How do federal regulations affect consumer prices? An analysis of the regressive effects of regulation

Abstract

This study is the first to measure the impact of federal regulations on consumer prices. By combining consumer expenditure and pricing data from the Bureau of Labor Statistics, industry supply-chain data from the Bureau of Economic Analysis, and industry-specific regulation information from the Mercatus Center’s RegData database, we determine that regulations promote higher consumer prices, and that these price increases have a disproportionately negative effect on low-income households. Specifically, we find that the poorest households spend larger proportions of their incomes on heavily regulated goods and services prone to sharp price increases. While the literature explores other specific costs of regulation, noting that higher consumer prices are a probable consequence of heavy regulation, this study is the first to provide a thorough empirical analysis of that relationship across industries. Irrespective of the reasons for imposing new regulations, these results demonstrate that in the aggregate, the negative consequences are significant, especially for the most vulnerable households.

Appendices

Appendix 1: Proofs of predictions from theoretical model

Proof of Prediction 1

Consumer i’s spending on the regulated commodity as a share of her income is:

$$S_{i} = \frac{{px_{i} (p,m_{i} )}}{{m_{i} }}$$

(7)

Therefore:

$$\ln (S_{i} ) = \ln (p) + \ln (x_{i} (p,m_{i} )) - \ln (m_{i} )$$

(8)

$$\frac{{\partial \ln (S_{i} )}}{{\partial m_{i} }} = \frac{1}{{x_{i} }}\frac{{\partial x_{i} (p,m_{i} )}}{{\partial m_{i} }} - \frac{1}{{m_{i} }} = \frac{1}{{m_{i} }}\left[ {\varepsilon_{m} - 1} \right]$$

(9)

where $\varepsilon_{m}$ is the income elasticity of demand. Therefore, if demand with respect to income is inelastic, i.e., $0 < \varepsilon_{m} < 1$, then ${{\partial \ln (S_{i} )} \mathord{\left/ {\vphantom {{\partial \ln (S_{i} )} {\partial m_{i} }}} \right. \kern-0pt} {\partial m_{i} }} < 0$ □

Proof of Prediction 2

Application of the Implicit Function Theorem to Eq. (1) yields:

$$\frac{\partial p}{\partial R} = \frac{{\frac{\partial x(p,R)}{\partial R}}}{{\sum\nolimits_{i} {\frac{{\partial x_{i} (p,m_{i} )}}{\partial p} - \frac{\partial x(p,R)}{\partial p}} }} > 0$$

(10)

which shows that p is increasing in R. □

Proof of Prediction 3

Consumer i’s budget constraint is:

$$px_{i} (p,m_{i} ) + y_{i} = m_{i}$$

(11)

where y_i denotes consumer i’s expenditure on all commodities other than x_i. Dividing the budget constraint by m_i yields:

$$\frac{{px_{i} (p,m_{i} )}}{{m_{i} }} + \frac{{y_{i} }}{{m_{i} }} = 1$$

(12)

and differentiating with respect to R yields:

$$\frac{{x_{i} }}{{m_{i} }}\frac{\partial p}{\partial R} + \frac{p}{{m_{i} }}\frac{{\partial x_{i} (p,m_{i} )}}{\partial p}\frac{\partial p}{\partial R} + \frac{1}{{m_{i} }}\frac{{\partial y_{i} }}{\partial p}\frac{\partial p}{\partial R} = 0$$

(13)

Recall ${{\partial p} \mathord{\left/ {\vphantom {{\partial p} {\partial R}}} \right. \kern-0pt} {\partial R}} > 0$ that (Prediction 2). Equation (13) can be simplified to:

$$\frac{1}{{m_{i} }}\frac{{\partial y_{i} }}{\partial p} = - \frac{{x_{i} }}{{m_{i} }}\left[ {1 + \varepsilon_{p} } \right]$$

(14)

where is the regulated commodity’s price elasticity of demand. The left-hand side of Eq. (14) shows the proportional change in consumer i’s spending on other commodities when p increases. Recall that ${{x_{i} } \mathord{\left/ {\vphantom {{x_{i} } {m_{i} }}} \right. \kern-0pt} {m_{i} }}$ is decreasing in $m_{i}$ if demand with respect to income is inelastic (Prediction 1). Therefore, if demand with respect to price is also inelastic, i.e.,$- 1 < \varepsilon_{p} < 0,$ then an increase in p causes lower-income consumers to reduce their spending on other commodities by proportionately more than it does for higher-income consumers. □

Appendix 2: methodological description of the construction of the consumer expenditure survey and regulation datasets

To determine the disparate effects of government regulations on households in different socioeconomic strata, we construct a dataset that maps goods and services from the Consumer Expenditure Survey (CE) onto industry regulations from the Mercatus Center at George Mason University’s industry regulation database (RegData).

The CE provides detailed household spending and price data for a wide array of goods and services by income group. These goods and services are organized using the Universal Classification Codes (UCC) system. RegData 2.0, however, reports the level of industry regulation by the two-, three-, and four-digit North American Industry Classification System (NAICS) code for each year between 1997 and 2012. Therefore, to construct a usable database, we map regulations from the NAICS space onto goods and services in the UCC space. The resulting balanced panel dataset contains 9872 observations, covering 617 UCC-based goods and services over a 16-year period.

