The Direct and Indirect Costs of Food-Safety Regulation

The Food Safety and Inspection Service (FSIS) promulgated the Pathogen Reduction Hazard Analysis Critical Control Program (PR/HACCP) rule in 1996 as its primary vehicle for regulating meat and poultry processing plants. The anticipated regulatory costs imposed on large and small plants were controversial then and remain contentious. At issue are the costs of sanitation and monitoring tasks, planning and reporting requirements, and testing mandates, and the relative costs of large and small plants.

Cost estimates of the proposed PR/HACCP rule prior to its promulgation sparked considerable controversy. FSIS's Federal Register announcement in 1996 projected costs of 0.12 cents per pound, but Knutson et al. had much higher estimates. Later, Antle, Nganje and Mazzocco, and Ollinger and Mueller used econometric analyses and various measures of food safety to estimate costs of 1.3, 0.04–43.5, and 0.9 cents per pound. The first direct cost estimates, based on post-PR/HACCP data, came from a national survey (Ollinger, Moore, and Chandran), a survey of Midwestern plants (Boland, Peterson-Hoffman, and Fox), and a survey of small and very small plants in Texas (Hooker, Nayga, and Siebert). These surveys indicated costs of 0.7, 0.9, and 2–20 cents per pound.

Most of the studies offered evidence that the PR/HACCP rule favors large plants over small ones. However, the econometric estimates were based on data taken before the rule was issued. Surveys taken after the regulation was mandated overcame this problem, but the subsequent analyses did not control indirect effects. For example, previous research has shown that regulations favor large plants and firms and more capital-intensive plants.

This paper evaluates the costs of the PR/HACCP rule econometrically. It differs from other papers in three critical ways. First, we use actual cost data that can be directly linked to several types of regulatory costs. Second, we examine both the direct and indirect effects of the PR/HACCP rule by extending a model of regulatory costs developed by Bartel and Thomas in which they examined regulations promulgated by the Occupation Safety Health Administration and other government agencies. Third, we show that private actions incur substantial food-safety costs. Direct effects come from the regulation itself (e.g., the costs of planning and implementing Standard Sanitation Operating Procedures (SSOPs) and HACCP). Indirect effects arise from the comparative advantage some plants have in meeting regulatory requirements (e.g., large plants may have lower per-unit regulatory costs because they can spread fixed regulatory costs over more volume). Private actions are food-safety practices taken and investments made by the plant that are not mandated by the regulation.

Our model also controls for wages and private actions. Private actions include (a) food-safety capital investments that plants make in order to satisfy the general demands of their customers, (b) standards imposed on a plant by fast-food restaurants and other major customers, and (c) internal organizational arrangements.

An important feature of this study is the use of three unique datasets. Data for one of those datasets come from a survey conducted by the Economic Research Service of the United States Department of Agriculture that garnered responses from 996 of the 1,720 plants considered manufacturers in administrative data files maintained by FSIS. This dataset has the costs of complying with the PR/HACCP rule and the various food-safety practices and technologies those plants use. Datasets from the FSIS provide information on food-safety monitoring practices and plant characteristics. Finally, Census data have detailed plant-level production data. All of these datasets are discussed in more detail in the data section.

Background

The Economic Framework for Analyzing the Sources of the Costs of the PR/HACCP Rule

We follow a regulatory cost model pioneered by Bartel and Thomas, who argue that regulatory costs have direct and indirect effects. Direct effects come from the regulation itself. For the PR/HACCP rule, these costs include performance of SSOP and HACCP tasks and compliance with the Salmonella and E coli standards. Indirect effects occur because plants have different characteristics that result in different regulatory effects. In addition to these regulatory costs, we control for wages and plant capital investments and costs due to private actions.

In equation (1) we express food-safety costs divided by sales (y) as a function of the cost of labor (W), private actions that include human and physical capital (K) and other market-driven private actions (M), and indirect (ID) and direct (D) regulatory effects. The analysis focuses on costs to domestic incumbents since imports are relatively small and we use cross-sectional data in which all plants have existed at least one year.

