Area-Based Socioeconomic Characteristics of Industries at High Risk for Violence in the Workplace

Abstract

This study examined socioeconomic factors associated with the presence of workplaces belonging to industries reported to be at high risk for worker homicide. The proportion of 2004 North Carolina workplaces in high-risk industries was computed following spatial linkage of individual workplaces to 2000 United States Census Block Groups (n = 3,925). Thirty census-derived socioeconomic variables (selected a priori as potentially predictive of violence) were summarized using exploratory factor analysis into poverty/deprivation, human/economic capital, and transience/instability. Multinomial logistic regression models indicate associations between higher proportion of workplaces belonging to high-risk industries and Block Groups with more poverty/deprivation or transience/instability and less human/economic capital. The relationship between human/economic capital and Block Groups proportion of high-risk industry workplaces was modified by levels of transience/instability. Community characteristics therefore contribute to the potential for workplace violence, and future research should continue to understand the relationship between social context and workplace violence risk.

Appendices

Appendix 1

Spatial Data Reference and Linkage

Spatial boundary information for North Carolina census-defined BG were obtained from ESRI, originally founded as Environmental Systems Research Institute (ESRI Copyright 1999–2005). The spatial data were edited to exclude BG that were predominately water or where the 2000 United States (US) Census listed total population was zero (n = 10). LEA jurisdictional boundaries were obtained from individual agencies and compiled together into a common geographic information system (GIS) database. The geocoded workplaces were spatially linked to Census 2000 BG and North Carolina LEA jurisdictions using ArcGIS 9.1.

In the GIS, BG were used as the unit of analysis. Crime data, however, was measured at the level of the LEA. BG located completely within a given agency’s jurisdictional boundaries were assigned the crime data for the given agency. BG that crossed agency boundaries were assigned a weighted average of crime rates, where the weights were equal to the proportion of the BG area covered by each LEA jurisdiction.

Of the 314,344 geocoded workplaces, 73,054 (23.24%) could not be assigned to a LEA due to lack of spatial data. As a consequence the effective sample size was reduced to 241,290 workplaces where both LEA and census BG identifiers were available.

Appendix 2

Workplace Exclusions

InfoUSA described each workplace by up to three SIC codes: one primary and two secondary SICs, where each SIC was detailed at 6-digits. SIC codes represent a classification system used by the US government in the compilation of establishment-based economic statistics classified by industry. Industries are defined in relation to what is produced and the composition and structure of the economy (Executive Office of the President, Office of Management and Budget, 1987).

Workplaces which were categorized as non-classifiable establishments by InfoUSA for all three SIC codes were excluded (n = 3,601, 1.50%) from the analysis. Non-classifiable establishments (SIC = 9999) reflect those establishments which can not be classified into any other industry under the current version (1987) of the SIC manual. Additionally, establishments classified as Police/Correctional/Fire Protection (SIC: 9221, 9223, 9224) or Farming (SIC: 01, 02, 07) for any of the three SIC codes were also excluded, due to different etiologic exposures than those hypothesized to contribute to fixed-location workplaces. Following these exclusions (n = 10,506) from the effective sample of 241,290 workplaces, the remaining 230,034 workplaces were distributed across 81 of 100 counties in North Carolina.

There is no clear means of making inferences about community characteristics for workplaces that are mobile (e.g. taxicabs, trucking, delivery services, construction). Thus the analysis was restricted to fixed location workplaces (workplaces that did not have a 2-digit SIC in the construction or transportation industry divisions). This excluded a further 12% (n = 27,407) from the total 230,034 available workplaces. Following the exclusion of mobile workplaces, 202,627 fixed-location workplaces remained, representing 4,078 BG and 177 LEA. A small number of BG (n = 153, 4%) had three or fewer total workplaces and were excluded, leaving a total of 3,925 BG available for analysis. BG with three or fewer total workplaces were excluded; we considered the computed proportion of high-risk industry workplaces in the BG unreliably estimated if the total number of workplaces was less than three.