To construct the final dataset, the following steps are employed:

• 1. The RegData 2.0 dataset consists of two-digit, three-digit, and four-digit NAICS-based tables. Each regulation record in the tables contains the name of the government agency imposing the regulation, the year of the regulation, the industry affected by the regulation, the regulatory word count, the restriction count, and the industry regulation index value. For our purposes, we use the industry regulation index value, which equals the regulatory restriction count weighted by industry relevance.³¹

  For each industry-and-year pair, the industry regulation index values are summed across federal regulators. Therefore, for each industry-and-year combination, a single-industry regulation index value is derived, equaling the sum of all regulatory restrictions (weighted by industry relevance) imposed on that industry by all federal regulators for that year. The result is three aggregated datasets, one for each two-digit, three-digit, and four-digit NAICS-based table. Last, the three aggregated datasets are combined (stacked) to form a single dataset.

• 2. The spreadsheet containing the 2007 commodity-by-industry direct requirements (after redefinitions) table was downloaded from the Bureau of Economic Analysis (BEA) website (http://www.bea.gov/industry/xls/io-annual/CxI_DR_2007_detail.xlsx). This spreadsheet contains two work sheets, both of which are used below:

  • a. The first work sheet is a concordance that converts the BEA’s input–output (I–O) commodity/industry codes into 2007 NAICS codes.

  • b. The second work sheet is the I–O direct requirements table, which contains I–O weights ($\alpha_{ij}$) equal to the amount of input (measured in dollars) from industry ($i$) required to produce a dollar’s worth of output by industry ($j$). By construction, these weights sum to 1 because, in addition to actual inputs, the BEA includes employee compensation, taxes, and gross operating surplus in the weighting schema.

• 3. The I–O commodity/industry code to NAICS concordance described in step (2a) above is matched with the aggregate industry regulations from step (1), to create a new table that lists the aggregate industry regulations by I–O commodity/industry code; the resulting table is further summed over commodity code by year to derive a table with a single total regulation value for each commodity code–year pair. This second round of aggregation after the initial match is necessary because some commodity codes map onto multiple NAICS industries. I–O commodity/industry codes with no associated regulations are assigned an industry regulation index value of 0. The resulting table is a measure of the direct regulations (denoted $DirectReg_{it}$) applicable to a given I–O commodity/industry code.

• 4. To determine the level of regulation that applies to the inputs/supply chain of a given industry, the I–O direct requirements ($\alpha_{ij}$) from step (2b) are matched with the direct regulations for each I–O commodity ($DirectReg_{it}$) from step (3) by way of their I–O commodity/industry codes. Note that if a commodity/industry is not needed to produce a given output, the associated input value is 0. This produces a large result set with more than 2.4 million rows of data. This dataset is then “grouped by” output industry ($j$) and year ($t$) and summed over the product of the direct input regulations (indexed by i) and I–O weights, producing an estimate of input–supply chain regulation:$InputReg_{jt} = \mathop \sum \limits_{i} \alpha_{ij} \cdot DirectReg_{it} .$

  See Figure 4 for a graphical summary of steps (1) to (4).

  Fig. 4

  

  Mapping regulations onto input–output (I–O) codes. Note: NAICS = North American Industry Classification System

• 5. The direct regulations by industry and year are matched with the total input regulations by industry and year. The direct and input regulations are summed to determine the total direct and indirect regulations affecting a given industry: $TotalReg_{it} = DirectReg_{it} + InputReg_{it} .$

• 6. To map regulations onto the UCC codes, a separate set of queries is executed to map the codes onto I–O commodity/industry codes.

  • a. As a beginning step, we import the personal consumption expenditures (PCE) concordance from the Bureau of Labor Statistics (BLS) (http://www.bls.gov/cex/pce_concordance_2012.xlsx) This file maps UCC codes onto PCE codes from the BEA’s national income and product accounts (NIPAs).

  • b. Next, we import BEA table 2.4.5U (I–O, Personal Consumption Expenditures by Type of Product with 2007 Input–Output Commodity Composition). This latter bridge file (http://www.bea.gov/national/xls/2007-pcs-io-bridge.xls) maps NIPA line numbers onto PCE codes.

  • c. The tables from steps (6a) and (6b) are matched by way of their common PCE codes. The resulting table serves as a bridge file that maps UCC codes onto NIPA line numbers.

• 7. Finally, we import the BEA’s PCE bridge file, which maps NIPA line numbers onto I–O commodity/industry codes (www.bea.gov/industry/xls/io-annual/PCEBridge_2007_Detail.xlsx), along with the total value of all purchases of the linked I–O commodity/industry in 2007.

  • a. Matching the NIPA line items from the PCE bridge with the results from step (6c) provides a clear mapping from UCC code to I–O commodity/industry codes. See Figure 5 for a graphic summary of steps (6) and (7).

    Fig. 5

    

    Mapping input–output (I–O) codes onto consumer expenditure codes. Note: PCE = personal consumption expenditures; UCC = Universal Classification Codes; NIPA = national income and products accounts

• 8. The resulting table from step (7a) maps a given consumer product from the CE onto all I–O industries that produce that product. In many cases, more than one industry produces a given UCC product. To produce a single regulation value for each consumer product, we derive industry weights equal to a given industry’s 2007 level of output relative to the total output of all industries that supply a given UCC product.³² For example, the UCC code for flour is 10,110. This consumer product is produced by seven I–O industries. Assigning each of these industries a weight equal to its total output relative to the total output of all seven industries produces a set of weights that sum to 1 (see Table 8). Although it would be preferable to update these weights annually, the BLS derives these output data from the US Census Bureau’s Economic Census, which is conducted only every five years.

  Table 8 Input–output industries that produce flour (UCC: 10110)

• 9. Finally, UCC codes, I–O commodity/industry codes, and output shares from step (8) are matched with the regulation-by-industry data from step (5). These matched data are then “grouped by” UCC code and year and aggregated over the product of industry regulation and output shares.

See Table 9.

Table 9 Top 20 expenditure categories by income quintile and corresponding regulations