The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability,Wealth,and Race

Abstract

The present research examined the zip code level (177 zip codes) prevalence of and deaths associated with COVID-19 in New York City as of May 22, 2020. Walkable zip codes had consistently lower prevalence of (r = −.49) and deaths (r = −.15) associated with COVID-19. The mediation analysis showed that the degree of reduction in actual geographical mobility during the lockdown (measured by smartphone GIS data) accounted for geographical variations in the number of confirmed cases and deaths. Residents in wealthy zip codes and walkable zip codes were able to limit geographical mobility, whereas residents in poor zip codes and Black and Hispanic dominant zip codes were not. Finally, the spatial lag regression analysis showed that walkability was a robust predictor of zip code–level prevalence of and deaths associated with COVID-19. Overall, walkability seems to have provided protection against the spread of COVID-19.

Keywords

COVID-19 walkability racial disparity social ecology

Natural disasters and pandemics are part of human history and life. Humans have always had to deal with them. Recent research suggests that natural disasters and pandemics have shaped human values, norms, and culture (Gelfand et al., 2011; Oishi et al., 2017; Schaller & Murray, 2008; Van de Vliert, 2013). Recent research also shows that natural disasters and pandemics affect racial minority and lower socioeconomic status individuals disproportionately (e.g., Rhodes et al., 2010; Stephens et al., 2009; Yamada, 2009). The present research sought to identify some protective factors against such vulnerability. Following the socioecological psychological framework (Gelfand et al., 2011; Oishi, 2014), we explored a neighborhood factor that has not yet received much attention in the discussion on COVID-19: walkability, in relation to two other established factors: wealth (Grantz et al., 2016; Lowcock et al., 2012) and race (e.g., Pellowski et al., 2013). In addition, following the geographical psychology framework (Götz et al., 2020; Rentfrow, 2014; Rentfrow & Jokela, 2016), we examined zip code–level geographical variations in the number of confirmed cases and deaths in New York City, the early epicenter of COVID-19 in the United States.

In this work, we focused on walkability (defined as how easy it is to get things done by walking) as well as on race and wealth for two reasons. First, there is numerous research indicating that people living in walkable areas are healthier than those living in less walkable areas (Frank et al., 2007; Sallis et al., 2009). Those living in walkable areas engage in healthy behaviors more often than those living in less walkable areas, suggesting that those living in a walkable zip code might be more willing to engage in healthier behaviors (e.g., wearing a mask) than those living in a less walkable zip code. Second, recent research (Oishi et al., 2019) found that youth from lower income families who grew up in walkable cities (e.g., New York, Pittsburgh) were more likely to move up the economic ladder as adults than those who grew up in less walkable cities (e.g., Miami, Phoenix). The instrumental variable regression analysis and other analyses showed that this association was above and beyond other related variables (e.g., % Black, income inequality, quality of education). Moreover, residents of walkable cities reported a stronger sense of belonging to the city and were more optimistic about their own chance of moving up the economic ladder in the future than those of less walkable cities. Such a sense of belonging and optimism might help residents in walkable areas cope with threats and isolation during a pandemic better than those in less walkable areas.

Although previous research revealed the benefits of walkable neighborhoods, when it comes to a contagious disease, there may also be associated negatives. For instance, walkable neighborhoods are more densely populated (Smith et al., 2008), which tends to make contagion spread more easily and quickly. Therefore, walkable neighborhoods might be more vulnerable to a contagious disease. However, in a walkable neighborhood, residents can limit their exposure to the virus while doing the necessary daily chores during the lockdown (e.g., grocery shopping). In a less walkable neighborhood, residents need to go further to do the chores, thereby exposing themselves to the virus more frequently. Similarly, residents of wealthy neighborhoods could have others deliver their daily essentials to their homes and limit their geographical mobility, whereas those of less wealthy neighborhoods could not. Residents in the majority Black or Hispanic neighborhoods are likely to have a harder time limiting their geographical mobility even under lockdown because they are more likely to be essential workers (https://comptroller.nyc.gov/reports/new-york-citys-frontline-workers).

In sum, we examined geographical variations in the prevalence of COVID-19 cases in terms of walkability as well as race and wealth. Specifically, we hypothesized that (1) lower income neighborhoods would have more COVID-19 cases per capita than higher income neighborhoods, (2) predominantly Black and Hispanic neighborhoods would have more COVID-19 cases per capita than predominantly White neighborhoods, (3) walkable neighborhoods would have fewer COVID-19 cases per capita than less walkable neighborhoods, (4) the disadvantage of low-income neighborhoods and the concentration of racial minority residents would be ameliorated by walkability of the neighborhood (i.e., the association between median income/% Black/% Hispanic and COVID-19 cases would be moderated by walkability), and (5) the link between the walkability, wealth, and racial composition of neighborhoods and COVID-19 spread is explained in part by a reduction in geographical mobility during the exposition of COVID-19 in New York City.