(1)

Ollinger and Moore show that private actions played a major role in Salmonella control in meat and poultry products. Private actions include investments in human capital and innovative food-safety technologies and practices. Managers make these investments to avoid costly recalls or other food-safety catastrophes, enhance their reputation with buyers, etc. (Ollinger and Mueller; Golan et al.).

Roberts gives several examples of food-safety-related technologies and Golan et al. discuss experiences of EXCEL and IBP, the two largest cattle slaughter firms, which purchased steam pasteurizers for all of their cattle processing plants, and Texas American Beef, which instituted a sophisticated pathogen-detection and control system.

Other market-driven private actions include explicit agreements between plants and large buyers, such as fast food restaurant chains, in which plant managers agree to undertake food-safety process control tasks and make specific investments in return for guaranteed markets, higher volume orders, higher prices, or some other benefit (Ollinger and Mueller; Golan et al.; Codron, Giraud-Heraud, and Soler; Codron, Fares, and Rouvière).

Private actions also occur when contractual arrangements with suppliers or upstream buyers are so burdensome and production processes are compatible enough that plant managers choose to vertically integrate (Williamson; Reimer). Vertical integration offers greater control over product quality since negotiations over contractual terms are eliminated, as one management controls the entire process.

Previous research points to four variables that regulation indirectly affects and should be included in the vector ID. Antle, Pashigian, Boland, Peterson-Hoffman, and Fox and Ollinger, Moore, and Chandran provide ample evidence that regulation favors larger plants. Large firms may also have lower costs (Pashigian; Bartel and Thomas). Additionally, Pashigian found that regulation favors more capital-intensive industries. Finally, the PR/HACCP rule requires meat and poultry manufacturers to establish and implement different HACCP plans for different HACCP products, suggesting that plants with less product specialization (i.e., more products) have higher costs and more specialized plants (fewer products) have lower costs. Hooker, Nayga, and Siebert offer some support for this specialization hypothesis.

There are other indirect regulatory effects. Pashigian provides evidence that regulation favors union workers because unionization tends to raise the costs of providing worker benefits. Bartel and Thomas found that regulation favors importers because domestic producers have to comply with regulatory costs. Finally, Moore's analysis indicates that regulation favors incumbents because regulation raises industry entry costs. We do not consider these factors because (a) unionization in the meat and poultry industry existed in a much weaker condition by 2000 than it did prior to 1985, (b) we are considering only domestic plants, and (c) all plants in our dataset have existed at least one year.

The direct regulatory costs of the PR/HACCP rule include doing SSOP and HACCP tasks and complying with performance standards. Compliance with performance standards requires plants to increase product testing or cleaning, purchase steam vacuum units or other food-safety process-control equipment, or use some other means to meet the standard.

Three Unique Datasets Provide the Data

Data are a matched dataset coming from a national survey conducted by the Economic Research Service in 2001 on the costs of the PR/HACCP rule and food-safety technology, the Enhanced Facilities Database (EFD) of FSIS for 2000, and the Longitudinal Research Database (LRD) from the Bureau of the Census.

The survey-garnered responses from 996 of the 1,720 plants considered manufacturers in the administrative data files maintained by FSIS. FSIS regulates more than 6,000 retail stores, restaurants, and manufacturing facilities that process meat or poultry. Plants may or may not ship product across state lines. We defined an establishment as a manufacturer if it slaughtered animals or was assigned to SIC 2011, 2013, or 2015 in the EFD and had either sales exceeding $7.0 millions per year or production greater than 1.0 million pounds of meat and poultry per year.

The ERS data include only plants from the EFD that responded to the survey, so they are not nationally representative and it may not be valid to generalize results. However, several reasons lead us to believe that the bias is small. First, the final dataset has a large number of plants, including 161 federally inspected cattle and hog slaughter plants, 66 federally inspected poultry slaughter plants, and 298 federally inspected cooked and raw meat processors with no slaughter operations. Second, the share of total output closely tracks the share of plants responding to the survey. Third, a regression analysis by the authors suggests that no correlation exists between plant size and survey response.

To account for remaining biases in the data, we treated it with a post-stratification adjustment (Gelman and Carlin). Under this approach, the regression is adjusted with a response weight equal to the reciprocal of the share of plants responding to the survey in its strata. The dataset was stratified by plant size.

