Skip to main content

Federal regulation and aggregate economic growth

Abstract

We introduce a new time series measure of the extent of federal regulation in the U.S. and use it to investigate the relationship between federal regulation and macroeconomic performance. We find that regulation has statistically and economically significant effects on aggregate output and the factors that produce it—total factor productivity (TFP), physical capital, and labor. Regulation has caused substantial reductions in the growth rates of both output and TFP and has had effects on the trends in capital and labor that vary over time in both sign and magnitude. Regulation also affects deviations about the trends in output and its factors of production, and the effects differ across dependent variables. Regulation changes the way output is produced by changing the mix of inputs. Changes in regulation offer a straightforward explanation for the productivity slowdown of the 1970s. Qualitatively and quantitatively, our results agree with those obtained from cross-section and panel measures of regulation using cross-country data.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11
Fig. 12

Notes

  1. 1.

    Friedman and Friedman (1979), Becker and Mulligan (1999), and Mulligan and Shleifer (2003).

  2. 2.

    See the Appendix for details on the method of construction. Note in particular that we have accounted for changes in typeface and page sizes.

  3. 3.

    The predecessors are Nicoletti et al. (2000, 2001), Bassanini and Ernst (2002), Djankov et al. (2002, 2006), Nicoletti and Scarpetta (2003), Loayza et al. (2004, 2005), Alesina et al. (2003) and Kaufman et al. (2003).

  4. 4.

    Similarly, Loayza et al. (2005) found very high correlations among their seven indices of regulation.

  5. 5.

    We use Granger causality rather than simple intertemporal cross-correlations because Granger causality controls for autocorrelation and so gives a stronger measure of genuine independent predictive power of one kind of regulation for another kind. Indeed, that is the whole point of Granger causality. Our goal here is to show that measures of regulation that are restricted to just a few types of regulation may lead to spurious inference because they may attribute to one kind of regulation effects arising from other unrelated types. Granger causality seems a better indicator of such a problem than simple correlation.

  6. 6.

    An issue that afflicts all existing measures of regulation is that some regulation comes into existence in response to innovation: changes in the production process or the invention of new products. Regulations of that type may merely keep the regulatory burden constant as the economy grows rather than increase it, so increases in either the page count of the CFR or the number of regulations would overstate to some extent the increase in regulatory burden over time. Endogenous growth theory suggests that the way to handle that issue is to measure regulation per product or per firm, but time series data on the number of products or number of firms do not exist for our sample period.

  7. 7.

    Furthermore, growth theory suggests that anti-monopoly regulation, which would raise output in the short run by eliminating monopoly restrictions on supply, may reduce output in the long run by reducing the monopoly returns necessary to justify R&D and thereby reducing the rate of output growth.

  8. 8.

    See the Wall Street Journal’s front page article “Sour Economy Spurs Government to Grab a Bigger Oversight Role”, 25 July 2008.

  9. 9.

    One relevant detail is the nature of government purchases. Equation (1) is based on the assumption that the government sets the tax rates and adjusts purchases to satisfy the government budget constraint. If instead the government sets purchases and adjusts tax rates, a somewhat different form arises. As we discuss below, evidence on the behavior of government expenditures and tax rates suggests that equation is the right form to use. Another detail is that Peretto restricts analysis to constant tax rates. Intertemporal changes in tax rates introduce additional elements in the A, B, and C functions. We address this issue briefly in our discussion of the specification of the estimating equation.

  10. 10.

    For example, the intercept term \(A\)(\({t}_{L})\) in Peretto (2007a) is the simplest of the three functions and has the form:

    $$\begin{aligned} \frac{\left( {1-t_L } \right) \left( {1-\theta } \right) }{\left( {1-t_L } \right) \left( {1-\theta } \right) \left( {1+\gamma } \right) +\gamma \left( {\rho -\eta } \right) \beta \theta ^{2}} \end{aligned}$$

    which is neither linear nor log-linear in \({t}_{L}\) or even \(1-{t}_{L}\). The functions \(B\)(.) and \(C\)(.) are far more complicated.

  11. 11.

    Fractional integration models have been shown to be observationally equivalent to trend-break models (Diebold and Inoue 2001).

  12. 12.

