posc data
DTLG
methodsanddata.pdf
Violent Crime Across U.S. States and Jurisdictions
Violent Crime Across U.S. States and Jurisdictions
This section is based on the violent crime subject of the previous POSC proposal. Structural
factors can explain variation in violent crime across U.S. states. The provided dataset
operationalizes structural context using three predictors in the file: poverty, Title I spending per
student, and health insurance coverage. This section aims to demonstrate how each concept is
quantified, the distribution of the variables, and the level of confidence a reader should have in
the data.
Research Design
The research design employed in this study is a non-experimental, cross-sectional, ecological
research design. The SPSS file contains 52 observations of the state or jurisdiction that is the unit
of analysis. The data is a compilation of publicly available secondary data available through the
FBI Uniform Crime Reporting Program on violent crime, the U.S. Census Bureau on poverty
and health insurance, and Title I spending. Since these data are official administrative and
survey-based, rather than the result of original fieldwork, the design enables broad comparisons
across units at relatively low cost. Meanwhile, the data capture summarizes conditions, not
personal experiences; hence, any results should be kept at the ecological level of analysis.
There is also a significant temporal constraint on the design. The dependent variable represents
violent crime in 2020, with the independent variables representing 2021 and 2024, and thus, the
paper should not assume that the relationships are causal but only associative. Also, there is one
missing value in Title I spending, which decreases the number of valid cases of that variable to
51. These options continue to facilitate an exploratory comparison across states and jurisdictions
but undermine causal ordering and diminish comparability across measures. This section thus
highlights clear measurement, descriptive distributions, and a cautious account of validity and
reliability.
Dependent Variable: Violent Crime Rate
Violent crime is defined as serious interpersonal crimes that involve force or the threat of force.
The category of murder and nonnegligent manslaughter, rape, robbery, and aggravated assault is
included in the FBI Uniform Crime Reporting program, and the present study operationalizes the
concept as the average violent crime rate per 100,000 inhabitants in 2020 (FBI, 2019). The face
validity of this operationalization is high since the variable is a direct measure of the exact
behavior that the hypothesis aims to explain. It is also helpful in comparing across jurisdictions
since the rate normalizes reported offenses based on population size. To replicate this measure, a
researcher would need the number of violent offenses reported to the UCR per annum and the
corresponding population denominator for the same year.
Variable N Mean Median SD Min to
Max
Range Shape
Violent
crime rate
(2020)
52 391.65 373.68 177.11 108.58
to
999.84
891.26 Moderately
right skewed
Poverty rate
(2024)
52 12.81 11.70 4.77 7.50 to
41.07
33.57 Strongly
right skewed
Title I
spending per
student
(2021)
51 328.28 310.30 121.00 122.70
to
874.20
751.50 Strongly
right skewed
Health
insurance
coverage
(2024)
52 90.89 91.44 3.05 80.95 to
96.41
15.46 Mildly left
skewed
Table 1: Descriptive statistics for study variables
Figure 1: Histogram of the violent crime rate
Table 1 and Figure 1 summarize the distribution of violent crime. The average rate of violent
crime is 391.65, the median is 373.68, the standard deviation is 177.11, and the observed values
range from 108.58 to 999.84. The variable is moderately skewed to the right because the mean is
greater than the median and the histogram has a long right tail. The reliability is fairly high since
the FBI has a standard reporting system and provides estimates of those agencies that do not
report, but voluntary reporting, underreporting to the police, and agency-level discrepancies in
classification may still result in a measurement error (FBI, 2021). It is also a cautionary measure
by the FBI that simplistic rankings are not to be trusted, and therefore, this is best used as a
rough comparative measure, but not as an exhaustive account of public safety in all jurisdictions.
Independent Variable 1: Poverty Rate
Poverty is used to define the degree of material deprivation in a state or jurisdiction. The 2024
estimated rate of poverty is the measure of this concept taken in the study and denoted in the
dataset as the percentage in poverty. The official measure of poverty categorizes individuals as in
poverty when their family's money income is below a threshold that depends on family size and
family composition, and the Small Area Income and Poverty Estimates program provides annual
estimates by combining survey data, population estimates, and administrative records (U.S.
Census Bureau, 2026a). This operationalization provides the study with a direct measure of
economic hardship as opposed to a more abstract measure of inequality. A researcher wishing to
reproduce the variable would obtain the 2024 state or jurisdiction estimate from the Census
poverty series and note the percentage reported for each unit.
Figure 2: Histogram of poverty rate
Table 1 and Figure 2 indicate that there is a wide range of poverty among the 52 observations.
The average poverty level is 12.81%, the median is 11.69%, and the standard deviation is 4.77
percentage points, with a range of 7.50% to 41.07%. The huge difference between the maximum
and the rest of the distribution gives rise to a strong right skew, indicating at least one unusually
high poverty value. This variable is the best content validity of economic deprivation since it is a
direct reflection of the official definition of poverty, but it lacks the depth of poverty, wealth
disparities, and general income inequality. The reliability is also quite high, as SAIPE relies on
model-based estimation to minimize sampling error, but the estimates are still based on modeling
assumptions rather than direct enumeration (U.S. Census Bureau, 2026b).