Method

The sample size was determined by the total number of zip codes in New York City. We used all the zip codes with available data. The total number of zip codes used in the following analyses was 177, which exceeds the suggested sample size, 134 (calculated using G*Power 3.1.9.2, Faul et al., 2009, assuming f ² = .10, power = .80, number of predictors = 5, which are used for the multiple regression analysis in the Result section).

Dependent Variables

Because New York City has been the epicenter of the COVID-19 pandemic in the United States, we examined geographical variations in COVID-19 in New York City. The data were available from New York City’s Department of Health at https://www1.nyc.gov/site/doh/COVID/COVID-19-data.page. We analyzed the cumulative data as of May 22, 2020, as the number of cases dropped substantially afterward; the first case in New York City was confirmed on March 1, and the peak was on April 6, the day there were 6,374 new cases; 1 month later on May 6, there were 1,401 new cases; on May 20, there were 1,069, on May 21 there were 1,069, and on May 22 1,055; then on May 23, there were 463 new cases, followed by 475 on May 24, and 463 on May 25. The results are largely the same when the cumulative data on May 18 are used (see Online Supplementary Materials). As of May 22, 2020, the mean number of confirmed COVID-19 cases across 177 zip codes was 2,165.68 (SD = 842.74). The mean number of deaths associated with COVID-19 was 178.24 (SD = 96.37).

Predictors

The walkability score for each zip code was taken from www.walkscore.com. A walk score is calculated for each street address based on access to various amenities (e.g., grocery store, library, school) and physical factors such as population density, block length, and intersection density. Walk scores range from 0 (nothing to walk to) to 100 (most daily needs can be met by walking). Walkscore.com provides the mean walk score at the zip code level (see Carr et al., 2011; Duncan et al., 2011, for validity). The mean walk score across 177 zip codes was 86.40 (SD = 12.81).

We obtained median income, % Asian, % Black, and % Hispanic for all the zip codes in New York City from the 2014–2018 American Community Survey (ACS) taken by the U.S. Census Bureau (see census.gov/content/dam/Census/programs-surveys/acs/about/ACS_Information_Guide.pdf for more information). The ACS is an annual nationwide survey conducted by the U.S. Census Bureau, including over 3.5 million households, on social, economic, housing, and demographic characteristics about the U.S. population. To get the stable estimates, we took the mean of the five most recent surveys (2014–2018). The mean median household income across 178 zip codes in New York City was $73,614 (SD = 36,439.44). The mean % Asian residents was 15% (SD = 14). The mean % Black residents was 20% (SD = 24). The mean % Hispanic residents was 26% (SD = 20).

Control Variables

Life expectancy (2010–2015) was taken from United States Department of Health and Human Services (2018). This is the most recent life expectancy data at the zip code level. Deaths occurring in 2010–2015 were geocoded based on their residential addresses, then matched with the 2010 decennial census and the 2011–2015 ACS data (see www.cdc.gov/nchs/data/series/sr_02/sr02_181.pdf for details). The mean life expectancy across 176 zip codes was 80.71 (SD = 2.70).

% Asian, % Black, and % Hispanic, occupants per room, and median age were taken from the 2014–2018 ACS. The mean median age across 178 zip codes was 37.86 (SD = 5.68). The occupants per room reported in the ACS is obtained by dividing the reported number of current residents in each residential unit by the number of rooms in the unit. The mean number of occupants per room across 178 zip codes was 8.26 (SD = 4.90). The political orientation of zip codes was taken from the 2017 Cook Partisan Voting Index (PVI; retrieved from https://cookpolitical.com/index.php/introducing-2017-cook-political-report-partisan-voter-index). The PVI is computed for each congressional district by taking the difference in voting %, the higher number indicating more Republican votes. Zip code–level data were not available. Thus, we matched the PVI of congressional districts with zip codes and used the congressional district PVI score as the proxy to the zip code–level PVI. The mean PVI across 178 zip codes was −28.21 (SD = 13.73).

The Mediator Variable

Reduction in geographical mobility data were obtained from SafeGraph (safegraph.com), which generates the visitation data using a panel of GPS pings from millions of anonymous mobile devices. SafeGraph holds 108,851 points of interest (POI) in New York City and 28 million unique visitors were geocoded as visiting the POI from mid-February to mid-April, 2020 (see Benzell et al., 2020, for the use of similar GIS data on dangerous locations for COVID-19). We calculated reduction in geographical mobility using two indices: the number of visits to various locations in a given zip code and the number of visitors to a given zip code area in two time periods: 4 weeks before the new cases started to explode (February 17–March 15) and 4 weeks afterward (March 16–April 12). On March 16, President Trump issued guidelines for social distancing and gatherings of more than 10 people. The official lockdown order in New York went in effect on March 22. Thus, February 17–March 15 was a period of relatively free mobility (baseline geographical mobility), whereas 4 weeks afterward was a period of limited mobility, followed soon by the official lockdown (lockdown geographical mobility). We thus took the difference between the average daily mobility during the 4 weeks to March 15 and the average daily mobility during the 4 weeks afterward (March 16–April 12), then divided the difference by the average daily mobility during the 4 weeks before March 15. Thus, these two indices measure the % reduction in geographical mobility. Because these two variables were highly correlated (r = .987, p < .001), we created the composite score by taking the mean of the two. The higher the composite score, the greater the degree of reduction in geographical mobility (i.e., more staying at home than before the lockdown). The mean reduction in geographical mobility across 171 zip codes was 59.25% (SD = 11.43).

