A data set on enforcement operations; illustrates Poisson regression

Holland (2015) data set used in Rainey (2023) to illustrate the difference between the "average-of-simulations" and directly transformed quantities of interest. These are the data to reproduce Model (1) for Bogota, Lima, and Santiago from Holland's Table 2 on p. 368. Note that the coefficients in Table 2 are standardized and exponentiated with (Stata's) robust standard errors.

Usage

holland2015

Format

A data frame with 89 district-level observations across three cities:

city: the observation's city (Holland performs separate analyses for each city)
district: the name of the district (the unit of observation)
operations: the number of enforcement operations conducted per month by a district (averaged across three or more months per district)
lower: share of lower-class residents in a district
vendors: the number of unlicensed street vendors (in thousands)
budget: district budget per capita
population: district population (unclear scale)

For further details, see Holland (2015, p. 364) and pp. 3-6 of the file ajps12125-sup-0001-SupMat.docx available on the journal website.

Details

In Rainey (2023), I focus on the following hypothesis:

"My first hypothesis is that enforcement operations drop off with the fraction of poor residents in an electoral district. So district poverty should be a negative and significant predictor of enforcement, but only in politically decentralized cities [Lima and Santiago]. Poverty should have no relationship with enforcement in politically centralized cities [Bogota] once one controls for the number of vendors (p. 362)"

References

Holland, Alisha C. 2015. "The Distributive Politics of Enforcement." American Journal of Political Science 59(2): 357–71. doi:10.1111/ajps.12125 .

Holland, Alisha. 2014. "Replication data for: The Distributive Politics of Enforcement." Harvard Dataverse, V2. doi:10.7910/DVN/24859 .

Rainey, Carlisle. 2023. "A Careful Consideration of CLARIFY: Simulation-Induced Bias in Point Estimates of Quantities of Interest." Political Science Research and Methods. Forthcoming. doi:10.1017/psrm.2023.8 .

Examples


# a simple example

# load data
holland <- crdata::holland2015

# table of observations per city
table(holland$city)
#> 
#>   bogota     lima santiago 
#>       19       36       34 

# formula corresponds to model 1 for each city in holland (2015) table 2
f <- operations ~ lower + vendors + budget + population

# fit poisson regression model for Santiago
fit <- glm(f, family = poisson, data = holland, subset = city == "santiago")
summary(fit)
#> 
#> Call:
#> glm(formula = f, family = poisson, data = holland, subset = city == 
#>     "santiago")
#> 
#> Coefficients:
#>               Estimate Std. Error z value Pr(>|z|)    
#> (Intercept)  2.6190747  0.5185130   5.051 4.39e-07 ***
#> lower       -0.0376151  0.0088506  -4.250 2.14e-05 ***
#> vendors     -0.2209762  0.1216055  -1.817   0.0692 .  
#> budget      -0.0005983  0.0004912  -1.218   0.2232    
#> population   0.0154822  0.0093948   1.648   0.0994 .  
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> 
#> (Dispersion parameter for poisson family taken to be 1)
#> 
#>     Null deviance: 226.08  on 33  degrees of freedom
#> Residual deviance: 167.49  on 29  degrees of freedom
#> AIC: 226.79
#> 
#> Number of Fisher Scoring iterations: 7
#>