Independent Variable 2: Title I Spending per Student
Title I expenditure per pupil is a state expenditure on economically disadvantaged student
programs. The variable is based on NCES Table 8 in fiscal year 2021 and operationalized as Title
I expenditures per pupil. According to NCES, this figure is current and carryover Title I spending
divided by the total student membership, both Title I eligible and non-eligible (NCES, n.d.). This
measure is not a full measure of education expenditure, but it is a measure of a policy initiative
that is focused on directing more resources to educational disadvantage. A researcher would
replicate the measure by gathering the reported Title I spending and student membership by each
state or jurisdiction, and dividing the spending by the number of pupils using the NCES formula.
Figure 3: Histogram of Title I spending per student
Table 1 and Figure 3 show that the mean and median of Title I spending per student are 328.28
and 310.30, respectively. The standard deviation is 121.00, and the values range from 122.70 to
874.20, indicating a clearly right-skewed distribution. There is also one missing value in the
SPSS output; thus, the statistics of this variable are computed using 51 valid values. Face validity
is moderate since the measure effectively measures the intended educational expenditure, but it is
only an indirect measure of the institutional support that may influence the long-term crime
outcomes. The reliability of NPEFS is greater than that of most survey measures, as it relies on
administrative finance records, review procedures, edit checks, and imputation processes;
however, comparability may be affected by cross-jurisdictional accounting differences and the
unusual pandemic-era context of fiscal year 2021 (NCES, 2023).
Independent Variable 3: Health Insurance Coverage
Health insurance coverage is a proportion of the population that has access to some health
coverage. The dataset quantifies this concept as the proportion of the insured in 2024 in terms of
health insurance. According to the Census Bureau, the estimates of ACS health insurance are an
average of current coverage annually, and the ACS estimates are available on a state level since
its sample size is around 3.5 million addresses every year (U.S. Census Bureau, 2024). This
variable is not a direct measure of crime or law enforcement, but it provides a plausible measure
of social protection and access to institutions that could alleviate social strain. A researcher
might replicate the variable by removing the 2024 percentage insured from the ACS state health
insurance tables for each unit.
Figure 4: Histogram of health insurance coverage
The distribution of health insurance coverage does not appear similar to the other predictors.
Table 1 and Figure 4 indicate a mean of 90.89%, a median of 91.44%, a standard deviation of
3.05 percentage points, and a range of 80.95% to 96.41%. Since the majority of observations are
concentrated around the upper part of coverage, the histogram exhibits a weak left skew as
compared to the right skew of Poverty and Title I spending. The criterion-based validity of this
variable is reasonable since federal health insurance series like ACS and SAHIE are expected to
move in the same direction, although the methods used are different, and the ACS publishes
separate documentation on sampling and data accuracy to justify reliability (U.S. Census Bureau,
2024). Nonetheless, insurance coverage can only be an indirect indicator of the broader notion of
social well-being; thus, its substantive relationship to violent crime is more abstract than
concrete.
Assessment of Data Quality
The descriptive statistics indicate that the independent variables are not distributed in the same
pattern. Poverty and Title I spending are highly right-skewed, violent crime is moderately
right-skewed, and health insurance coverage is compressed at high values and slightly
left-skewed. These patterns are important because skewed variables may shift the mean away
from the median and make relationships sensitive to a few large-value observations. The
histograms thus indicate that any subsequent statistical test must verify the presence of outliers
and also account for robust approaches if the model is adopted and requires more robust
distributional assumptions. At least, the descriptive phase demonstrates that the conditions of the
state and jurisdiction levels are different enough to be compared.
In general, the data are sufficient, as each variable is based on a transparent federal source with
published definitions and methods. Violent crime and poverty are the most direct measures,
whereas Title I spending and health insurance coverage are more indirect proxies of institutional
support and social protection. The primary weaknesses are the timing of measures across years,
the ecological design, the absence of Title I value, and the fact that official reports may not
reflect underlying behavior or hardship. These limits do not render the project invalid, but they
limit the claims a research paper can make. The most justifiable inference from this design will
be a patterned association across states and jurisdictions rather than a causal explanation.
References
Federal Bureau of Investigation. (2019). Violent crime.
https://ucr.fbi.gov/crime-in-the-u.s/2019/crime-in-the-u.s.-2019/topic-pages/violent-crime
Federal Bureau of Investigation. (2021). FBI releases 2020 crime statistics.
https://www.fbi.gov/news/press-releases/fbi-releases-2020-crime-statistics
National Center for Education Statistics. (2023). Documentation for the NCES Common
Core of Data National Public Education Financial Survey (NPEFS), school year 2020 to
2021, fiscal year 2021: Provisional file version 1a.
https://nces.ed.gov/ccd/pdf/2023302_FY_21_NPEFS_Documentation_August_18_2023.
National Center for Education Statistics. (n.d.). Table 8. Title I allocations and Title I
expenditures per pupil for public elementary and secondary education, by state or
jurisdiction: FY 2021.
https://nces.ed.gov/ccd/tables/NPEFS_FinanceTable8_FY21_1a.asp
U.S. Census Bureau. (2024). Health insurance statistics: CPS, SIPP, ACS, and SAHIE.
https://www.census.gov/data/developers/data-sets/Health-Insurance-Statistics.html
U.S. Census Bureau. (2026a, January 27). Poverty statistics: CPS and SAIPE.
https://www.census.gov/data/developers/data-sets/Poverty-Statistics.html
U.S. Census Bureau. (2026b, January 12). SAIPE methodology.
https://www.census.gov/programs-surveys/saipe/technical-documentation/methodology.h
tml
