7 Social Infrastructure & Well-Being

7.1 Libraries Used

library(tidyverse)
#> ── Attaching core tidyverse packages ──── tidyverse 2.0.0 ──
#> ✔ dplyr     1.1.4     ✔ readr     2.1.5
#> ✔ forcats   1.0.1     ✔ stringr   1.6.0
#> ✔ ggplot2   4.0.0     ✔ tibble    3.3.0
#> ✔ lubridate 1.9.4     ✔ tidyr     1.3.1
#> ✔ purrr     1.2.0     
#> ── Conflicts ────────────────────── tidyverse_conflicts() ──
#> ✖ dplyr::filter() masks stats::filter()
#> ✖ dplyr::lag()    masks stats::lag()
#> ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(nycOpenData)
library(dplyr)
library(stringr)

7.2 Data Loading

First, I loaded records of NYC permitted events and NYC borough community reports using the NYC Open Data package that my professor (Christian Martinez) created.

Events <- nyc_permit_events_historic(limit = 10000, filters = list())
BoroReport <- nyc_borough_community_report(limit = 10000, filters = list())

knitr::kable(
  head(Events, 25),
  caption = "First 25 rows of Events"
)

knitr::kable(
  head(BoroReport, 25),
  caption = "First 25 rows of BoroReport"
)

7.3 Cleaning

eventscleaner <- Events %>%
  mutate(
    cd_id =
      community_board |> 
      str_replace_all("[^0-9]", "") |>  
      as.numeric()                       
  )

BoroReport <- BoroReport %>%
  
  mutate(
    snap_borough = str_extract(community_district, "^[A-Za-z]") |> str_to_upper(),
    snap_cb      = str_extract(community_district, "[0-9]+") |> as.numeric()
  ) %>%
  
  mutate(
    snap_borough_num = case_when(
      snap_borough == "M" ~ 100,  # Manhattan
      snap_borough == "B" ~ 200,  # Bronx
      snap_borough == "K" ~ 300,  # Brooklyn
      snap_borough == "Q" ~ 400,  # Queens
      snap_borough == "S" ~ 500,  # Staten Island
      TRUE ~ NA_real_
    ),
    
    cd_id = snap_borough_num + snap_cb
  )

eventscleaner <- eventscleaner %>%
  
  mutate(
    borough_num = case_when(
      event_borough == "Manhattan"     ~ 100,
      event_borough == "Bronx"         ~ 200,
      event_borough == "Brooklyn"      ~ 300,
      event_borough == "Queens"        ~ 400,
      event_borough == "Staten_Island" ~ 500,
      event_borough == "Staten Island" ~ 500,  
      TRUE ~ NA_real_
    ),
    
   
    cd_id = borough_num + cd_id
  )

7.4 Events Count

After this, I glanced at the number of events per community district – just to garner a better understanding of the data.

events_cd <- eventscleaner %>%
  count(cd_id, name = "n_events")

knitr::kable(
  head(events_cd, 30),
  caption = "Number of Events Per CD"
)

Across community districts, the mean number of permitted events was 312.5, with a median of 63. (Note: The right skew in the data was due to the number of events in joint-interest areas. These were dropped from the later analysis, due to the lack of SNAP recipients in those areas).

I then created a graph to display the number of events per district, in descending order:

• Community district numbers correspond to the final two digits shown on the y-axis.
• District numbers starting with 1 indicate Manhattan.
• District numbers starting with 2 indicate the Bronx.
• District numbers starting with 3 indicate Brooklyn.
• District numbers starting with 4 indicate Queens.
• District numbers starting with 5 indicate Staten Island.
• New York City has 59 community districts in total:
  • Manhattan: 12 districts
    
  • The Bronx: 12 districts
    
  • Brooklyn: 18 districts
    
  • Queens: 14 districts
    
  • Staten Island: 3 districts
• District numbers that do not follow this schema (for example, 55 and 64) refer to joint-interest areas rather than standard community districts.
  • District 55 corresponds to Prospect Park.
  • District 64 corresponds to Central Park.

A full list of community districts and joint-interest areas is available here.

events_cd %>%
  slice_max(n_events, n = 25) %>%
  ggplot(aes(x = reorder(cd_id, n_events), y = n_events)) +
  geom_col(fill = "steelblue") +
  coord_flip() +
  labs(
    title = "Number of Events by Community District (Top 25)",
    x = "Community District",
    y = "Number of Events"
  ) +
  theme_minimal()

Figure 7.1: Fig 1: Number of permitted events by community district, ordered from highest to lowest for the top 25. Each horizontal bar represents a distinct community district. Community district identifiers are displayed on the y-axis and event counts are displayed on the x-axis.

7.5 SNAP Benefits Count

I also looked over the number of SNAP recipients per district.

The table below shows the mean number of SNAP recipients per month per community district. (Note: There are not necessarily equal amounts of people per community district, so number of SNAP recipients within a given district is not a de facto indication of the proportional amount of poverty in the area. That said, it still functions as a meaningful snapshot of poverty rates).

