Pith. sign in

REVIEW 4 major objections 5 minor 13 references

An Investigation of the Relationship Between Crime Rate and Police Compensation

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Higher police pay is associated with lower crime rates in Baltimore, the paper claims.

desk verdict A Baltimore police-pay vs. crime correlation resting on a single unquantified regression figure; the paper's own conceded limitations undercut the headline claim. read the letter →

arxiv 2411.14632 v1 pith:KRTW46IE submitted 2024-11-21 cs.CY stat.AP

Jhancy Amarsingh , Likhith Kumar Reddy Appakondreddigari , Ashish Nunna , Charishma Choudary Tummala , John Winship , Alex Zhou , Huthaifa I. Ashqar This is my paper · ORCID
classification cs.CYstat.AP
keywords policecompensationcrimerateBaltimorenegativecorrelationlinearregressionpublicpolicysociallosspayrolldata
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether police pay and crime move together in Baltimore, using public crime incident records and city payroll data from 2011 to 2021. It compresses more than 250,000 crime records into a month-to-month series, builds a parallel police salary and gross-pay series, and fits a linear regression to the two. The central result is a negative slope: periods with higher average police compensation are associated with lower crime rates. The authors read this as initial evidence that investing in officer pay could help reduce crime, and they connect it to economic theories of crime to suggest how state policymakers could use the relationship.

What carries the argument

The central device is a simple linear regression fitted to two aggregated time series: a month-to-month Baltimore crime rate built from over 250,000 incident records, and an annual police salary and gross-pay series from city payroll data. The regression's negative coefficient is the entire empirical result; the paper's stated next step is to embed that coefficient in a multiple regression based on an established social-loss function, $L = L(D, C, bf, O)$, where $D$ is crime damage, $C$ is the cost of combating offenses including police pay, $bf$ is punishment cost, and $O$ is the crime rate.

What would settle it

Recompute the regression on detrended series, for example first differences or residuals after removing a year trend; if the negative coefficient vanishes or changes sign, the reported correlation is an artifact of shared trends. A second check is to refit the model with the 2015-2016 spike window excluded; if the slope disappears, that window was the main driver.

Watch

Extended reading notes

Core claim

On its own terms, the paper's finding is that Baltimore police compensation and citywide crime rate are negatively correlated over the study period, with the linear regression showing that as average salary and gross pay rise, the crime rate falls. The paper treats this correlation as a signal that higher compensation is associated with greater officer competence and better enforcement, a 'get what you pay for' effect, and it argues the finding is interesting precisely because public-sector pay is not set by a market mechanism. It stops short of claiming causation, instead describing the result as an initial finding that needs refinement before it can guide policy.

Load-bearing premise

The result assumes the negative slope between the salary curve and the crime curve is a meaningful association rather than a coincidence of two long-run trends moving in opposite directions.

Editorial extensions

If this is right

  • If the correlation is correct, raising police compensation is associated with lower crime, giving state leadership a quantitative argument for increasing police budgets.
  • The association points to a policy trade-off between a smaller, more senior and higher-paid force and a larger force with more junior officers; the cited crime-prevention literature suggests both specialized units and extra patrols matter.
  • The paper's own conclusion requires the salary series to be adjusted for inflation and for seniority mix before the result can be used in budgeting.
  • The correlation can be extended into a social-loss minimization model that weighs crime damage, enforcement cost, and punishment cost together, rather than pay alone.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The negative slope probably tracks two long-run trends, falling national crime rates and rising wages, so a detrended regression (first differences or a year term) is the natural check on whether the association survives.
  • The 2015 crime spike and the 2020 compensation dip are treated as disruptions to smooth over; excluding or explicitly modeling those windows could flip the sign of the fit.
  • If the association survives detrending, the mechanism is likely the seniority mix of the force rather than pay level alone, which the paper itself flags as an open question.
  • Applying the same two-series regression to another city, such as the San Francisco comparison data already used as a sanity check, would show whether the pattern generalizes beyond Baltimore.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper investigates whether police salaries/compensation and crime rates are correlated in Baltimore from 2011 to 2021 using public Open Baltimore data. The authors describe a linear regression comparison between the two time series and claim in the Abstract and Conclusion that their 'initial results show a negative correlation between salary/compensation levels and crime rates.' They then discuss policy implications, echoing Becker's economic model of crime and Sherman et al.'s review of crime prevention, and propose future work such as inflation adjustment and multiple regression.