    The probability of dying affects the rate of time preference in the household choice model extended to include random time of death. See Blanchard (1985).

  13. 13.

    Barro and Ridlick (2011) present an alternative series on marginal income tax rates. It is an extension of the series originally published by Barro and Sahaskul (1983, 1986). However, Akhand and Liu (2002) have shown that the Barro–Sahaskul construct is considerably inferior to Stephenson’s as a measure of average marginal tax rates, so we restrict attention to Stephenson’s series.

  14. 14.

    The \(U\) component of the compound residual can include any exogenous variable not subject to analysis. The trends in such variables are included in the trend-apart term \(\beta \), so that \(U\) captures the transient components.

  15. 15.

    See, for example, Perron (1989, 1997), Banerjee et al. (1992), Zivot and Andrews (1992), and Vogelsang and Perron (1998) for further details.

  16. 16.

    We have not pursued the possibility of decomposing \(G\) into major parts (such as federal versus state and local, or national defense versus road building), even though different kinds of expenditure almost certainly have different effects on the economy and may interact with regulation in different ways. We also ignore government debt on the assumption that Ricardian equivalence holds, a proposition with much support in the empirical literature.

  17. 17.

    Both log-levels and levels of \(R, T\), and \(G\) appear in, so we examined Granger causality using levels of \(R, T\), and \(G\) as well. The results are qualitatively the same as those reported in Table 4.

  18. 18.

    We also performed Granger-causality tests on the individual titles of the CFR. Few individual titles Granger-cause any of the macroeconomic variables considered here, but the whole set of titles is jointly significant. With nearly 50 individual titles and only 57 observations, the tests have few degrees of freedom, leaving the results uninformative. This problem with degrees of freedom arises again later, where we discuss it in more detail.

  19. 19.

    More specifically, in the trend term we replace \(R_{t}\) with \([(R_{t}-R_{ 1949})+R_{ 1949}]^{2} =\Delta R_{t}+R_{t}\) and replace \(R_{t}^{2}\) with \([(R_{t}-R_{ 1949})+R_{ 1949}]^{2} = (\Delta R_{t})^{2}+2R_{t}\Delta R_{t}+(R_{t})^{2}\), multiply by the appropriate estimated trend coefficients (\(\gamma \) and \(\delta \)), and collect all terms containing \(\Delta R_{t}\). Similarly, in the cycle term we replace \(R_{t}\) with \(R_{ 1949}({R}_{t}/{R}_{ 1949})\), raise to appropriate estimated power (\(\omega \)), and collect all terms involving ratios of the form \((R_{t}/R_{ 1949})^{\omega }\). Finally, dividing actual output by the trend arising from the terms containing \(\Delta {R}_{t}\) and by the cycle components (\(R_{t}/{R}_{ 1949})^{\omega }\) is equivalent to setting \(\Delta R\) to zero and (\(R_{t}/{R}_{ 1949})^{\omega }\) to one, thus giving a counterfactual value of output under the restriction that regulation remained at its 1949 level.

  20. 20.

    Of course, because we have restricted our functional form to a quadratic, ridiculous results can be obtained by extrapolating far beyond the sample period. If regulation continues to grow, the negative terms take over and eventually output growth becomes negative, driving output toward zero as time passes. Such behavior obviously would not be tolerated by society, and the process governing the evolution of regulation would change. The problem is exactly the same as using a quadratic utility function to approximate the true function: it can work quite well locally but will give nonsensical results if abused. These problems of extrapolation are not relevant to our discussion here, which is confined to behavior within the sample period.

  21. 21.

    Indeed, Peretto (2008) finds that changes in tax rates can reduce the growth rate of output but raise social welfare. The same divergence could happen with regulation.

  22. 22.

    This 8 % excludes the cost of tax compliance, which Crain and Hopkins included. We exclude tax compliance cost because taxes generally are not considered “regulations.” Tax compliance cost amounts to about one half of 1 % of GDP.

  23. 23.

    See especially Parente and Prescott’s chapters 6 and 7.

  24. 24.

    Note that there is a misprint in Djankov et al.’s published Table 3. The entry in the first row of the last column of Table 3 should be \(-2.3241\), not \(-0.3241\). See their earlier working paper (2005) for the correct entry.