BoroReport <- BoroReport %>%
  mutate(
    bc_snap_recipients = as.numeric(bc_snap_recipients)
  )

snap_plot_data <- BoroReport %>%
  group_by(cd_id) %>%
  summarise(
    bc_snap_recipients = mean(bc_snap_recipients,na.rm = TRUE),
    .groups = "drop"
  )

snap_plot_data %>%
  slice_max(bc_snap_recipients, n = 25) %>%
  ggplot(
    aes(
      x = reorder(cd_id, bc_snap_recipients),
      y = bc_snap_recipients
    )
  ) +
  geom_col(fill = "steelblue") +
  coord_flip() +
  labs(
    title = "SNAP Recipients by Community District (Top 25)",
    x = "Community District",
    y = "SNAP Recipients"
  ) +
  theme_minimal()

Figure 7.2: Fig 2: Mean number of SNAP recipients by community district, ordered from highest to lowest for the top 25. Each horizontal bar represents a distinct community district. Community district identifiers are displayed on the y-axis and SNAP recipients are displayed on the x-axis.

Across community districts, the mean number of SNAP recipients per month was 28580, with a median of 25456.

7.6 Merging

Finally, I merged the two datasheets using the community district names I created earlier.

merged <- BoroReport %>%
  left_join(events_cd, by = "cd_id")

knitr::kable(
  head(merged, 25),
  caption = "First 25 rows of merged"
)

7.7 Linear Regression

I then conducted a linear regression to determine whether number of permitted events predicts number of SNAP recipients. The model was statistically significant, F(1, 723) = 45.34, p < .001, and explained approximately 6% of the variance in SNAP recipients (R² = .059). The number of events was a significant negative predictor of SNAP recipients, b = −21.30, SE = 3.16, t(723) = −6.73, p < .001.

(Note: The model dropped all rows with missing event counts. This means that all joint-interest areas were dropped from the analysis, as well as any months for which there was no event count data)

model1 <- lm(bc_snap_recipients ~ n_events, data = merged)
summary(model1)
#> 
#> Call:
#> lm(formula = bc_snap_recipients ~ n_events, data = merged)
#> 
#> Residuals:
#>    Min     1Q Median     3Q    Max 
#> -28081 -12534  -2646   8391  44572 
#> 
#> Coefficients:
#>              Estimate Std. Error t value Pr(>|t|)    
#> (Intercept) 30043.800    715.843  41.970  < 2e-16 ***
#> n_events      -21.303      3.164  -6.733 3.38e-11 ***
#> ---
#> Signif. codes:  
#> 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 16210 on 723 degrees of freedom
#>   (986 observations deleted due to missingness)
#> Multiple R-squared:  0.059,  Adjusted R-squared:  0.0577 
#> F-statistic: 45.34 on 1 and 723 DF,  p-value: 3.385e-11

ggplot(
  merged,
  aes(
    x = n_events,
    y = bc_snap_recipients
  )
) +
  geom_point(alpha = 0.6) +
  geom_smooth(method = "lm", se = TRUE, color = "red") +
  labs(
    title = "Permitted Events and SNAP Recipients by Community District",
    x = "Permitted Events",
    y = "SNAP Recipients"
  ) +
  theme_minimal()
#> `geom_smooth()` using formula = 'y ~ x'
#> Warning: Removed 986 rows containing non-finite outside the scale
#> range (`stat_smooth()`).
#> Warning: Removed 986 rows containing missing values or values
#> outside the scale range (`geom_point()`).

Figure 7.3: Fig 3: Relationship between permitted events and SNAP recipients by community district. Each point represents a distinct community district, and the line shows the linear association between event counts and SNAP recipients.

7.8 Conclusion

Despite the significant p-value of this analysis, there are a number of limitations. As mentioned in the introduction, the number of SNAP recipients is an imperfect measure of economic well-being (not to mention holistic well-being). Likewise, permitted events are an imperfect indicator of social gatherings in an area. At a more granular level, community districts are not normalized by population size, and major hubs of social activity—such as parks—are excluded from the regression.

However, these limitations point to ways in which data collection could be improved. Below, I outline several possibilities for instantiating such improvements:

Better Dependent Variables

To meaningfully assess quality of life in NYC, future datasets should include more varied indicators of well-being and capture outcomes across the income distribution. Ideally, validated population-level measures of well-being and social connectedness would be available for use as dependent variables. In addition, economic proxies for well-being (such as median income) should be collected. Diverse datasets of this sort would provide a more complete picture of the psychological and economic well-being of NYC residents.

More Information about Social Gatherings

Currently, NYC Open Data has information about permitted events. Yet, there are countless other social gatherings that could be quantified as well. These include volunteer opportunities, Meetup groups, Eventbrite activities, Reddit meetups, and more. While an exhaustive catalog of social gatherings is not feasible, expanded coverage of accessible, low-barrier events would strengthen any analyses of social life in the city. It would also allow analysts to subdivide events in meaningful ways.

Geographic Information

Community districts provide a useful organizational unit, but many NYC datasets lack this data. In addition, even more detailed neighborhood-level data on events might provide information about areas with a shortage (or surplus) of social activity. Identifying such areas might support more strategic intervention. Finally, knowledge of individuals’ willingness (or lack of willingness) to travel might provide yet more valuable information. The prominence of parks in the event data suggests that social life is often organized around specific hubs. The practical accessibility of these hubs is yet another concept worth exploring.

Concrete Suggestions

There is no “control New York City.” As such, causality cannot be established through the analyses I describe. Nevertheless, if evidence were to suggest that certain types of social activities were associated with positive psychological outcomes, it would then be possible to recommend concrete actions to citizens who wished to improve civic and social life in New York. In this way, improved data infrastructure could help foster a stronger sense of civic autonomy among New Yorkers – as well as a happier, healthier New York City.