Significance. If supported, a robust negative association between police compensation and crime rates would be policy-relevant for municipal budgeting and public safety debates. The paper has the merit of using publicly available data and addressing a timely question. However, as submitted, the central claim is not statistically substantiated: no regression coefficients, confidence intervals, p-values, or goodness-of-fit measures are reported, and the analysis does not address the obvious confounds of time trends, seasonality, inflation, or the 2015 Freddie Gray crime spike. The paper also explicitly concedes in the Conclusion that salaries were not inflation-adjusted and that the seniority mix was not examined. These gaps are load-bearing because the headline conclusion rests entirely on Figure 8, a raw scatter-style regression of two co-trending series. The scientific contribution is therefore not currently established.

major comments (4)
  1. [Results, Figure 8; Abstract; Conclusion] The central claim of a negative correlation relies on a 'negative correlated linear regression' shown in Figure 8, yet the paper reports no regression equation, slope, intercept, R-squared, confidence interval, p-value, or number of observations. Without these statistics, a regression of two trending series (rising nominal salaries and generally declining crime) can produce a negative slope even when no meaningful association exists. The paper neither detrends the series nor includes a time index, and it does not control for seasonality or the 2015 Freddie Gray spike. Since this figure is the only quantitative evidence for the abstract's claim, the conclusion is unsupported.
  2. [Step 2 and Conclusion] The manuscript admits the salary analysis is incomplete: the Conclusion states that 'the salary analysis should be adjusted for inflation using the CPI index' and that assessing 'the seniority mix that undergirds the salary data is important.' Both are acknowledged to be missing, meaning the reported negative correlation could be an artifact of nominal salary growth over time. Additionally, the sentence in Step 2 is garbled ('some additional salary / compensation analysis required to adjust for things such as the mix of seniority levels techniques to adjust for inflation') and does not clarify whether any adjustment was actually performed.
  3. [Step 1 and References] The 'Freddy Grey effect' adjustment is cited to Kolodrubetz, but the in-text citation number (reference 3) points to an unrelated IEEE paper on bike-sharing systems; the correct reference appears to be reference 4, the Kolodrubetz and Ashqar report. More importantly, no method for this adjustment is described anywhere, so a reader cannot verify how the crime time series was constructed, how the 2015 spike was handled, or whether different adjustment choices would change the result. This matters because the crime series is a direct input to the regression underlying the paper's main claim.
  4. [Data Facts] The Data Facts section states that the crime analysis covers '5 years (2016 to 2020),' while the study period is 2011-2021. This inconsistency makes it unclear which time window is used for the regression in Figure 8. The paper also never defines the crime rate metric (e.g., per capita rate vs. raw monthly counts), and Figure 8 lacks axis labels and units, making the magnitude of the alleged correlation impossible to assess.
minor comments (5)
  1. [Throughout] The name 'Freddy Grey' is misspelled; it should be 'Freddie Gray.'
  2. [Methods (Pandas, Matplotlib, Seaborn sections)] Large portions of the Methods text are generic tutorial descriptions of Pandas, Matplotlib, and Seaborn that are not specific to this research and should be removed or drastically condensed.
  3. [References] The reference list is misnumbered relative to the in-text citations. For example, the text cites 'Kolodrubetz3' and 'Press4,' but in the reference list reference 3 is an Ashqar et al. bike-sharing paper and reference 6 is the Press study; the correct mapping needs to be fixed.
  4. [Figure 8] The figure lacks axis labels, units, and any indication of what points or series are plotted, so the reader cannot interpret the regression or its range.
  5. [Conclusion] The paper repeatedly states 'our initial results show a negative correlation' as if it were an established finding. The language should be clearly qualified as exploratory and hypothesis-generating, not inferential, given the absence of statistical reporting.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the negative correlation is an empirical regression result; the Freddie Gray adjustment cites a co-authored report, but the central claim does not reduce to that citation.