Results

All data can be found at https://osf.io/nj3he/?view_only=67465455934f40ac831a53bcfe6ce9bb and the specific codes can be found in the Online Supplementary Materials. Figure 1A shows geographical variations in the number of confirmed COVID-19 cases, whereas Figure 1B shows geographical variations in the number of confirmed COVID-19-related deaths in New York City.

Figure 1.

(A) Geographical variations in the number of confirmed COVID-19 cases per 100,000 residents in New York City by Zip Code. (B) Geographical variations in the number of confirmed COVID-19-related deaths per 100,000 Residents in New York City by Zip Code.

As predicted by Hypothesis 1, COVID-19 cases and deaths per 100,000 residents were substantially higher in lower median income zip codes in New York City than higher median income zip codes (Table 1). Consistent with Hypothesis 2, the greater the proportion of Black and Hispanic residents there were in a given zip code, the higher the confirmed COVID-19 cases and COVID-19-related deaths were. Consistent with Hypothesis 3, there were fewer COVID-19 cases and deaths in more walkable zip codes than less walkable zip codes in New York City.

Table 1.

Correlations and Descriptive Statistics at the Level of Zip Code.

	1	2	3	4	5	6	7	M (SD)
1. Confirmed Cases	—	.69**	−.49**	−.54**	−.19*	.35**	.43**	2,165.68 (842.74)
2. Deaths		—	−.15*	−.54**	−.11	.34**	.36**	178.24 (96.37)
3. Walkability			—	.09	.04	−.15*	.16*	86.40 (12.81)
4. Median income				—	.14	−.36**	−.57**	73,614 (36,439.44)
5. % Asian					—	−.47**	−.25**	.15 (.14)
6. % Black						—	−.01	.20 (.24)
7. % Hispanic							—	.26 (.20)

Note. Confirmed cases and deaths are per 100,000 residents per zip code as of May 22, 2020. Walkability score ranges from 0 to 100. When the log-linear transformed median income is used instead, correlations are quite similar to the ones reported above: confirmed case = −.50**, deaths = −.57**, walkability = −.01, % Asian = .20**, % Black = −.38**, % Hispanic = −.65**.

Next, we tested our Hypotheses 1–3 simultaneously using the multiple regression analysis. Specifically, we regressed the number of confirmed cases per 100,000 residents on walkability, median income (log-transformed), % Asian, % Black, and % Hispanic. As seen in Table 2, Hypotheses 2 (% Black/% Hispanic) and 3 (walkability) were supported, whereas Hypothesis 1 (median income) was not. In terms of the effect size, 1 SD difference in walkability was associated with 452 fewer cases per 100,000 residents. One SD difference in % Hispanic corresponded to 412 more cases. One SD difference in % Black corresponded to 243 more cases.

Table 2.

Predicting the Number of Confirmed Cases per 100,000 Residents (Standardized Regression Coefficients).

Predictors	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Walkability	−.54***	−.49***	−.56***	−.48***	−.59***	−.45***
Median Inc	−.10	−.11	−.17	−.11	−.04	−.04
% Asian	.12	.10	.09	.10	.06	.05
% Black	.29***	.32**	.32***	.31***	.27***	.36***
% Hispanic	.49***	.50***	.50***	.52***	.40***	.44***
Conservatism		.10				.15*
Life expectancy			.14			.04
Median age				.12		.23**
Occupants per room					.21**	.34***
df	171	170	169	170	170	166
R ²	.604	.611	.607	.612	.626	.663

Note. Median Inc = median income (log transformed).

*p < .05. **p < .01. ***p < .001.

Next, we ran another regression analysis; this time predicting the number of COVID-19-related deaths from the same set of predictors (see Table 3). This regression analysis showed that Hypotheses 1 (median income) and 3 (walkability) were fully supported, whereas Hypothesis 2 was partially supported. One SD difference in median income was associated with 41 fewer deaths. One SD difference in walkability was associated with 14 fewer deaths. The greater the proportion of Black residents, the more deaths (1 SD difference corresponded to 21 more deaths). In contrast, % Hispanic was no longer significant.

Are the Wealth and Race Effects Moderated by Walkability?