The ERS data include approximately ten questions dealing strictly with costs and benefits of HACCP regulation, thirty-five on food-safety technologies and practices, and fifteen miscellaneous questions about plant and other characteristics. The questions about the costs of the PR/HACCP rule dealt with the number and types of workers hired, planning costs, nonlabor variable costs, and capital investments. The HACCP questions also asked subjective questions, such as the aspect of the PR/HACCP that was most costly. The questions are provided and the responses are tabulated and summarized at http://www.ers.usda.gov/data/haccpsurvey/.

The EFD contains observations for more than 9,000 manufacturing and other establishments monitored by FSIS and state food-safety agencies. These establishments include all meat and poultry manufacturing plants and other establishments that process meat or poultry as a minor business, for example some grocery stores. The EFD provides very little production data for plants monitored by state agencies but data for plants inspected by FSIS include counts of the number of slaughtered animals, estimated sales and employment, types of processing operations (e.g., animal carcasses or ready-to-eat products), and other establishment characteristics.

The LRD includes information on all manufacturing plants from the Survey of Manufacturers taken every five years by the Census Bureau. The most recent survey available for this study was in 2002. The LRD also has data on a subset of larger plants and a sampling of smaller plants for the inter-Census years. LRD data include value of shipments, number of workers, wages, end of period value of buildings, end of period value of machinery, etc.

The Empirical Model

The model specification follows from equation (1). Food-safety costs divided by sales (y) are regressed on independent (x) variables serving as proxies for wages, human and physical capital, other market-driven private actions, and indirect and direct regulatory effects,

(2)

We divided food-safety costs by plant sales because a plant's profit margin is based on costs per unit (pound or dollar of sales). If regulatory costs per unit rise, then profits drop. Large plants naturally have higher regulatory costs because they have more operations, but, since they also have higher sales over which to spread these costs, they may suffer a very small drop in profits per unit when regulatory costs rise. By contrast, a small plant may have low regulatory costs but realize high costs per unit since there is a lower sales base over which to spread costs.

Food-safety costs equal nonlabor variable costs of complying with the PR/HACCP rule plus HACCP and SSOP planning costs plus the labor costs associated with performing tasks mandated under PR/HACCP plus labor for costs needed to maintain new technologies. Nonlabor variable costs come from question 14 of the ERS survey. Planning costs are the number of days required to make HACCP plans (question 15 of the ERS survey) times the average annual wage from Census files divided by 270 days (the number of workdays in a year). Labor costs for performing mandated tasks and staffing new technologies equal the number of production and quality control workers hired to meet PR/HACCP requirements (question 7 of the ERS survey) times the average annual wage of meat and poultry slaughter and processing workers from Census files divided by 270.

The wage rate (x₁) is defined as the average state wage for meat and poultry production workers in the state in which the plant was located. Human capital (x₂) equals one for plants that had a formal food-safety process control system prior to implementation of the PR/HACCP rule in 1996 and zero otherwise. A pre-HACCP food-safety process-control system was considered formal if the plant (a) used schematics or flow diagrams to identify critical pathogen control points and (b) systematically reviewed plant operations prior to promulgation of the PR/HACCP rule. These data come from questions 16 and 17 of the ERS survey.

The physical capital variable (x₃) comes from Ollinger, Moore, and Chandran and is an index value of food-safety plant technology. It is a continuous variable between zero and one and is monotonic in that plants with higher index values use equipment that is more sophisticated, do more frequent cleaning, have superior worker training systems, and/or have other practices and technologies that are superior in controlling pathogens than plants with lower index values. Data come from thirty-five to forty questions on five types of food-safety technologies given in the ERS survey. The five technologies are: sanitation, operations, food-safety processing equipment, plant capital investments, and hide removal technologies.

Other market-driven private actions include contractual agreements between plants and their customers and vertical integration. Contractual agreements with customers (x₄) equal one for plants that have customers that specify requirements that are more stringent than those FSIS demanded and zero otherwise (question 44 of the ERS meat survey).