  25. 25.

    Other estimated aggregate effects of regulation also are large. For example, Loayza et al. (2005) find that increasing a country’s index of labor regulation by one standard deviation in the cross-country sample would increase the size of the informal sector relative to GDP by nearly 3 percentage points. Nicoletti et al. (2001) find that cross-country differences in product market regulations in the OECD account for up to 3 percentage points of deviations of the employment rate from the OECD average.

  26. 26.

    Nicoletti et al. report (in their Table 2) a semi-elasticity of \(-\)0.75. We converted that to an elasticity using the average product market regulation value of 1.94, found by averaging the individual country values reported in (Nicoletti et al. (2000), Table A3.6). We made similar conversions for semi-elasticity estimates pertaining to employment protection regulation, using the average value of employment protection regulation from Table A3.11 in Nicoletti et al. (2000).

  27. 27.

    The OECD’s (2003) regressions for entry rates are in levels, so the estimated coefficients are simple derivatives: d(entry rate)/d(regulation). We converted to elasticities by multiplying by \((\text{ regulation })/(\text{ average } \text{ entry } \text{ rate } \approx 0.10)\), where the country entry rates are reported in (Nicoletti et al. (2001), Table A2.1).

  28. 28.

    In this regard, our results differ from those in Alesina et al. (2003), whose estimates of regulatory impact are insensitive to inclusion or omission of fiscal policy variables.

  29. 29.

    See the OECD’s (2003, Table 3.3) for estimated coefficients and see (Nicoletti et al. (2000), Table A3.2), for values of barriers to entrepreneurship.

  30. 30.

    Note that disaggregation by title is not the same as disaggregation by industry affected, nor does it necessarily capture all regulations of a general type. “Agriculture” and “Animals and Animal Products” are separate titles that both affect the agriculture industry. “Banks and Banking” is a title that may affect many industries. For some purposes, it might be preferable to measure some group of related regulations, such as all regulations pertaining to agriculture. This is the approach taken in some of the literature cited in the Introduction.

  31. 31.

    Throughout the appendix, the following citation format is used: volume or title number followed by name of publication followed by page or section number. For example, “49 Stat. 500” designates Volume 49 of the United States Statutes at Large, p. 500. The following abbreviations are also used in the citations: USC for United States Code, FR for Federal Register, and CFR for Code of Federal Regulations.

  32. 32.

    The U.S. Statutes at Large and U.S. Code are comparable to the Federal Register and Code of Federal Regulations, respectively, except that the former are primarily concerned with the publication and codification of laws, whereas the later are concerned with transmitting to the public written requirements to be carried out and enforced by government agencies (i.e., regulations). Thus, the CFR is more appropriate than the U.S. Code as a measure of regulation.

  33. 33.

    No supplement was published for 1942.

  34. 34.

    Due to the imminence of the second edition of the CFR, no supplement was issued for 1948. Regulatory changes published in the FR during 1948 were codified for the first time in the 1949 edition of the Code.

  35. 35.

    The term “pocket supplement” derives from pockets which were made in the books of the 1949 edition of the CFR for placement of the forthcoming supplements.

  36. 36.

    On several occasions, an “added pocket part” (APP) was published instead of a pocket supplement. The APP served as an addition or supplement to the previous year’s pocket supplement. APPs were not cumulative unless they appeared in consecutive years, in which case the old APP was replaced by the current APP as a supplement to the most recent pocket supplement.

  37. 37.

    Beginning with the 1973 revision of the CFR, the effective revision date of each title varies within the year according to the following quarterly schedule: Titles 1–16 as of January 1; Titles 17–27 as of April 1; Titles 28–41 as of July 1; and Titles 42–50 as of October 1.

  38. 38.