We tested Hypothesis 4 using the multiple regression models with the interaction terms between median income/% Black/% Hispanic and walkability. Hypothesis 4 was not supported in terms of the number of confirmed cases as the inverse association between walkability and the number of confirmed cases was not moderated by median income (p = .233), % Black (p = .476), nor % Hispanic (p = .172). That is, regardless of median income and % racial minorities, walkability was consistently associated with fewer COVID-19 cases (see Figure 2).

Figure 2.

The inverse association between walkability and the number of confirmed COVID-19 cases in New York City is not moderated by % Black. Regardless of % Black, walkability is inversely associated with the number of confirmed cases.

In contrast, Hypothesis 4 was supported in terms of median income and COVID-19-related deaths. Specifically, the inverse association between median income and the number of confirmed COVID-19-related deaths was moderated by walkability, β = .367, t (173) = 4.882, p < .001. As seen in Figure 3, in walkable zip code areas (+1 SD), median income was less strongly associated with the number of deaths (b = −32.25, SE = 7.24, 95% CI [−46.54, −17.96], t = −4.45, p < .001) than in less walkable zip code areas (−1 SD): b = −109.02, SE = 12.24, 95% CI [−133.38, -−84.66], t = −8.84, p < .001). That is, among not so walkable neighborhoods, the wealth of the neighborhood mattered a lot in predicting the number of COVID-19-related deaths, whereas among walkable neighborhoods, the wealth of the neighborhood did not matter as much. However, Hypothesis 4 was not supported in terms of % Black and % Hispanic, as the inverse walkability–death association was not moderated by % Black (p = .625) nor % Hispanic (p = .572).¹ Walkable neighborhoods had fewer deaths than less walkable neighborhoods, regardless of % Black or Hispanic.

Figure 3.

The inverse association between median income (standardized log-linear transformed median income) and COVID-19 deaths per 100,000 residents is moderated by walkability: The regression slope is −69.65 (SE = 15.63), t = −4.45, p < .001 in highly walkability zip codes (+1 SD); the regression slope is −235.48 (SE = 26.65), t = −8.84, p < .001 in not so walkable zip codes (−1 SD).

Possible Alternative Explanations

Before moving on to test Hypothesis 5, we explored four alternative explanations for the findings associated with walkability, as the type of correlational analyses above could be still distorted by third variables. First, the previous research showed that people living in walkable neighborhoods are healthier than those living in less walkable neighborhoods (Frank et al., 2007; Sallis et al., 2009). Then, the inverse correlation between walkability and COVID-19 cases and deaths could be driven by the preexisting, prepandemic differences in general health conditions. Indeed, the healthier, longer prepandemic life expectancy zip codes had fewer COVID-19 cases, r (174) = −.388, p < .001, and deaths, r (174) = −.382, p < .001, than the less healthy, shorter life expectancy zip codes. The residents of walkable neighborhoods had a longer life expectancy than those of less walkable neighborhoods, r (174) = .166, p = .027. The prepandemic life expectancy was also positively associated with median income, r (174) = .655, p < .001, and % Asian, r (174) = .447, p < .001, and negatively associated with % Black, r (174) = −.515, p < .001, and % Hispanic, r (174) = −.408, p < .001.

However, controlling for the prepandemic life expectancy, walkable zip codes still had fewer COVID-19 cases (see Table 2). Similarly, % Black and % Hispanic were still positively associated with the number of cases. Likewise, walkability was still associated with fewer deaths (see Table 3). The walkability–median income interaction effect, the main effect of median income, and % Black remained significant, while % Hispanic was no longer significant.

Table 3.

Predicting the Number of COVID-19-Related Deaths per 100,000 Residents (Standardized Regression Coefficients).

Predictors	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Walkability	−.33***	−.33***	−.34***	−.15*	−.38***	−.11
Median Inc	−.59***	−.59***	−.60***	−.61***	−.56***	−.43***
% Asian	.13	.13	.12	.07	.10	−.03
% Black	.18*	.18*	.19*	.23**	.17*	.22**
% Hispanic	.17	.17	.18	.27**	.12	.18*
Conservatism		.01				.04
Life expectancy			.05			−.17*
Median age				.38***		.56***
Occupants per room					.10	.36***
Walkability × Med Inc	.36***	.36***	.37***	.37***	.36***	.38***
df	170	169	168	169	169	165
R ²	.449	.449	.444	.527	.454	.571

Note. Med Inc = median income (log transformed).

*p < .05. **p < .01. ***p < .001.

Second, walkable neighborhoods in New York City might have more young and healthy people than less walkable neighborhoods, and that could explain the walkability effects found above. Indeed, the median age was substantially lower in walkable zip codes than less walkable zip codes, r (175) = −.496, p < .001. The regression analysis showed that additionally controlling for the median age did not dramatically change the regression estimates for walkability (see Table 2) on the number of confirmed cases. Again, controlling for median age additionally, % Black and % Hispanic were also associated with more confirmed cases. In terms of the number of confirmed deaths (see Table 3), the zip codes with older residents (higher median age) had more deaths. However, when controlling for median age additionally, the regression coefficients for walkability and the walkability–median income interaction effect on the confirmed deaths remained significant. Median income, % Black, and % Hispanic also remained significant.