Vertical integration (x₅) occurs in slaughter plants that also process meat into raw or cooked products. We account for it in our model with a dummy variable equal to one for slaughter plants that also produce ground meat, fabricated cuts, or other raw or processed meat and zero otherwise. Cooked and raw meat processors, as we have defined them, are not vertically integrated because, by definition, these plants are strictly processors without slaughter operations. Nevertheless, we include a dummy variable for processing industries to control for the cost differences between raw meat processors that produce cooked meat products and vice versa. For these plants, the dummy variable equals one for cooked meat processors that also produce raw products and vice versa for raw meat processors.

The indirect regulatory variables include plant size (x₆), firm size (x₇), the capital to labor ratio (x₈), and the specialization ratio (x₉). Plant size is the number of plant employees. Firm size equals one for plants owned by firms that own more than one establishment and zero otherwise. The capital to labor ratio equals the ratio of the plant's value of buildings and equipment at the end of the period divided by the plant's total employment. Plant specialization is the Census primary product specialization ratio.

Census defines the specialization ratio as the sum of a plant's value of shipments from its primary products divided by its total value of shipments. A slaughter plant's primary products include all raw meat production from one slaughter species (cattle, hogs or chickens) and corresponds to at most three HACCP product categories requiring separate plans: carcasses, fabricated cuts, and ground meat. We do not use the specialization ratio for cooked and raw meat processors because Census product categories differ substantially from HACCP categories. We could have used the number of Census products instead of the specialization ratio but there is little correspondence between processed Census products and processed HACCP products, making this a poor measure of planning requirements.

There are three direct regulation variables. Tasks per employee (x₁₀) equal the number of SSOPs and HACCP tasks performed in 2001 in order to comply with the PR/HACCP rule divided by the total number of employees. More tasks should mean higher costs.

We also distinguish between the costs of SSOP and HACCP tasks by including a variable defined as the number of HACCP tasks divided by the combined total of SSOP and HACCP tasks required to comply with SSOP and HACCP plans (x₁₁). If HACCP tasks are more costly than SSOPs, then this variable should be positive.

The PR/HACCP rule requires plants to meet tolerances for Salmonella, generic E coli, and fecal matter. Plant managers that cannot meet these performance standards with their existing food-safety process controls or fear that they might not meet them in the future must hire production workers to operate food-safety equipment, such as steam vacuum units, staff new operating practices, or do more intensive cleaning. Plants that are able to meet performance standards do not have to hire any production workers. All plants had to hire quality control personnel to perform SSOP and HACCP tasks. Thus, the ratio of production workers hired to comply with the PR/HACCP rule divided by all production and quality workers hired in response to the PR/HACCP rule (x₁₂) should correspond to the regulatory effort required to comply with the Salmonella and generic E coli performance standards and should be positively related to the costs of the PR/HACCP rule.

Endogenous Tasks Per Employee and the Econometric Approach

We evaluated all variables of equation (2) at their sample means and adjusted the model for heteroskedasticity with White's estimator. We also tested the right-hand side variables for endogeneity with Hausman tests. Since we could not reject endogeneity for tasks per employee, we used an instrumental variable for that variable. Hausman tests rejected endogeneity for all other right-hand side variables.

The method of instrumental variables is a two-stage procedure in which an instrumental variable is determined in the first stage and used in the second as an independent variable. To estimate an instrumental variable, we needed an instrument that is correlated with the endogenous variable—tasks per employee in our case—but not the error term. We used economic tasks performed per employee as that instrument for the cattle, hog, and chicken slaughter and raw meat processing regressions and lagged tasks per employee as an instrument for cooked meat. Economic tasks are actions mandated by FSIS and undertaken by a plant to ensure that non-food-safety quality attributes, such as product fat and water content, adhere to FSIS standards. Data come from FSIS files.

In the first-stage regression, we included the instrument for tasks per employee and all of the exogenous variables of equation (2) as right-hand side variables and made an estimate of tasks per employee. Since a Hausman test of this instrumental variable rejected endogeneity, we replaced tasks per employee in equation (2) with its estimate (the instrumental variable) and proceeded with the analysis. Appendix Table A1 contains the results of the model predicting tasks per employee. All variables have been defined.