    There is the possibility that the nature of the language used has changed over time, becoming more or less verbose as time passes, so that the word-to-regulatory-content ratio changes over time. We have no way to control for or even check the existence of such a problem. We know of no evidence that such changes in language occurred over our sample period. Kimble (2002, 2007) has suggested changes in usage that would replace the legalese of federal laws and regulations with more straightforward English, but Kimble himself notes that such changes sometimes result in longer rather than shorter documents. Also, we know of no evidence that such suggested changes ever have been adopted. Indeed, there is anecdotal evidence that simplification has not occurred. President Carter signed an executive order in 1978 requiring that federal regulations be written as simply and clearly as possible. Twenty years later, in 1998, the Clinton administration demanded that regulations be written in plainer prose, suggesting that President Carter’s order had not had achieved its goal. We thus proceed on the assumption that there have been no significant changes in the verbosity of regulation over time.

  39. 39.

    Recall from the discussion above that the timing of revisions to the 1949 edition of the CFR varies across titles between the years 1949 and 1969.

  40. 40.

    Dawson (2007) discusses the “double-counting” problem in more detail and offers some alternative methods for constructing the regulatory series based on interpolation in the non-revision years. The results of the analysis in this paper are not sensitive to the construction method, thus we restrict attention to the series discussed here.

  41. 41.

    See Nicoletti et al. (2000, 2001), Djankov et al. (2002, 2005), Kaufman et al. (2003), Nicoletti and Scarpetta (2003), and Loayza et al. (2004) for detailed descriptions of these alternative measures.

  42. 42.

    Similarly, Loayza et al. (2005) found very high correlations among their seven indices of regulation.

  43. 43.

    Nicoletti and Pryor (2001) argue that the OECD measure is objective, but it clearly is not. Indeed, in describing its construction, Nicoletti et al. (2000) say that 90 % of the data underlying the OECD measure is “survey data” (p. 11) and that both questionnaire responses and the procedure for scoring them involved “subjective judgement” (p. 16).

  44. 44.

    At least for federal regulations. Our measure does not include state and local regulations.

  45. 45.

    Friedman and Friedman (1979) use the number of pages in the Federal Register to measure the growth of regulation. Becker and Mulligan (1999) use pages in the U.S. Code as a measure of growth in the size of government. Mulligan and Shleifer (2003) use kilobytes of unannotated state law, where 1kb approximately equals one printed page, to study the causes of regulation.

References

  1. Akhand, H., & Liu, H. (2002). Marginal income tax rates in the United States: A non-parametric approach. Journal of Monetary Economics, 49, 383–404.

    Article  Google Scholar 

  2. Alesina, A., Ardagna, S., Nicoletti, G., & Schiantarelli, F. (2003). Regulation and Investment. Working paper # 9560. National Bureau of Economic Research, Cambridge, MA.

  3. Averch, H., & Johnson, L. L. (1962). Behavior of the firm under regulatory constraint. American Economic Review, 52, 1053–1069.

    Google Scholar 

  4. Bandiera, O., Caprio, G., Honohan, P., & Schiantarelli, F. (2000). Does financial reform raise or reduce saving? Review of Economics and Statistics, 82, 239–263.

    Article  Google Scholar 

  5. Banerjee, A., Lumsdaine, R. L., & Stock, J. H. (1992). Recursive and sequential tests of the unit-root and trend-break hypothesis: Theory and international evidence. Journal of Business and Economic Statistics, 10, 271–287.

    Google Scholar 

  6. Barro, R. J. (1981). Output effects of government purchases. Journal of Political Economy, 89, 1086–1121.

    Article  Google Scholar 

  7. Barro, R. J., & Ridlick, C. J. (2011). Macroeconomic effects from government purchases and taxes. Quarterly Journal of Economics, 126, 51–102.

    Article  Google Scholar 

  8. Barro, R. J., & Sahaskul, C. (1983). Measuring the average marginal tax rate from the individual income tax. Journal of Business, 56, 419–452.

    Article  Google Scholar 

  9. Barro, R. J., & Sahaskul, C. (1986). Average marginal tax rates from social security and the individual income tax. Journal of Business, 59, 555–566.

    Article  Google Scholar 

  10. Bassanini, A., & Ernst, E. (2002). Labor market institutions, product market regulations and innovation: Cross country evidence. OECD Economics Department Working Papers, No. 316.

  11. Becker, G. S., & Mulligan, C. B. (1999). Accounting for the growth of government. Working paper, University of Chicago.