Third, walkable neighborhoods in New York City might have more single households than less walkable neighborhoods. Given that crowded housing is a major risk factor for a pandemic (Aligne, 2016), walkability’s positive effects could be explained by less crowded (more single) occupancy. In reality, however, occupants per room were larger in walkable neighborhoods than less walkable neighborhoods, r (175) = .312, p < .001. The regression analysis showed that occupants per room were associated with more confirmed cases (see Table 2). Additionally controlling for the occupants per room did not change the regression estimates for walkability on the number of confirmed cases as well as for % Black and % Hispanic.

Occupants per room were not associated with the number of deaths above and beyond other predictors. Additionally controlling for the occupants per room, both the main and the interactive effects of walkability remained significant on the number of deaths (see Table 3). % Black remained a significant predictor of the number of deaths, whereas % Hispanics no longer predicted the number of deaths.

Fourth, walkable neighborhoods are preferred by politically liberal individuals than conservative individuals (Lewis, 2015). Indeed, in New York City, walkable zip codes had more liberal residents than less walkable zip codes, as measured by Cook’s partisan voting index, r (175) = −.420, p < .001. Cook’s partisan voting index (the higher number means more conservative) was not associated with the number of confirmed cases, r (175) = .06, p = .451, nor deaths, r (175) = −.14, p = .065. Controlling for Cook’s partisan voting index, walkability, % Black, and % Hispanic remained significant (see Table 2). In terms of the number of deaths, the main effect of walkability and the interactive effect of walkability also remained significant (see Table 3). % Black also remained a significant predictor of the number of deaths, whereas % Hispanic no longer predicted the number of deaths.

Finally, controlling for the prepandemic life expectancy, median age, occupants per room, and political orientation simultaneously, walkable zip codes still had fewer cases than less walkable zip codes. The greater the proportion of Black and Hispanic residents there were, the more confirmed cases there were (see Table 2). In terms of the number of deaths, the walkability main effect disappeared, but the interaction effect remained significant. Wealthier zip codes and zip codes with longer prepandemic life expectancy had fewer deaths (see Table 2). The greater the proportion of Black and Hispanic residents there were, the more COVID-19-related deaths there were. The older the median age, the more occupants per room, the more deaths as well.

Spatial Autocorrelation Issues

The zip code–level data within a given city pose a spatial autocorrelation issue (e.g., neighboring zip codes are more similar to one another than distant zip codes, see Legendre, 1993, for conceptual details; see Ebert et al., 2020, for an example in personality psychology). For instance, the number of confirmed cases was highly correlated with the number of confirmed cases in neighboring zip codes, r (174) = .853, p < .001. Similarly, the number of deaths was correlated with the number of deaths in neighboring zip codes, r (174) = .553, p < .001. To address the spatial autocorrelation issue, we reran the key analyses reported in Table 2 and 3, using the spatial lag regression analysis. Specifically, we included the mean number of cases/deaths in neighboring zip codes as an additional predictor in the regression analysis (Model 6) reported in Table 2 and 3. Additionally controlling for the spatial autocorrelation (i.e., neighboring zip code’s confirmed cases), walkability, β = −.146, t (164) = −2.299, p = .023, % Black (β = .173, t = 2.921, p = .004), % Hispanic (β = .178, t = 2.552, p = .012), and occupants per room (β = .187, t = 2.911 p = .004) remained significant predictors of COVID-19 cases.

Likewise, controlling for the mean number of deaths per 100,000 residents in neighboring zip codes additionally to Model 6 in Table 3, the deaths per 100,000 residents were still predicted by median income, β = −.431, t (163) = −3.835, p < .001, % Black (β = .202, t = 2.619, p = .010, median age (β = .496, t = 5.581, p < .001), and the walkability–median income interaction effect (β = .368, t = 5.438, p < .001). Thus, the spatial autocorrelation did not substantially change our regression estimates reported above.

How to Explain Walkability, Median Income, Race Effects: Reduction in Geographical Mobility

Now that the roles of walkability, wealth, and racial compositions on COVID-19 cases and deaths are clarified, we went on to test Hypothesis 5. To test this exposure hypothesis, we used actual mobility data collected through smartphones using the GIS. The GIS geographical mobility data indicate that after the lockdown order, residents in wealthier (r = .653, p < .001) and walkable zip code areas (r = .262, p = .001) were able to limit their movement and exposure to the virus to a greater degree than those in less wealthy, less walkable zip code areas. In contrast, residents in more Black (r = −.500, p < .001) and Hispanic neighborhoods (r = −.370, p < .001) were less able to reduce geographical mobility.

