Environmental Regulation and Relative Food Cost

EPA Regulation and Relative Food Cost

Copyright 2017, Levi A. Russell

Abstract

Cost-benefit analysis of agri-environmental regulation is limited in the sense that it only examines the effects a single regulation will have on the public and polluters. Further, important mechanisms through which the public might bear part of the cost of regulation are not examined. This paper uses new data that allows for examination of regulation by a specific government agency on a specific industry to determine the extent to which relative food costs are affected by regulation of agriculture by the Environmental Protection Agency (EPA). The index allows for an examination of the overall effect of regulation, which is an important addition to the existing literature. Findings indicate that the costs of EPA regulation have not been borne solely by producers and that the costs associated with EPA regulation have offset the food-cost-reduction effect of increased agricultural productivity from 1976 to 2010. Though this is not a comprehensive study of the net benefits of regulation, this paper adds to the literature on calculating the cost of regulation.

Keywords: food cost, environmental regulation, agricultural regulation

JEL codes: L51, Q11, Q58 

1. Introduction

Environmental regulation is designed to benefit the public by internalizing the external environmental costs associated with agricultural production. There is considerable academic literature and analysis by regulators measuring the benefits to the public and the costs to agricultural producers of specific, incremental regulations. However, there is comparatively little analysis examining the potentially negative effects of environmental regulation, as a whole, on consumers. This paper makes use of newly-available data on regulation to determine the extent to which there is an association between the share of income spent on food and EPA agri-environmental regulation. It should be noted at the outset that these results are not indicative of a net benefit or cost to society of environmental regulation; such an analysis would require a general equilibrium model. However, the results of general equilibrium analysis would likely be driven by estimates of gross domestic product and other macroeconomic variables. This paper examines the potential effects of environmental regulation on relative food cost given the existing macroeconomic conditions.

 Though most cost-benefit analysis of environmental regulation is focused on the costs to producers and benefits to consumers, few attempts examine the cost of environmental regulation on consumers. One notable exception is the considerable literature on the ban of battery cage egg production in California, particularly a study by Malone and Lusk (2016). Similar regulation intended to increase the quality of treatment to hens but resulted in a negative impact on consumers. Initially, the ban affected over 90% of the egg industry in California, since less than 10% used alternative methods to battery cages. As the ban in California broadened to include all domestic products, the supply of eggs in California became heavily restricted. Laws which were intended to benefit animal welfare had a negative result to consumers by putting upward pressure on egg prices.  

In a similar way, environmental regulation is designed to internalize external costs associated with the production of agricultural products. The increase in producer costs due to regulatory compliance can be modeled as a leftward shift of the supply curve, resulting in higher food prices. This higher food cost offsets, to some extent, the benefit the general public receives from improved environmental conditions due to changes in agricultural production practices required by environmental regulation.

Examination of the overall effects of other types of regulation leads to similar conclusions. Analysis by Davies (2014) on regulation in general and its effect on productivity across a range of industries indicates that more-regulated industries have lower output per hour, lower output per worker, and higher unit labor costs than less-regulated industries. Taken together, these studies indicate that it is reasonable to expect that compliance and other costs of regulation are not fully mitigated by the reduction in cost due to innovation. Given that these regulations shift supply curves in their respective markets to the left, a portion of the costs are borne by consumers. This research seeks to provide insight into this empirical question in the case of environmental regulation and food cost.

        The costs of regulation are theoretically expected to be borne by consumers is through higher food prices. Recent reviews of the agri-environmental regulation literature by Gurtoo and Antony (2007) and Vercammen (2011) do not include studies examining this potential effect. However, in their study of consumer prices and overall federal regulation, Chambers and Collins, and Krause (2017) find that consumer prices associated with industries that are relatively more regulated are higher and more variable than prices associated with less regulated industries. Further, they find that lower-income households spend a larger share of their income on consumer goods from relatively more regulated industries, which indicates a regressive distributional impact of regulation. The findings of these studies necessitate an examination of agri-environmental regulation’s effects on food cost.