  12. Blanchard, O. (1985). Debt, deficits, and finite horizons. Journal of Political Economy, 93, 223–247.

    Article  Google Scholar 

  13. Blanchard, O., & Giavazzi, F. (2003). Macroeconomic effects of regulation and deregulation in goods and labor markets. Quarterly Journal of Economics, 118, 879–908.

    Article  Google Scholar 

  14. Christandl, F., & Fetchenhauer, D. (2009). How laypeople and experts misperceive the effect of economic growth. Journal of Economic Psychology, 30, 381–392.

    Article  Google Scholar 

  15. Crain, W. M., & Hopkins, T. D. (2001). Impact of regulatory costs on small firms. Report SBAHQ-00-R-0027, Office of Advocacy. Small Business Administration, Washington, DC. U.S .

  16. Dawson, J. W. (2007). Measuring federal regulation in the U.S.: 1938–2005. Working paper, Appalachian State University.

  17. Diebold, F. X., & Inoue, A. (2001). Long memory and regime switching. Journal of Econometrics, 105, 131–159.

    Article  Google Scholar 

  18. Djankov, S., LaPorta, R., Lopez-De-Silanes, F., & Shleifer, A. (2002). The regulation of entry. Quarterly Journal of Economics, 117, 1–37.

    Article  Google Scholar 

  19. Djankov, S., McLiesh, C., & Ramalho, R. (2005). Regulation and growth. Working paper, World Bank, March.

  20. Djankov, S., McLiesh, C., & Ramalho, R. (2006). Regulation and growth. Economics Letters, 92, 395–401.

    Article  Google Scholar 

  21. Elliott, G., Rothenberg, T. J., & Stock, J. H. (1996). Efficient tests for an autoregressive unit root. Econometrica, 64, 813–836.

    Article  Google Scholar 

  22. Friedman, M., & Friedman, R. (1979). Free to choose. New York: Avon Books.

    Google Scholar 

  23. Goff, B. (1996). Regulation and macroeconomic performance. Boston: Kluwer.

    Google Scholar 

  24. Greenwood, J., & Yorukoglu, M. (1997). 1974. Carnegie-Rochester Conference Series on Public Policy, 46, 49–95.

    Article  Google Scholar 

  25. Hakkio, C. S., & Rush, M. (1991). Cointegration: How short is the long run? Journal of International Money and Finance, 10, 571–581.

    Article  Google Scholar 

  26. Hamilton, J. D. (1994). Time Series Analysis. Princeton, NJ: Princeton University Press.

    Google Scholar 

  27. Hopkins, T. D. (1991). Cost of regulation. Rochester, NY: Rochester Institute of Technology.

    Google Scholar 

  28. Kaufman, D., Kraay, A., & Mastruzzi, M. (2003). Governance matters. Working paper #3106. The World Bank, Washington, DC.

  29. Kimble, J. (2002). A crack at federal drafting. Court Review, Spring, 44–47.

  30. Kimble, J. (2007). Lessons in drafting from the new federal rules of civil procedure (Part 1). Michigan Bar Journal, Spring, 62–65.

  31. Loayza, N. V., Oviedo, A. M., & Servén, L. (2004). Regulation and macroeconomic performance. Policy Research Working Paper, No. 3469, World Bank, September.

  32. Loayza, N. V., Oviedo, A. M., & Servén, L. (2005). The impact of regulation on growth and informality: cross-country evidence. AEI-Brookings Joint Center for Regulatory Studies, Related Publication 05–11, May.

  33. Lumsdaine, R. L., & Papell, D. H. (1997). Multiple trend breaks and the unit-root hypothesis. Review of Economics and Statistics, 79, 212–218.

    Article  Google Scholar 

  34. Maddison, A. (2000). The world economy: A millennial perspective. Paris: OECD.

    Google Scholar 

  35. McChesney, F. S. (1987). Rent extraction and rent creation in the economic theory of regulation. Journal of Legal Studies, 16, 101–118.

    Article  Google Scholar 

  36. Mulligan, C. B., & Shleifer, A. (2003). Population and regulation. Working paper, University of Chicago.

  37. Murray, C. J., & Nelson, C. R. (2000). The uncertain trend in U.S. GDP. Journal of Monetary Economics, 46, 79–95.

    Article  Google Scholar 

  38. Nicoletti, G., Bassanini, A., Ernst, E., Jean, S., Santiago, P., & Swaim, P. (2001). Product and labor markets interactions in OECD countries. OECD Economics Department Working Papers, No. 312, OECD, Paris.