Next, we went on to test the reduction in geographical mobility as a mediator model, using Hayes (2018) Process Procedure for SPSS version 3.3. The number of bootstrap was set at 10,000. As seen in Figure 4, there was a significant mediation effect: standardized indirect effect = −.118 (SE = .035), 95% CI [−.190, −.054]. Walkability was associated with the reduction in geographical mobility after the lockdown order, which in turn was associated with fewer confirmed cases. Likewise, there was a significant mediation effect on the number of COVID-19-related deaths: standardized indirect effect = −.139 (SE = .042), 95% CI [−.226, −.062]. Walkability was associated with a greater reduction in geographical mobility, which in turn was associated with fewer deaths (β = −.531, t = -−7.79, p < .001). Once the reduction in geographical mobility was taken into account, walkability was no longer associated with the number of deaths (β = .014, t = .211, p = .833), though it continued to be associated with the number of confirmed cases.

Figure 4.

The walkability–COVID-19 cases association is partially mediated by reduction in geographical mobility after the lockdown order.

The next mediation analysis revealed that the wealthier the zip code areas were, the greater the reduction in geographical mobility (β = .662, t = 11.45, p < .001), which in turn was associated with fewer cases (β = −.437, t = −5.10, p < .001): standardized indirect effect = −.290 (SE = .058), 95% CI = [−.405, −.180], and deaths (β = −.260, t = −3.17, p = .002): standardized indirect effect = −.172 (SE = .061), 95% CI [−.290, −.052]. Once the reduction in geographical mobility was taken into account, median income was no longer associated with the number of confirmed cases (β = −.162, t = −1.89, p = .061), though median income remained a significant predictor of deaths (β = −.403, t = −4.92, p < .001).

As seen in Figure 5, the next mediation analysis showed that the greater the proportion of Black residents, the lower the reduction in geographical mobility after the lockdown. To the extent that the reduction in geographical mobility is associated with fewer cases and deaths, the smaller reduction in geographical mobility in high proportion Black zip codes partially explained why there were more confirmed cases (standardized indirect effect = .270, SE = .046, 95% CI [.184, .362]) and deaths (standardized indirect effect = .253, SE = .044, 95% CI [.171, .341]) in high proportion Black zip codes (see Figure 4). Notably, once reduction in geographical mobility is taken into account, % Black no longer predicted the number of confirmed cases (β = .051, t = .676, p = .500), nor deaths.

Figure 5.

The % Black-COVID-19 deaths association is mediated by reduction in geographical mobility after the lockdown order.

Similarly, the final mediation analysis found that the greater the proportion of Hispanic residents, the lower the reduction in geographical mobility (β = −.395, t = −5.57, p < .001), and the lower the reduction, the more cases (β = −.456, t = −6.66, p < .001; standard indirect effect = .180, SE = .038, 95% CI [.110, .259]) and more deaths (β = −.468, t = −6.62, p < .001; standardized indirect effect = .185, SE = .039, 95% CI [.112, .265]). The main effect of % Hispanic remained significant: case (β = .223, t = 3.26, p = .001) and death (β = .150, t = 2.12, p = .035).

Discussion

This study explored geographical variations in the number of confirmed COVID-19 cases and deaths in New York City. There were more confirmed cases of COVID-19 in less walkable, poorer, more Black, and more Hispanic neighborhoods. We found that walkable neighborhoods had fewer COVID-19 cases, even when accounting for racial, demographic, and economic variations as well as preexisting health conditions across zip code areas. Thus, walkability appears to help contain the quick spread of the virus. The GIS-based geographical mobility data showed that being able to accomplish daily chores easily is likely to protect people from COVID-19 by limiting the outing. Reduced geographical mobility also partially explained the link between median income, % Black, and % Hispanic and COVID-19 cases and deaths.

Before closing, several limitations should be noted. First, because New York City is far more walkable than most U.S. cities, suburbs, and rural areas, it is questionable whether walkability has a similar effect elsewhere. It is important to replicate the current findings in other cities. Second, the present data set did not allow us to explore psychological mechanisms regarding why walkability seems to help contain the spread of COVID-19. One possibility is that people living in a walkable neighborhood feel a strong sense of belonging (Oishi et al., 2019) and will feel that they are in this pandemic together with their neighbors. They may also get more help from neighbors than those living in less walkable neighborhoods (Leyden, 2003). Another possibility is that walkability of the neighborhood is systematically associated with the personality of residents. For instance, the residents of walkable neighborhoods are more conscientious than those of less walkable neighborhoods (Götz, Yoshino, & Oshio, 2020). Conscientiousness is associated with more precautions toward COVID-19 (Aschwanden et al., 2020) and a smaller likelihood of infection (Qian & Yahara, 2020). Conscientiousness is also associated with healthy behaviors in general (Bogg & Roberts, 2004). As such, personality could explain in part why walkable neighborhoods had lower rates of and deaths related to COVID-19 in New York City. The precise mechanisms underlying the link between walkability and the spread of COVID-19 need to be explored in the future.