        The objective of this paper is to examine the impacts of agri-environmental regulation by the Environmental Protection Agency (EPA) on relative food costs. The EPA oversees enforcement of 12 major regulatory acts directly impacting agricultural producers, including the Clean Air Act; the Clean Water Act; the Federal Insecticide, Fungicide, and Rodenticide Act; and the Food Quality Protection Act. Regulations impact the day-to-day operations of a farm because they restrict the range of technologies available to farmers to deal with pests and soil fertility issues. Furthermore, these regulations affect producers from the farm up the supply chain to the processor. Specifically, the costs of crop production and storage, animal production and feeding operations, dairies, greenhouses, and other agricultural operations are all affected by agri-environmental regulations. 

Though the EPA is required to conduct a cost-benefit analysis of all significant agri-environmental regulations written and enforced by the EPA, the analysis only examines a single regulatory effort in isolation. This is inadequate, since each new regulation likely interacts differently with many existing regulations. Thus, an examination of the effects of the regulations as a whole should be an integral part of the body of work evaluating regulatory effects. In an examination of cost-benefit analyses, Dudley (2013) notes that regulatory cost-benefit analysis is often incomplete. Specifically, Dudley states, 

On the benefit side of the equation, they quantify or list every conceivable good thing that they can attribute to a decision to issue new regulations, while on the cost side they only consider the most obvious direct and intended costs of complying with the regulation. (Dudley, 2013, pp 30)

This paper contributes to the regulation literature by examining an unintended consequence of environmental regulation of agricultural firms: specifically, the cost imposed on the public through a higher relative food cost. We make use of a new dataset called RegData created by Al-Ubaydli and McLaughlin (2015). This database is generated by counting the number of restrictive words such as “must,” “shall,” or “required” in the Code of Federal Regulations (CFR). The word counts are transformed into regulatory indices that measure regulation by agency by industry. This is done by adjusting the regulatory restriction count by the probability that a given part in the CFR applies to an industry (in the present paper, we consider only those regulations written by the EPA and enforced on agriculture). The probability is calculated by dividing the number of mentions of the NAICS industry code of interest (NAICS code 11, in this case) by the total number of words in that part of the CFR.

We proceed as follows: section 2 discusses the identification of the empirical model, section 3 describes the data, section 4 details the results of specification tests, section 5 provides a discussion of the results, including a detailed discussion of the problem of omitted variables bias, and section 6 concludes.

2. Model

Environmental regulation of agriculture is designed to ensure that external environmental costs associated with agricultural production are internalized by producers, which promotes the public interest (Pasour and Rucker, 2005, pp. 49). However, the leftward shift of the supply curve associated with the regulations also results in higher food prices.

Omitted variables bias is a significant concern which can be addressed in part by leaning on basic economic theory. First, agricultural productivity growth is a key driver of relative food cost. Improvements in agricultural productivity reduce the cost of food over time since fewer inputs are required to produce more agricultural output. Second, food cost is also affected by the intermediate elements of the food supply chain. Processing, distribution, and retailing make up the value-added components of the food supply chain. The farm-to-retail price spread is used to account for the contributions of the food supply chain to relative food cost. Finally, per-capita disposable income likely affects relative food cost. As consumers are increasingly able to purchase a wider array of products, food purchases as a share of total income are likely to fall. Failure to adequately account for these factors would result in biased estimates of regulatory impacts. Supply and demand factors are not simply ad hoc additions to the conceptual model; they are fundamental factors determining food prices and quantities, and thus relative food cost.

To measure the effects of EPA agri-environmental regulation on relative food cost, we use data from 1976 to 2010 and specify a time series model. As Miller and Coble (2007) note, the ratio of food expenditures to personal disposable income is the predominant measure of relative food cost. To identify the effect of agri-environmental regulation on relative food cost, we follow Miller and Coble in the specification of the following model:

FESt=α+1EPAt+2APt+3FRPt+4PCINCt+et                            (1)

where FES is the ratio of food expenditure to personal disposable income in year t, AP is agricultural total factor productivity, FRP is the farm-to-retail price spread, PCINC is per-capita disposable income (which is not the same as the denominator in the dependent variable), and EPA is an annual index of EPA regulation of agriculture from 1976 to 2010. The error term, et, is defined below in section 4.