  39. Nicoletti, G., & Pryor, F. L. (2001). Subjective and objective measures of the extent of governmental regulations, Working Paper. OECD and Swarthmore College, December.

  40. Nicoletti, G., & Scarpetta, S. (2003). Regulation, productivity and growth: OECD evidence. Economic Policy, April 2003, pp. 9–72.

  41. Nicoletti, G., Scarpetta, S., & Boylaud, O. (2000). Summary indicators of product market regulation with an extension to employment protection legislation. OECD Economics Department Working Papers, No. 226, OECD, Paris.

  42. Nordhaus, W. D. (2004). A retrospective on the 1970s productivity slowdown. NBER Working Paper # 10950.

  43. OECD. (2003). The sources of economic growth in OECD countries, organization for economic Co-OPERATION and development. Paris: OECD.

  44. Parente, S. L., & Prescott, E. C. (1999, October). Barriers to Riches. Third Walras-Pareto Lecture. Lausanne: University of Lausanne.

  45. Peretto, P. (2007a). Corporate taxes, growth, and welfare in a Schumpeterian economy. Journal of Economic Theory, 137, 353–382.

    Article  Google Scholar 

  46. Peretto, P. (2007b). Schumpeterian growth with productive public spending and distortionary taxation. Review of Development Economics, 11, 699–722.

    Article  Google Scholar 

  47. Peretto, P. (2007c). Energy taxes and endogenous technological change. Working paper. Duke University.

  48. Peretto, P. (2008). A Schumpeterian analysis of deficit-financed dividend tax cuts. Working paper. Duke University.

  49. Perron, P. (1989). The great crash, the oil price shock, and the unit root hypothesis. Econometrica, 57, 1361–1401.

    Article  Google Scholar 

  50. Perron, P. (1997). Further evidence on breaking trend functions in macroeconomic variables. Journal of Econometrics, 80, 355–385.

    Article  Google Scholar 

  51. Pigou, A. C. (1938). The economics of welfare (4th ed.). London: Macmillan.

    Google Scholar 

  52. Sa, N. (2007). Sunk costs, market structure, and welfare: A general equilibrium interpretation. Working paper. Duke University.

  53. Saikkonen, P. (1992). Estimation and testing of cointegrated systems by an autoregressive approximation. Econometric Theory, 8, 1–27.

    Article  Google Scholar 

  54. Stephenson, E. F. (1998). Average marginal tax rates revisited. Journal of Monetary Economics, 41, 389–409.

    Article  Google Scholar 

  55. Stigler, G. J. (1971). The theory of economic regulation. Bell Journal of Economics and Management Science, 2, 3–21.

    Article  Google Scholar 

  56. Stock, J. H., & Watson, M. (1993). A simple estimator of cointegrating vectors in higher order integrated systems. Econometrica, 61, 783–820.

    Article  Google Scholar 

  57. Ventura, J. (1997). Growth and interdependence. Quarterly Journal of Economics, 112, 57–84.

    Article  Google Scholar 

  58. Vogelsang, T. J., & Perron, P. (1998). Additional tests for a unit root allowing for a break in the trend function at an unknown time. International Economic Review, 39, 1073–1100.

    Article  Google Scholar 

  59. Zivot, E., & Andrews, D. W. K. (1992). Further evidence on the great crash, the oil-price shock, and the unit-root hypothesis. Journal of Business and Economic Statistics, 10, 251–270.

    Google Scholar 

Download references

Acknowledgments

We thank Michele Boldrin, Tim Bollerslev, Michelle Connolly, Simeon Djankov, Ronald Gallant, Thomas Grennes, Bruce Hansen, Bang Jeon, Christopher Laincz, Oksana Leukhina, Aart Kraay, John Lapp, Marcelo Oviedo, Douglas Pearce, Denis Pelletier, Pietro Peretto, Martin Rama, George Tauchen, an anonymous referee, and an anonymous Associate Editor for helpful comments. We are also grateful to Amit Sen for providing finite-sample critical values for the Zivot–Andrews test. The regulation and marginal tax rate data are available on Seater’s website at: http://www4.ncsu.edu/~jjseater/index_003.htm.