full rationale

The paper's derivation chain is: (1) build a monthly crime-rate series from BPD crime data; (2) build a salary/gross-pay series from Open Baltimore employee data; (3) run a linear regression comparing the two; (4) read the negative slope in Figure 8 as evidence of a negative correlation; (5) interpret via Becker and Sherman. At no point is crime rate defined in terms of salary, nor is a fitted parameter relabeled as an independent prediction. The negative slope is the regression output itself; Figure 8 is described as 'a negative correlated linear regression' and the Abstract/Conclusion restate that fitted result. That is descriptive statistics, not a derivation from an input that already contains the conclusion. The only self-citation in the load-bearing chain is the Freddie Gray adjustment: 'Here we leveraged work done by Kolodrubetz3 to account for the ''Freddy Grey effect.''' The intended reference (reference 4, Kolodrubetz and Ashqar) is co-authored by Huthaifa Ashqar, an author of this paper; the in-text reference number is wrong, pointing to an unrelated bike-sharing paper. The adjustment method is not described, so the reader cannot audit its effect. However, the cited report is an external crime-incident analysis, and the paper does not claim the adjustment is what manufactures the negative slope; there is no exhibited equation showing the correlation reduces to the report's output. Thus this is a minor, underdescribed self-citation and a reference error, not a circular step. The Conclusion admits missing controls: 'assessing the seniority mix that undergirds the salary data is important' and 'the salary analysis should be adjusted for inflation using the CPI index.' These are validity threats (spurious correlation due to common trends), not circularity patterns. Similarly, the absence of coefficients, intervals, and detrending in Figure 8 is a correctness risk. Per the hard rules, weak or nonstandard statistics do not constitute circularity. Overall score 2 due to the minor non-load-bearing self-citation and reference error; the central claim is not forced by construction or by a self-citation chain.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim depends on fitting a regression line to two under-described time series (free parameter 1), a smoothing choice (free parameter 2), and an undocumented adjustment (free parameter 3). The analysis also assumes the crime and salary series are valid and comparable (domain assumptions) and that linear regression is appropriate without diagnostics. No new entities are introduced.

free parameters (3)
  • Linear regression slope and intercept for crime vs compensation = not reported
    The negative correlation claim rests on the fitted slope in Figure 8, but no coefficient, R-squared, or p-value is reported.
  • Crime rate smoothing window = 12 months
    The paper says it smoothed the crime rate over a 12-month period to handle seasonality, but the exact smoothing method is not specified.
  • Freddie Gray effect adjustment = not specified
    The paper says it adjusted for the 2015 Freddie Gray spike using prior work, but the adjustment method and parameter values are not described.
assumptions (4)
  • domain assumption The crime rate time series derived from incident data is a valid measure of crime intensity.
    The paper summarizes over 250,000 crime entries into a monthly rate but does not specify per-capita normalization or crime-type inclusion.
  • domain assumption Police salary and gross pay data are comparable across years as a measure of compensation.
    The paper does not adjust for inflation or seniority mix; it admits in the Conclusion that CPI adjustment is still needed.
  • domain assumption A negative slope in a simple regression of the two raw time series indicates a meaningful relationship.
    No controls for common time trends, autocorrelation, or confounders are included, so the slope may reflect unrelated trends.
  • standard math Ordinary least squares assumptions are satisfied.
    The paper uses linear regression without checking linearity, independence, or residual diagnostics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Investigation of the Relationship Between Crime Rate and Police Compensation." pith.science (2026). https://pith.science/paper/KRTW46IE

@misc{pith2026241114632,
  author       = {Pith},
  title        = {Pith review of: An Investigation of the Relationship Between Crime Rate and Police Compensation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KRTW46IE}},
  note         = {Machine review of arXiv:2411.14632}
}
read the original abstract

The goal of this paper is to assess whether there is any correlation between police salaries and crime rates. Using public data sources that contain Baltimore Crime Rates and Baltimore Police Department (BPD) salary information from 2011 to 2021, our research uses a variety of techniques to capture and measure any correlation between the two. Based on that correlation, the paper then uses established social theories to make recommendations on how this data can potentially be used by State Leadership. Our initial results show a negative correlation between salary/compensation levels and crime rates.

Figures

Figures reproduced from arXiv: 2411.14632 by the authors.

Figure 1
Figure 1. Crime Rate - City of Baltimore Crime Data from 2011-2021 [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Salary/Gross Pay Data for Baltimore Employees, specific to the Police [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Pandas Example Method for Crime rate analysis Data Visualization: Everything we touch seems to generate a lot of data. While buzzwords like "big data" may have disappeared, the data itself still exists. When it comes to understanding, processing, analyzing, and communicating this data, the old technology just won't cut it. It doesn't matter how big your screen is, or if it's curved, the CSV you import into Excel is … view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Seaborn example Introduction to TreeMap Chart: Treemap Chart is intended for the visualization of hierarchical data in the form of nested rectangles. Each level of such a tree structure is depicted as a colored rectangle, often called a branch, which contains other rec…
Figure 5
Figure 5. Figure 5: Treemap Example Our team also leveraged several other methods, described briefly below. countplot ● Show the counts of observations in each categorical bin using bars. ● A count plot can be thought of as a histogram across a categorical, instead of quantitative, variab…
Figure 6
Figure 6. Figure 6: Salary Data Summary There are around 1897 job classes and 71 agency IDs. Gross pay values are left skewed which indicates that the lowest gross pay values are more, skilled police officers alone get highly paid. There is an increase in gross pay and annual salary over …
Figure 7
Figure 7. Figure 7: Police Compensation Over Time. Results: Crime Rates vs Salaries/Compensation: Negative Correlation - when there is a negative correlation between two variables, the variables move relative to each other which means if one variable increases then the other decreases. Wh…
Figure 8
Figure 8. Figure 8: Negative Correlation between Crime Rate and Police Compensation [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