3. Data

Data on both components of the dependent variable from 1976 to 2010 are taken from the USDA Economic Research Service. The food expenditure data only measures expenditures on food purchased from grocery stores or other food retailers and food grown and prepared on farms. One limitation of the study is that food purchased from restaurants is not included in the food expenditure data. This results in an understatement of percentage of income spent on food. However, adding expenditures on food purchased from restaurants to the data would not necessarily result in less bias overall. The retail price of prepared food served or delivered by a restaurant is not the same as the underlying wholesale price paid for the food due to the cost of preparation and service or delivery. EPA regulations likely only have ancillary effects on food preparation and service or delivery (e.g. regulations related to disposal of waste), so it is unlikely that the exclusion of expenditures on food away from home bias our results more than their inclusion would.

Although millennials are more likely to eat out they also place a higher premium on quality food than other age demographics. Compared to the youth of 1980 they spend significantly more of their budgets on meats and fruits than teens in other generations or even older generations now. While the youth of today admittedly do spend more eating out they also spend the most of their total food income on beef, pork, poultry eggs, and fresh fruit (Conley and Lusk, 2018). 

In addition to market income, the personal disposable income data includes government transfer payments for the purposes of food purchases. As Figure 1 indicates, this ratio has fallen from 16% in 1976 to 11% in 2010. As our results will show, the decline in this ratio represents a decrease in relative food costs over time as improvements in agricultural productivity and increases in incomes have outstripped additional costs associated with value-added components of food and EPA regulation. Summary statistics data on all variables in the model can be found in Table 1.

        The USDA Economic Research Service is also the source of the supply-side data employed in the model. Total factor productivity (Figure 2) is calculated as an index of outputs relative to inputs. Agricultural productivity has increased 78% from 1976 to 2010 primarily due to technological improvements in purchased inputs, improved yields, and the development and improvement of labor-saving devices. The farm-to-retail price spread is an index of value added to food products based on a market basket of goods produced in the 1982-1984 time period. The retail cost of the basket is compared to the price received by farmers for the agricultural commodities that correspond to this basket.

        Per-capita disposable income data (Figure 3) are taken from the U.S. Bureau of Economic Analysis. Incomes have risen from about $18,600 per year in 1976 to over $37,000 before the Great Recession and falling to roughly $35,000 in 2010. This variable is adjusted for inflation and is expressed in chained 2009 dollars.

        The EPA regulation index (Figure 4) is taken from the RegData database created by Al-Ubaydli and McLaughlin (2015). The data represent the level of regulation of agriculture by the EPA from 1976 to 2010. Regulatory restrictions data have two significant advantages over other measures such as page counts in the Federal Register or annual regulator budgets. First, since the regulatory index is based on the number of restrictions, there is likely to be less noise from other factors such as changes in bureaucratic costs or recession (a problem for spending measures) or changes in extraneous legislative language (a problem for page count measures). Second, the ability to determine the level of regulation on a specific industry by a given agency reduces the potential for noise from changes in that agency’s regulation on other industries. This index has been used to determine the effect of regulation on entrepreneurship and employment by Bailey and Thomas (2017).  

4. Specification Tests

We conduct a range of specification tests on the model in Equation 1 to ensure that serial correlation, heteroskedasticity, and nonstationarity do not influence the results. We first conduct unit root tests to determine whether the data are stationary. We then test for heteroskedasticity. Finally, we test for serial correlation and use information criteria to select a model and specify the error process for the empirical model described in Equation 1. For interpretive convenience, all variables are transformed using the natural log function. This allows us to interpret the coefficients in Equation 1 as elasticities. Robustness checks confirm the stability of our findings and are discussed in the results section.