As in previous pandemics, poor and minority–majority neighborhoods were more susceptible to the spread of COVID-19 in New York City. Most important, the current research found that the walkability of the neighborhood provides some protection against the spread of COVID-19. As we prepare for the second wave of COVID-19, and new viruses in the future, it is worth considering how to make cities more walkable and how to help racial minorities and people with limited means avoid going out during the lockdown.

Supplemental Material

Supplemental Material, Supplementary-data - The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race

Supplemental Material, Supplementary-data for The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race by Shigehiro Oishi, Youngjae Cha and Ulrich Schimmack in Social Psychological and Personality Science

Supplemental Material

Supplemental Material, Supplementary-supplementary_results - The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race

Supplemental Material, Supplementary-supplementary_results for The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race by Shigehiro Oishi, Youngjae Cha and Ulrich Schimmack in Social Psychological and Personality Science

Supplemental Material

Supplemental Material, Supplementary-syntax - The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race

Supplemental Material, Supplementary-syntax for The Social Ecology of COVID-19 Cases and Deaths in New York City: The Role of Walkability, Wealth, and Race by Shigehiro Oishi, Youngjae Cha and Ulrich Schimmack in Social Psychological and Personality Science

Footnotes

Acknowledgments

We would like to thank Bansi Javiya for preparing some of the zip code level data used in our paper.

Author Contributions

Shigehiro Oishi designed the research, performed research, analyzed data, and wrote the first draft of the paper. Youngjae Cha contributed the data, analyzed data, and revised the paper. Ulrich Schimmack provided the data, analyzed data, and revised the paper.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Youngjae Cha

Ulrich Schimmack

Supplemental Material

The supplemental material is available in the online version of the article.

Note

References

Aligne

C. A.

(2016). Overcrowding and mortality during the influenza pandemic of 1918: Evidence from U.S. army camp A. A. Humphreys, Virginia. American Journal of Public Health, 106, 642–644.

Aschwanden

Strickhouser

J. E.

Sesker

A. A.

Lee

J. H.

Luchetti

Stephan

Sutin

A. R.

Terracciano

(2020). Psychological and behavioural responses to Coronavirus Disease 2019: The role of personality. European Journal of Personality.

Benzell

S. G.

Collis

Cicolaides

(2020). Rationing social contact during COVID-19 pandemic: Transmission risk and social benefits of US locations. The Proceedings of the National Academy of Science, 117, (26), 14642–14644.

Bogg

Roberts

B. W.

(2004). Conscientiousness and health-related behaviors: A meta-analysis of the leading behavioral contributors to mortality. Psychological Bulletin, 130, 887–919.

Carr

L. J.

Dunsiger

S. I.

Marcus

B. H.

(2011). Validation of walk score for estimating access to walkable amenities. British Journal of Sports Medicine, 45, 1144–1148.

The Cook Political Report (2017). Introducing the 2017 Cook Political Report Partisan Voter Index. Retrieved October 12, 2020 from https://cookpolitical.com/index.php/introducing-2017-cook-political-report-partisan-voter-index

Duncan

D. T.

Aldstadt

Whalen

Melly

S. J.

Gortmaker

S. L.

(2011). Validation of walk score for estimating neighborhood walkability: An analysis of four U.S. metropolitan areas. International Journal of Environmental Research and Public Health, 8, 4160–4179.

Ebert

Götz

F. M.

Gladstone

J. J.

Müller

S. R.

Matz

S. C.

(2020). Spending reflects not only who we are but also who we are around: The joint effects of individual and geographic personality on consumption. Journal of Personality and Social Psychology.

Faul

Erdfelder

Buchner

Lang

A.-G.

(2009). Statistical power analysis using G*Power 3.1: The correlation and regression analyses. Behavioral Research Methods, 41, 1149–1160.

10.

Frank

L. D.

Sallis

J. F.

Conway

T. L.

Chapman

J. E.

Saelens

B. E.

Bachman

(2007). Many pathways from land use to health: Associations between neighborhood walkability and active transportation, body mass index, and air quality. Journal of the American Planning Association, 72, 75–87.

11.

Gelfand

M. J.

Raver

J. L.

Nishii

Leslie

L. M.

Lun

Lim

B. C.

Duan

Almaliach

Ang

Arnadottir

Aycan

Boehnke

Boski

Cabecinhas

Chan

Chhokar

D’Amato

Ferrer

Fischlmayr

I. C.

… Aycan

(2011). Differences between tight and loose cultures: A 33-nation study. Science, 332(6033), 1100–1104.

12.