  1. [1]

    Crime and Punishment: An Economic Approach

    Becker, Gary S., William M. Landes (1974). “Crime and Punishment: An Economic Approach.” National Bureau of Economic Research (https://www.nber.org/books-and- chapters/essays-economics-crime-and-punishment)

    work page 1974

  2. [2]

    Preventing Crime: What Works, What Doesn’t, What’s Promising

    Sherman, Lawrence W., Denise C. Gottfredson, Doris L. MacKenzie, John Eck, Peter Reuter, and Shawn D. Bushway (1998). “Preventing Crime: What Works, What Doesn’t, What’s Promising.” U.S. Dept. of Justice (https://www.ojp.gov/pdffiles/171676.pdf)

    work page 1998

  3. [3]

    I., Elhenawy, M., Almannaa, M

    Ashqar, H. I., Elhenawy, M., Almannaa, M. H., Ghanem, A., Rakha, H. A., & House, L. (2017, June). Modeling bike availability in a bike-sharing system using machine learning. In 2017 5th IEEE International Conference on Models and Technologies for Intelligent Transportation Systems (MT-ITS) (pp. 374-378). IEEE

    work page 2017

  4. [4]

    Crime in Baltimore: A Retroactive Analysis of Crime Incidents in Baltimore City from 2014 to 2019

    Kolodrubetz, Samuel, and Huthaifa Ashqar (2020). “Crime in Baltimore: A Retroactive Analysis of Crime Incidents in Baltimore City from 2014 to 2019.” University of Maryland, Baltimore County

    work page 2020

  5. [5]

    Predicting residential property value: a comparison of multiple regression techniques

    Whieldon, Lee, and Huthaifa I. Ashqar. "Predicting residential property value: a comparison of multiple regression techniques." SN Business & Economics 2, no. 11 (2022): 178

    work page 2022

  6. [6]

    Press, S.J. (1971). Some Effects of an Increase in Police Manpower in the 20th Precinct of New York City. New York: New York City Rand Institute

    work page 1971

  7. [7]

    What Is Known About Deterrent Effects of Police Activities

    Chaiken, Jan M. (1978). “What Is Known About Deterrent Effects of Police Activities.” In James A. Cramer, ed., Preventing Crime. Beverly Hills, California: Sage. 18

    work page 1978

  8. [8]

    An Experimental Evaluation of the Phoenix Repeat Offender Program

    Abrahamse, Allan F., Patricia A. Ebener, Peter W. Greenwood, Nora Fitzgerald, and Thomas E. Kosin (1991). “An Experimental Evaluation of the Phoenix Repeat Offender Program.” Justice Quarterly 8:141–168

    work page 1991

Show all 13 references
  1. [9]

    R., Aven, R., Drohan, C., & Ashqar, H

    Perfilyeva, A., Miskin, V. R., Aven, R., Drohan, C., & Ashqar, H. I. (2024). Estimating Variability in Hospital Charges: The Case of Cesarean Section. arXiv preprint arXiv:2411.08174

    2024 arXiv

  2. [10]

    https://governor.maryland.gov/2021/10/18/transcript-october-15-press-conference/

    2021

  3. [11]

    https://www.baltimorepolice.org/crime-stats/open-data

  4. [12]

    The cost of crime to society: new crime-specific estimates for policy and program evaluation

    McCollister KE, French MT, Fang H. The cost of crime to society: new crime-specific estimates for policy and program evaluation. 2010 Apr 1; doi: 10.1016/j.drugalcdep.2009.12.002. Epub 2010 Jan 13. PMID: 20071107; PMCID: PMC2835847

    2010 doi

  5. [13]

    https://www.brennancenter.org/our-work/research-reports/predictive-policing-explained

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.