        Heteroskedasticity, or non-constant variance of the error term, can lead to incorrect estimates of standard errors. To determine whether heteroskedasticity is present, we employ the Breusch-Pagan test. For Equation 1 in levels, we fail to reject the null hypothesis of constant variance (p-value 0.825). Estimating equation 1 in first differences yields a p-value of 0.688, meaning that we fail to reject the null hypothesis that the error term has constant variance.

Next, we test for nonstationarity in the data series. The Dickey-Fuller test indicates that the dependent variable is not stationary in levels (p-value 0.212) and is not trend-stationary (p-value 0.994). However, the test indicates that the first-differenced dependent variable is stationary (p-value 0.001). All independent variables are also non-stationary in levels with the exception of agricultural total factor productivity, which is trend stationary. Thus, the model is specified in differences for all variables except total factor productivity and a time trend is added to the model. It is important to note that the differences in stationarity of the models preclude the use of the more complex vector error correction model.

Finally, we test for serial correlation using the Cumby-Huizinga (1992) test. This test is similar to the more common Breusch-Godfrey test in that it tests for serial correlation along a range of orders. However, the Cumby-Huizinga test also allows for testing of serial correlation of each order individually. This is important because the Breusch-Godfrey test cannot directly determine the order of serial correlation. A Breusch-Godfrey serial correlation test may overstate the order of serial correlation because it is possible to reject the null hypothesis in the test of, say, serial correlation of orders 1 through 4 simply because the autocorrelation of the first order is very strong. The Cumby-Huizinga test yields p-values of 0.019, 0.055, 0.044, and 0.076 for serial correlation ranges of 1-1, 1-2, 1-3, and 1-4, respectively. However, the test of specified orders of serial correlation yields p-values of 0.019, 0.7475, 0.184, an 0.669 for orders of correlation 1, 2, 3, and 4, respectively. Thus, the tests conclude that there is serial correlation of the first order.

The tests specified above indicate that 1) We do not find evidence of heteroskedasticity, 2) the error term is stationary, and 3) there is evidence of serial correlation such that the model should be specified with at least 1 autoregressive term.

The results of a series of ARIMA regressions can be found in Table 2. We report both the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The problem of selecting the number of autoregressive and moving average terms is not symmetric. That is, selecting too few lags can result in biased parameter estimates while selecting too many lags merely reduces efficiency. Thus, we report the results of a range of specifications to increase confidence in the estimate of the food cost effects of EPA regulation of agriculture. Further discussion of alternative specifications is provided below, but it is necessary at this point to specify the error term in Equation 1. Given that both AIC and BIC are minimized with the ARMA(1,0) model, we specify the error term as follows:

et=ρet-1+t                                                                             (2)

where ρ is the autocorrelation parameter and t~N(0,2).

5. Results

We examine the extent to which there is an association between EPA regulation and relative food costs controlling for supply and demand factors. A positive and statistically significant association would be indicative of an unintended consequence of agri-environmental regulation on consumers. The magnitude of the effect is compared to the magnitudes of the effects of supply and demand factors to provide an interpretation of the relative importance of EPA regulation in the determination of the relative cost of food. 

To increase confidence in the robustness of our estimates, we estimate Equation 1 using log-log, linear-log, and linear-linear specifications. We also estimate regressions with additional AR and MA terms. In these other models, the coefficient on EPA regulation (1) is statistically significant at the 5% level across specifications.

Further, the magnitude, direction, and statistical significance of the regulation variable are robust to reductions in the sample size as well. The EPA regulation index increases dramatically from 2009 to 2010 (Figure 4) and we want to ensure that this single data point is not driving our results. Thus, we ran the models specified in Table 2 with data from 1976 to 2008 and found no statistically or economically significant differences in the coefficients.

The model with the lowest AIC and BIC values in Table 2 are in bold. The signs on the coefficients are in line with basic economic theory and are all statistically significant in the log-log specification reported in Table 2. Per-capita disposable income has the largest effect on relative food costs, indicating that increases in incomes over time have been the most important factor in expanding the food purchasing power of consumers. The farm-to-retail price spread is the next largest elasticity, indicating that the food supply chain has added significant value for consumers over the 1976 to 2010 time frame.