Götz

F. M.

Stieger

Gosling

S. D.

Potter

Rentfrow

P. J.

(2020). Physical topography is associated with human personality. Nature Human Behaviour, 4(11), 1135–1144.

13.

Götz

F. M.

Yoshino

Oshio

(2020). The association between walkability and personality: Evidence from a large socioecological study in Japan. Journal of Environmental Psychology, 69. https://doi.org/10.1016/j.jenvp.2020.101438

14.

Grantz

K. H.

Rane

M. S.

Salje

Glass

G. E.

Schachterle

S. E.

Cummings

D. A. T.

(2016). Disparities in influenza mortality and transmission related to sociodemographic factors within Chicago in the pandemic of 1918. The Proceedings of the National Academy of Science, 113, 13839–13844.

15.

Hayes

A. F.

(2018). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (2nd ed.). The Guilford Press.

16.

Legendre

(1993). Spatial autocorrelation: Trouble or new paradigm. Ecology, 74, 1659–1673.

17.

Lewis

P. G.

(2015). Moral intuitions and smart growth: Why do liberals and conservatives view compact development so differently. Journal of Urban Affairs, 37, 87–108.

18.

Leyden

K. M.

(2003). Social capital and the built environment: The importance of walkable neighborhoods. American Journal of Public Health, 93, 1546–1551.

19.

Lowcock

E. C.

Rosella

L. C.

Foisy

McGeer

Crowcroft

(2012). The social determinants of health and pandemic H1N1 2009 Influenza severity. American Journal of Public Health, 102, e51–58.

20.

Oishi

(2014). Socioecological psychology. Annual Review of Psychology, 65, 581–609.

21.

Oishi

Koo

Buttrick

(2019). The social ecology of upward social mobility. American Psychologist, 74, 751–763.

22.

Oishi

Yagi

Komiya

Kohlbacher

Kusumi

Ishii

(2017). Does a major earthquake change job preferences and human values? European Journal of Personality, 31, 258–265.

23.

Pellowski

J. A.

Kalichman

S. C.

Matthews

K. A.

Adler

(2013). A pandemic of the poor: Social disadvantage and the U.S. HIV Epidemic. American Psychologist, 68(4), 197–209.

24.

Qian

Yahara

(2020). Mentality and behavior in COVID-19 emergency status in Japan: Influence of personality, morality, and ideology. PLoS One, 15 (7): e0235883.

25.

Rentfrow

P. J.

(2014). Geographical psychology: Exploring the interaction of environment and behavior. American Psychological Association.

26.

Rentfrow

P. J.

Jokela

(2016). Geographical psychology: The spatial organization of psychological phenomena. Current Directions in Psychological Science, 25, 393–398.

27.

Rhodes

Chan

Paxson

Rouse

C. E.

Water

Fussell

(2010). The impact of Hurricane Katrina on the mental and physical health of low-income parents in New Orleans. American Journal of Orthopsychiatry, 80, 237–247.

28.

SafeGraph (2020). Guide to points-of-interest data: POI data FAQ. Retrieved October 12, 2020 from https://www.safegraph.com/points-of-interest-poi-data-guide

29.

Sallis

J. F.

Saelens

B. E.

Frank

L. D.

Conway

T. L.

Slymen

D. J.

Cain

K. L.

Chapman

J. E.

Kerr

(2009). Neighborhood built environment and income: Examining multiple health outcomes. Social Science & Medicine, 68, 1285–1293.

30.

Schaller

Murray

D. R.

(2008). Pathogens, personality, and culture: Disease prevalence predicts worldwide variability in sociosexuality, extraversion, and openness to experience. Journal of Personality and Social Psychology, 95, 212–221.

31.

Smith

K. R.

Brown

B. B.

Yamada

Kowaleski-Jones

Zick

C. D.

Fan

J. X.

(2008). Walkability and body mass index: Density, design, and new diversity measures. American Journal of Preventive Medicine, 35, 237–244.

32.

Stephens

N. M.

Hamedani

M. G.

Markus

H. R.

Bergsieker

H. B.

Eloul

(2009). Why did they “choose” to stay? Perspectives of Hurricane Katrina observers and survivors. Psychological Science, 20, 878–886.

33.

U.S. Census Bureau (2020). American Community Survey: Information Guide. Retrieved October 12, 2020 from https://www.census.gov/content/dam/Census/programs-surveys/acs/about/ACS_Information_Guide.pdf

34.

U.S. Department of Health and Human Services (2018). U.S. small-area life expectancy estimates project: Methodology and results summary. DHHS Publication No. 2011–2018.

35.

Van de Vliert

(2013). Climato-economic habitats support patterns of human needs, stresses, and freedoms. Behavioral and Brain Sciences, 36, 465–480.

36.

Yamada

(2009). Poverty, wealth, and access to pandemic influenza vaccines. The New England Journal of Medicine, 361, 1129–1131.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.05 MB

0.07 MB

0.01 MB