Author information

Affiliations

Authors

Corresponding author

Correspondence to John J. Seater.

Electronic supplementary material

Appendix: Code of federal regulations

Appendix: Code of federal regulations

A.1 History and Background of the Code of Federal Regulations

Before 1935, no systematic process existed for the promulgation of federal regulations; regulations were simply typed and filed by individual agencies. The lack of public notification regarding regulatory activity later came to be known as “hip pocket” law, which led the government to embarrassment in Panama Refining Company v. Ryan (293 U.S. 388, 1935), also known as the “Hot Oil Case.”Footnote 31 The government’s case, which was based on a provision that was later nullified by a subsequent regulation, was dismissed by the Supreme Court, and both parties in the case were impugned for their ignorance of the law. This outcome led to the Federal Register Act of 1935 (49 Stat. 500; 44 USC Chapter 15), which established a consistent framework for codification of government regulations throughout the rulemaking process.

The Federal Register (FR), first published on March 14, 1936, is a daily publication in which proposed regulations appear first in draft form and eventually in final form, if passed into law. The FR also contains presidential proclamations, executive orders, announcements of agency hearings and meetings on regulatory issues, grant application instructions and deadlines, official agency decisions and actions, and agency establishments, reorganizations, and dissolutions. Sometimes, there also are long sections containing technical or economic analyses or discussion of issues arising during consideration of a proposed regulation. The final regulations (newly passed into law) contained in the FR ultimately are codified in the CFR. Divided into 50 subject categories called titles, the structure of the CFR is similar, but not identical, to that of the United States Code. Currently, each title of the CFR is revised annually and contains all regulations in effect as of the cover date.Footnote 32

The first edition of the CFR published regulations in force as of June 1, 1938. In the early years, the CFR was not revised annually. Instead, annual supplements carried in full text all changes and additions to the 1938 edition of the CFR as published in the FR. The supplements covered the periods June 2–December 31, 1938 and subsequent calendar years through 1941, listing regulatory changes promulgated during the period and in effect on December 31 of the year in question.Footnote 33 The first revision of the CFR, scheduled for June 1, 1943 under the Federal Register Act, was postponed because of the volume of rapidly changing regulations related to World War II and the preoccupation of all government agencies with the war effort. In its place, a cumulative supplement to the 1938 edition of the CFR compiled regulations in force as of June 1, 1943. However, regulations in effect at that date whose text was identical to that in the 1938 edition of the CFR are included only by reference to the original CFR. Also, emergency controls associated with the war period are recorded by tabulation rather than codification in the cumulative supplement. Thus, the cumulative supplement served as an adjunct to the original edition rather than a replacement of it. Following the cumulative supplement, annual supplements continued to update the 1938 edition of the CFR for regulatory changes published in the FR during the remainder of 1943 and each calendar year through 1947. The wartime suspension of the first revision of the CFR was terminated in 1948 and the second edition of the CFR, recording regulations in effect on January 1, 1949, was issued.Footnote 34

Following the 1949 edition of the CFR, “pocket supplements” were used to record regulatory changes published in the FR.Footnote 35 Pocket supplements differed from the annual supplements to the first edition of the CFR in that they were cumulative; that is, the pocket supplement for a given year recorded the full text of all changes to the 1949 CFR in effect at the end of the given year, irrespective of the year that the change occurred. The first pocket supplement covered changes during the June 2 to December 31, 1949 period and subsequent pocket supplements included any additional changes in effect at the end of each succeeding calendar year. So, for example, the 1950 pocket supplement documents changes to the 1949 edition of the CFR that occurred between June 2, 1949 and December 31, 1950. Some of those changes occurred between June 2 and December 31 of 1949 and so already were reported in the 1949 pocket supplement. The 1950 pocket supplement repeats them and adds all changes that occurred between January 1 and December 31 of 1950.Footnote 36