Results

Results are given in Table 1. The R² statistics vary from 0.18 to 0.57. Estimated costs, when all dummy variables and all other variables regardless of whether they are significant are set equal to their mean values, equal $0.022, $0.020, $0.0001, $0.029, and $0.016 per dollar of sales for cattle, hog, and chicken slaughter and cooked and raw meat processing, respectively. These estimates are within one standard deviation of their mean values (table 2). If only significant terms are used, the estimates are $0.021, $0.016, −$0.0041, $0.018, and $0.004 per dollar of sales for the same industries. All of these, except for chicken slaughter, are within one standard deviation of their mean values.

Consider Table 1 and the individual reported results. Results show that state wages are negatively associated with costs in four of five cases. A positive sign means that the productivity of workers in high-wage states equals productivity in low-wage states, resulting in relatively higher costs for plants in high-wage states. A negative sign means that the higher wages offered in high-wage states paid for more productive workers that enabled plants in those high-wage states to hire fewer workers and incur lower costs than plants in low-wage states.

Private actions include previous experience (human capital) and the food-safety technology index (physical capital). Previous experience with a food-safety process-control system prior to enactment of the PR/HACCP rule should confer an advantage on plants with existing formal food-safety process-control systems. Results show that plants with food-safety process-control systems prior to the PR/HACCP rule had significantly lower costs in two of the five industries. Two other cases were positive and one was negative and insignificant. Results also show that plants with higher food-safety technology index ratings had consistently higher food-safety costs. Recall that a high technology index results from more intensive cleaning, the use of advanced food-safety equipment, such as steam vacuum units, and more intense use of other food-safety practices. These activities enhance food-safety process-control, but they also raise costs.

By making marginal changes to the independent variables, we can examine their impact on food-safety costs. As reported in Table 3, a 10% increase in the food-safety technology index raises food-safety costs by 16.9–20.6% for the three industries with significant changes. Note, since the regressions variables are at sample means, 10% changes for continuous variables are 10% changes of their coefficient values. For dummy variables, we used the sample mean for the value of the dummy variable and multiplied by 10%.

Other market-driven private actions include customer standards and vertical integration. Results indicate that customers imposing their own standards had a substantial impact on food-safety process-control costs. Coefficients were significant and positive in three of five cases, and a 10% increase in customers imposing their own standards significantly raises food-safety costs from 3.4 to 6.8% in the cattle, hog, and chicken slaughter industries (Table 3).

Vertical integration also affected plants costs. Its coefficient was significant and positive for cattle and significant and negative for chicken slaughter. Signs vary because, on the one hand, vertical integration reduces costs by enhancing control over quality (Williamson) while, on the other hand, vertical integration raises costs by increasing the number of HACCP-related processes, resulting in more HACCP-related planning and monitoring costs.

Previous research has indicated that that regulation indirectly affects costs. Large, more capital-intensive plants and firms should have lower costs relative to smaller, more labor-intensive plants and firms. Additionally, the PR/HACCP rule requires plants to develop and implement separate HACCP plans for separate HACCP processes, suggesting more specialized plants should incur lower costs. Empirically, plant size, plants owned by firms owning more than one plant, capital intensity, and the specialization ratio should negatively affect costs. Results show that plant size (the number of employees) is negative in all cases and significant in four of them; the variable for plants owned by firms owning more than one plant is negative in four cases and significantly so in one of those; the capital-to-labor ratio is negative in four cases and significant in two of those; the specialization ratio is negative in all three of its cases and significant in two of those. These results suggest substantial indirect effects. Only two of the possible eighteen regression coefficients were of a sign that differed from what was expected.

The direct regulatory effects are tasks per employee, and the ratio of new production workers to new production and nonproduction workers. As expected, all coefficients are positive and eight of the ten are also significant. Positive signs mean that food-safety costs rise with more PR/HACCP requirements.

Direct effects are particularly important because a regulator can control them. Changes in the number of HACCP and SSOP tasks per employee imply changes in the monitoring and cleaning tasks required to comply with the PR/HACCP rule. A 10% increase in the number of tasks per employee significantly increases costs by 3.7–11.3% for the four industries with significant effects (Table 3).

