Relationship Between Gun Ownership and Ethnicity Groups

Data Project: Relationship Between Gun Ownership And Ethnicity Groups

Introduction:

The purpose of this project is to investigate if there is a relationship between gun ownership in the households of US citizens aged 18 or older and ethnicity groups. A clearer way to state the research question is “are people belonging to particular ethnicity groups more likely to own a gun in their households?”

Widespread gun ownership is the subject of many debates on crime rate in the US. In that respect this study could also provide useful insights into indentifying whether people belonging to a particular ethnicity group feel they are not sufficiently protected by police force.

Data:

The study uses American National Elections Study (ANES) data for the year 2012. ANES is a survey of voters in the United States, conducted before and after every presidential election. For the year 2012 the data were collected in 2 different modes (Internet mode and Face-to-face mode), using 2 separate samples. The global sample is composed of 5914 cases (2.054 face-to-face mode and 3.860 Internet mode).

The observations are US citizens aged 18 or older.

For the Internet mode, all study participants were members of the Knowledge Panel, a panel of regular survey participants administered by GfK (formerly Knowledge Networks). Panelists were recruited using two probability sampling methods: address-based sampling (ABS) and random-digit dialing (RDD). A sample of Knowledge Panelists received invitations to take the ANES Time Series survey. This sample was limited to U.S. citizens who would be at least 18 years old by Election Day, November 6, 2012, and was limited to one person per household.

For the face-to-face mode, all sampled persons were interviewed in person. The sample includes a nationally-representative main sample and two oversamples: one of blacks and one of Hispanics. The first stage of sampling consisted of stratifying the 48 contiguous states and the District of Columbia into nine regions corresponding to Census Divisions. These Census Divisions constituted the study’s strata. Within each region, a number of census tracts was then randomly selected. The number of tracts selected per region was proportional to the region’s proportion of the U.S. adult population. The second stage of sampling consisted of the random selection of residential addresses within each tract. Addresses for the black and Hispanic oversamples were selected from tracts with relatively high proportions of one or both of these populations. The third and final stage of sampling was the selection of one eligible person per household.

It is important to underline that the data come from a survey and not from an experiment (the researchers did not assign different groups of subjects to various treatments), consequently the study can be characterized as observational. For this reason, it can establish only correlation between the variables of interest and not causation. However, since the sample design criteria included: probability sampling, stratified sampling and random sampling, the study’s findings can be generalized to the entire population of interest (US citizens aged 18 or older). One last note on the scope of inference is that one potential source of bias might be represented by privacy concerns on respondents’ answers. In order to limit its effects, for a portion of the face-to-face interview respondents were allowed to answer questions privately.

The study uses two variables chosen from those collected by ANES survey:

Respondent’s race and ethnicity group (code name: “dem_raceeth”): it is a categorical variable whose values are: “White-Non Hispanic”, “Black-Non Hispanic”, “Hispanic”, “Other-Non Hispanic”;

Gun ownership in the respondent’s household (code name: “owngun_owngun”): it is a categorical variable whose values are " Yes“;”No“.

Exploratory data analysis:

Before exploring the data, we create a dataset with the two variables of interest and we give them a clearer label.

myvars <- c(114,117)

project_data <- anes[myvars]

colnames(project_data) <- c("Ethnicity.Group","Gun.Ownership")

There are some observations with missing values. Filtering them out brings the number of observations to 5704. The sample size remains significant for the study.

project_data = project_data[complete.cases(project_data),]

nrow(project_data)

## [1] 5704

The traditional way to explore a categorical value is through a frequency table. We can easily compute the number (and the proportion) of respondents who own a gun in their households.

table(project_data$Gun.Ownership)

##

## Yes No

## 1814 3890

table(project_data$Gun.Ownership)/5704

##

## Yes No

## 0.3180224 0.6819776

We are now able to describe graphically the distribution of this categorical variable.

barplot(table(project_data$Gun.Ownership)/5704,main="Gun Ownership")

The overall rate of gun ownership in the sample is roughly 31.8%. However, what we are really interested in is trying to figure out how this variable is related to the grouping variable:“Ethnicity Group”. We can explore this relation through a contingency table.

Contingency Table (Counts)

mytable <- table(project_data$Gun.Ownership,project_data$Ethnicity.Group)

mytable

##

## White Non-Hispanic Black Non-Hispanic Hispanic Other Non-Hispanic

## Yes 1367 164 179 104

## No