The elasticity of food cost with respect to agri-environmental regulation, 0.09, is remarkably stable across model specifications. A useful way to interpret this elasticity is to compare it to others in the model. The results indicate that EPA regulation and agricultural productivity have opposite effects on relative food cost and the magnitudes of the two effects are economically similar and are not statistically different from each other at the 10% level. The results suggest that EPA regulation offsets the oft-discussed food-cost-reduction effects of increased agricultural productivity over the 1976 to 2010 time period. 

More specifically, EPA regulation of agriculture increased 314% from 1976 to 2010. Assuming that the estimated association between EPA agri-environmental regulation and relative food cost is causal, we find that relative food cost would be 3.1 percentage points lower (7.9% rather than 11%) if EPA agri-environmental regulation had remained at 1976 levels. While this may seem like an extreme claim, similarly dramatic results have been found in the macroeconomic literature, such as Coffey, McLaughlin, and Peretto (2016); Dawson and Seater (2013); and Loayza, Oviedo, and Sevren (2004). Still, our result should be interpreted with caution since we use a reduced form model rather than a structural general equilibrium model. It is also possible that other regulations intended to benefit consumers may be correlated with environmental regulation of agriculture and thus constitute some of the cost spillovers measured in this paper.

This result does not imply a normative claim about EPA regulation itself. We do not claim that this result implies that EPA regulation should have remained at 1976 levels. Our normative claim is that cost-benefit analysis of regulation should consider 1) negative consequences of regulation on the public via higher product costs and 2) the empirical examination of the overall effect of regulation.

While standard welfare analysis is important in the determination of the incidence of regulatory costs, data that identifies regulatory restrictions at the industry level by regulator such as the RegData database used in this paper is crucial for understanding the real-world impacts of regulation. This dataset allows economists to examine the impacts of regulation without resorting to cross-country analysis of regulation using dummy variables for the existence of various policies that may or may not be comparable. Our paper is one of several recent efforts to examine the effects of regulation with this level of specificity.

5.1 Omitted Variables Bias

This study makes use of observational data to determine the effect of EPA regulation on relative food cost. To show the effects of omitted variables bias on the effect of interest, we estimate a range of more parsimonious alternative specifications by omitting or including the supply and demand variables (Table 3). 

The most parsimonious model is a regression of the food expenditure share on only EPA regulation. The model chosen by the selection criteria defined in sections 4 and 5 is reproduced in the far-right column. Moving from the far left to the far-right columns of Table 3, there is relatively little change in the standard errors of the coefficient of interest compared with the changes in the coefficient estimate itself. None of the estimates are negative, but the effect is only large enough to attain conventional levels of statistical significance when at least two of the control variables are included. Thus, the addition of the control variables identified above does not increase precision (i.e. a decrease in the standard error), but reduces the bias associated with omitting variables that, according to the theoretical discussion presented in section 5.0, should not be omitted. We include this discussion of the results of Table 3 to explicitly show the omitted variables bias and the effect it has on our analysis so that the reader can be more confident in our model specification and results.

6. Conclusion

Cost-benefit analysis of regulation is an important application of economic theory. As this paper demonstrates, it is also important to consider the possible negative consequences of regulation and the overall effects of regulation. While regulatory cost-benefit analysis is conducted on each new EPA regulation, these analyses are conducted in isolation and do not take into account the overall effects of regulation.

Our findings indicate that there is an economically and statistically significant positive association between EPA regulation of agriculture and the relative cost of food. This provides evidence of a causal relationship between the two, implying that upward pressure on food cost by EPA regulation is a real unintended consequence of said regulation. To guarantee our results are causal, we would need to observe an otherwise-identical U.S. food economy without EPA regulation. Since this is not possible, we make use of available data for variables that, according to economic theory, determine relative food cost. The robustness of our results and soundness and simplicity of the theory employed suggest that the association between EPA regulation and relative food cost is indicative of a causal relationship.