Remember that the ratio of newly hired production workers in response to the PR/HACCP rule to the combined total of newly hired production workers and quality control workers reflects changes in the food-safety process control system necessary to meet performance standards. Results show that a 10% change in this production worker ratio raises costs by about 1.5–3.4% in the four industries with significant effects (Table 3).

Discussion

Table 4 contains the contributions to costs made by direct and indirect effects of private actions and the PR/HACCP rule. All values are at their sample means. We focus our discussion on the significant variables associated with physical capital, customer standards, and the indirect and direct regulatory costs. We ignored the other variables because they are either exogenous to the plant (state wages), cannot be changed (whether a plant had experience with quality control programs prior to the PR/HACCP), or had ambiguous outcomes (processes and slaughter or processes raw and cooked meat). Parameter values of statistically insignificant parameters that had the expected sign are included in the table but not included in the individual subtotals. Significant parameters are distinguished from insignificant ones by the use of asterisks.

Physical capital investments and customer standards are private actions taken by a plant in order to (a) meet general consumer demands for their products and (b) comply with specific standards imposed on them by large customers, such as fast food chains. As shown in Table 4, the sum of parameter values of private actions exceeds direct regulatory effects in all industries, except cooked meat processing, at sample mean values. The food-safety technology index is the most costly of all private actions in all industries except chicken slaughter.

A central theme among economists is that the indirect effects of regulation, particularly plant size, are important. We found that large plants do have a cost advantage over smaller ones. A 10% change in plant size gives only a small cost advantage (economies of scale) to larger plants. However, there are huge differences in plant sizes in all of the meat and poultry industries. For example, there are many plants in all industries with more than 500 employees. This size plant is about 3.5, 2.5, 4, and 4.5 times larger than the mean size cattle and hog slaughter and cooked and raw meat processing plants reported in Table 2. In cattle slaughter, just a 100% increase in plant size (i.e., one times the coefficient in Table 1) leads to an 11% reduction in food-safety costs. One hundred percent increase in the number of employees at hog and chicken slaughter and raw meat processing plants leads to changes in costs ranging from reductions of 11% in hog slaughter to 40% in chicken slaughter.

These cost differences are quite large and are consistent with findings of Antle, Boland, Peterson-Hoffman, and Fox, Hooker, Nayga, and Siebert, and Ollinger, Moore, and Chandran. However, it is difficult to see how these differences affect survival. Large plants already enjoy substantial economies of scale, yet small plants persist by producing niche products and avoiding direct competition with their large competitors (MacDonald, Ollinger, Handy, and Nelson). Thus, the actual disproportionate impact on survival of the PR/HAACP on the survival of small plants relative to large ones may be quite small.

Now compare direct and indirect regulatory costs. We evaluate direct regulatory variables at their mean values and indirect regulatory costs due to plant size at the mean and 100% of their mean size. Table 4 shows that direct effects exceed indirect effects at sample mean values in the hog slaughter and cooked and raw meat processing industries. At 100% sample mean sizes, indirect effects exceed direct ones in all of the slaughter industries but none of the processing industries.

Table 4 also shows that the instrumental variable for tasks per employee accounted for more than 60% of all direct regulatory costs in cattle and hog slaughter and cooked and raw meat processing. The instrumental variable was not significant in the chicken slaughter industry.

Notice that the costs of complying with performance standards—the ratio of production workers hired to comply with the PR/HACCP rule to the total of production workers and quality control workers hired in response to PR/HACCP—contribute about 16%, 36%, 39%, and 23% of the direct regulatory effects in cattle and hog slaughter and raw and cooked meat. This means that performance standards could be made about six times more stringent (in terms of costs) than currently exist in cattle slaughter to result in the same cost to the plant as tasks per employee. In a similar manner, hog slaughter performance standards could be made 2.8 times and cooked meat and raw meat processing could be made 2.6 and 4.3 times more stringent than that which currently exists to result in the same cost to the plant as tasks per employee.

