers of inhabitants: (1) less or equal to 15; (2) between 16 and 30; (3) more than 30. (Middle-right) Difference between natives being exposed only to out-movers leaving from a point close to their own residential location (X = 1, upper) or from farther away (X = 0, lower), respectively for three different spatial levels: (1) 100m x 100m; (2) 200m x 200m; (3) 300m x 300m. (Bottom-right) Effect heterogeneity between natives exposed to out-movers (X = 1) and natives exposed to no change (X = 0) in 100m x 100m residential areas across three types of ethnic composition: (1) only natives (left); (2) less than 10% non-natives (middle); and (3) more than 10% non-natives (right). The estimates are the β coefficient from the LPM on the raw data (in white, only in Top-left to prevent clutter in the other plots) and after applying CEM (in black). Error bars indicate 95% confidence intervals, p-values in parentheses. The straight horizontal line indicates no effect. The upper bars indicate whether coefficients across models are statistically significantly different from one another. ‘***’ p-value<0.001, ‘**’ p-value<0.01, ‘*’ p-value<0.05, ‘NS’ p-value>0.05.

Fig 7. The macro-consequences of SI on ethnic residential segregation.

Results from simulations using an agent-based model of Schelling’s (1971) spatial proximity model (SPM) with SI partly determining out-mobility (see main text). Average Dissimilarity Index (Taeuber and Taeuber 1976) for 100 executions for each parameter combination, as a function of time according to different strengths of SI: 0 (no SI, black), 0.05 (lime), 0.1 (turquoise), 0.15 (light blue) and 0.2 (blue). The plots show time binned for clarity (maximum run time is 1e5 for all simulations) and confidence intervals at the 95% level (they are tiny). The black line represents Schelling’s SPM with no SI and it is used as reference. The share of the groups is 50%, and households are randomly distributed at the beginning of the simulation. See the S1 Appendix in S2 File for more information about the ABM.