Given the existence of industry- and regulator-specific data from the RegData database, future work should examine the overall effects of regulation in other key industry-regulator pairs. Some potentially interesting effects to examine are 1) EPA regulation on concentration in agriculture, 2) USDA regulation on concentration in the food production industry, and 3) FDA regulation on productivity and profitability in the food production industry.

7. References

Alpay, Ebru, Steven Buccola, and Joe Kerkvliet. 2002. "Productivity Growth and Environmental Regulation in Mexican and U.S. Food Manufacturing." American Journal of Agricultural Economics 84 (4): 887-901.

Alston, J. M., and J. S. James. 2002. "The Incidence of Agricultural Policy." In Handbook of Agricultural Economics, by Bruce L. Gardner and G. C. Rausser, 1689-1749. North-Holland.

Al-Ubaydli, Omar, and Patrick A. McLaughlin. 2015. "RegData: A Numerical Database on Industry-Specific Regulations for All United States Industries and Federal Regulations, 1997-2012." Regulation and Governance 11 (1): 109-123.

Ambec, Stefan, Mark A. Cohen, Stewart Elgie, and Paul Lanoie. 2013. "The Porter Hypothesis at 20: Can Environmental Regulation Enhance Innovation and Competitiveness?" Review of Environmental Economics and Policy 7 (1): 2-22.

Bailey, James B, and Diana W Thomas. 2017. "Regulating away competition: the effect of regulation on entrepreneurship and employment." Journal of Regulatory Economics 52 (3): 237-254.

Becker, Randy, and Vernon Henderson. 2000. "Effects of Air Quality Regulations on Polluting Industries." Journal of Political Economy 108 (2): 379-421.

Chambers, Dustin, and Courtney A. Collins. 2016. How Do Federal Regulations Affect Consumer Prices? Mercatus Working Paper, Arlington, VA: Mercatus Center at George Mason University.

Coffey, Bentley, Patrick A. McLaughlin, and Pietro Peretto. April, 2016. The Cumulative Cost of Regulations. Mercatus Working Paper, Arlington, VA: Mercatus Center at George Mason University.

Davies, Antony. 2014. Regulation and Productivity. Mercatus Research, Arlington, VA: Mercatus Center at George Mason University.

Dawson, J. W., and J. J. Seater. 2013. "Federal Regulation and Aggregate Economic Growth." Journal of Economic Growth 18 (2): 137-177.

Dudley, Susan E. 2013. "OMB's Reported Benefits of Regulation: Too Good to Be True?" Regulation 36 (2).

Greenstone, Michael. 2002. "The Impact of Environmental Regulations on Industrial Activity: Evidence from 1970 to 1977 Clean Air Act Amendments and the Census of Manufactures." Journal of Political Economy 110 (6): 1175-1219.

Gurtoo, Anjula, and S.J. Antony. 2007. "Environmental Regulations: Indirect and Unintended Consequences on Economy and Business." Management of Environmental Quality: An International Journal 18 (6): 626-642.

Lanoie, Paul, Jeremy Laurent-Lucchetti, Nick Johnstone, and Stefan Ambec. 2011. "Environmental Policy, Innovation, and Performance: New Insights on the Porter Hypothesis." Journal of Economics and Management Strategy 20 (3): 803-842.

Loayza, Norman V., Ana Maria Oviedo, and Luis Serven. 2004. Regulation and Macroeconomic Performance. Policy Research Working Paper No. 3469, Washington, D.C.: World Bank.

Miller, J. Corey, and Keith H. Coble. 2007. "Cheap Food Policy: Fact or Rhetoric?" Food Policy 32 (1): 98-111.

Pasour, Jr., E.C., and Randall R. Rucker. 2005. Plowshares and Pork Barrels: The Political Economy of Agriculture. Oakland, CA: The Independent Institute.

Porter, Michael E. 1991. "America's Green Strategy." Scientific American, April: 168.

Vercammen, James. 2011. "Agri-Environmental Regulations, Policies, and Programs." Canadian Journal of Agricultural Economics 59 (1): 1-18.