Now suppose that the generic E coli and Salmonella performance standards are as effective at controlling pathogens as are SSOP and HACCP tasks. This would mean that the performance standards provide food-safety process-control at 50% or less of the cost of the SSOPs and HACCP tasks, suggesting the most efficient way to increase regulatory oversight would be to increase the stringency of the performance standards.

Conclusion

This paper empirically examines the impacts of indirect and direct regulatory effects and private actions on the reported costs of complying with the PR/HACCP rule. One indirect effect is plant size. Direct effects include HACCP and SSOP tasks per employee and a measure of the effort devoted to performance standards.

Cost estimates, based on the model, varied from $0.0001 in chicken slaughter to $0.0180 in cattle slaughter at sample mean values. All estimates fell within one standard deviation of their mean costs per pound and were consistent with Antle, Nganje, and Mazzocco, and Ollinger and Mueller, who used pre-regulatory data and translog cost functions to estimate costs of $0.013, 0.0004–0.435, and 0.009 per pound of meat. They were also consistent with survey-based estimates of $0.007, $0.009, and $0.02 to 0.20 cents per pound of meat from Ollinger, Moore, and Chandran, Boland, Peterson-Hoffman, and Fox, and Hooker, Nayga, and Siebert.

Results suggest that indirect and direct regulatory effects and private actions significantly affected food-safety costs. Some of the more notable findings are: (a) economies of scale in food-safety process control give the very largest plants a substantial cost advantage over their smaller competitors, (b) an instrumental variable for HACCP and SSOP tasks per employee imposed the greatest indirect cost in four industries, (c) the costs of complying with the generic E. coli and Salmonella performance standards was less than one-half the costs of performing SSOP and HACCP tasks in cattle and hog slaughter and raw and cooked meat. The third finding means that if performance standards and process control tasks (SSOP and HACCP tasks) currently provide equal amounts of safety and if FSIS regulators wanted to enhance food-safety process control, then the same benefits at less than 40% the costs could be realized by raising the stringency of performance standards rather than the number of process control tasks.

Private actions (buyer contracts and plant food-safety capital investments) accounted for more than half of food-safety costs. There are two likely reasons why. First, plants have one food-safety process-control system and they may not distinguish between the costs of complying with the PR/HACCP rule and private actions, causing them to report all food-safety process-control cost as a cost of PR/HACCP. Second, if a plant or buyer makes superior food-safety process control a strategic goal, then the plant or a buyer would have to stiffen its stringency requirements whenever the benchmark quality increases. Thus, a rise in regulatory requirements results in a rise in the stringency (and cost) of private actions.

It is important to note that this paper attributes 64–100% of food-safety costs in the cattle, hog, and chicken slaughter and raw meat processing industries to private actions and Ollinger and Moore recently found that more than four-fifths of Salmonella control is due to private actions, suggesting that the cost of providing food-safety process control per unit of Salmonella control is about the same for private actions and regulation.

This study did not account for the effect of food-safety regulations on plant productivity. Yet, Antle points out that economists starting with Christianson and Haveman in the 1980s have recognized that regulations can reduce productivity. He adds that Klein and Brester found that line speed reductions, due to regulations at one plant, raised production costs by 7%.

The ERS survey offers some evidence that yields may have dropped. Question 10 of the ERS survey asks how the PR/HACCP rule affected finished product yields. About 25% of respondents indicated that the regulation reduced yields while 73% said there was no change and 2% said there was an increase. However, the ERS survey also suggests that product shelf life may have increased. Question 18 of the ERS survey asks how shelf life changed after 1996 (the year in which PR/HACCP was promulgated). About 62% indicated that shelf life did not change, 27% said that it increased more than one day, and only about 1% said that it decreased. Combined, these data suggest that PR/HACCP may have reduced yields but it may also have increased shelf life, leaving an ambiguous effect on the net value of output.

Limitations and Further Research

The main limitation of this study is that the ERS survey and the final dataset were not nationally representative, meaning that results cannot theoretically be generalized. The bias may have been minimal, however, because (a) the share of total output of respondents closely tracks the share of survey participants, (b) a regression showed no correlation between plant size and survey response, and (c) data were treated with a post-stratification adjustment (Gelman and Carlin).