From time to time, as warranted by growth of the pocket supplements, individual titles (or individual parts of a title) of the 1949 CFR were revised. These revisions represented a complete codification of regulations in effect as of December 31 of the year in which they were published. The timing of revisions varied considerably across titles. In all titles, however, revisions became more frequent over time. In 1950, for instance, only Parts 71–90 of Title 49 (Transportation and Railroads) were revised. In 1960, all or parts of Titles 1–5, 14, 18–20, 26, 27, 32, 40, 41, 49, and 50 were revised, and by 1968, all except Titles 34, 35, and 37 were revised. Beginning in 1969, all titles of the CFR have been revised annually.Footnote 37

A.2 Measuring regulatory activity using the CFR

The consistent codification of federal regulations in the CFR since its inception in 1938 provides a unique source of information on regulatory activity over the years. Dawson (2007) constructs series measuring regulatory activity based on the number of pages published in the CFR’s various editions and supplements. Although the number of pages of regulation cannot capture the differential effects of alternative regulations on economic activity, it affords new information on the temporal behavior of the total amount of regulation in place. The remainder of this section provides a summary of these CFR-based measures of regulation. For a complete description of the methodology used to construct the series and a statistical comparison of the various series, see Dawson (2007).

Before counting pages, we must standardize the pages in the CFR for different words per page across the years. That turns out to be almost effortless. The CFR uses the same font and page size in all years except the very first, 1938. We converted 1938 pages to “standard” pages simply by multiplying by an adjustment factor based on average words per page computed by sampling words per page in each title of the Code. Even this adjustment turns out to be irrelevant to our empirical work because, for reasons to be explained momentarily, we started our sample period in 1949, thus omitting the non-standard 1938 edition of the Code entirely.Footnote 38

Measuring regulatory activity using data on the number of pages in the CFR is straightforward in years when the CFR is revised. These include the years 1938, 1949, all years after 1969, and some years between 1949 and 1969.Footnote 39 Estimating total pages of regulation during the periods between the 1938, 1949, and subsequent revisions is more problematic. One approach, which explicitly uses all annual and pocket supplement data to estimate total pages of regulation during years in which no revision is published, adds the number of pages in a nonrevision-year’s supplement to the number of pages in its corresponding complete CFR. The series that results from this methodology exhibits rapid growth in pages of regulation during most of the 1940s followed by a drastic decline in 1949. This behavior in part may reflect the increase in regulation associated with World War II and the subsequent decrease following the war, but it also is likely to reflect in part an element of double counting that is, for practical purposes, unavoidable with the supplements used to codify regulatory changes between the 1938 and 1949 revisions of the CFR. The supplements print the entire text of any section of regulation that changed, even if only one word was different. Consequently, a page of text in a supplement may represent completely new text that was not present in 1938 or may be almost entirely repetition of previously existing text. The only way to avoid double counting repeated text would be to read each reported change to determine how much of it was repetition, an obviously impractical task. Growth in the estimated pages of regulation resumes in the early 1950s and moderates into the 1960s. The same double counting problem exists after 1949 as before but is less severe because revised volumes of the CFR were published intermittently between 1949 and 1969. The frequency of these intermittent updates increased as time passed, with almost the entire CFR being revised in 1968. Consequently, the growth in the CFR page count between 1949 and 1969 is much more likely to be a genuine phenomenon than is the pre-1949 growth. Double-counting ceases to be an issue after 1968 because the entire CFR is published every year after that. Because the counting problems are much more severe before 1949 than after, we restrict attention in our study to the period 1949–2005.Footnote 40 Also, because we are interested in the effects of regulation on the private economy, we exclude from our page count all regulations in the first six titles, which pertain to the internal organization and operation of the federal government itself.

A.3 Comparison with other measures of regulation

A3.1 Description

Our measure covers one country over 57 years; earlier measures cover many countries over much shorter periods of time.Footnote 41 Some of the earlier measures are purely cross-sectional, applying to a single year; others cover more years and so are panel data. The longest time span of the panel sets is 20 years.

Our measure is more comprehensive than any of its predecessors. Federal law requires that all federal regulations be published in the CFR, so our measure includes literally every regulation issued by the federal government. No other measure of regulation comes close to that extent of coverage. For example, the most widely used of the earlier data sets is the OECD cross-section measure described by Nicoletti et al. (2000) and extended in part to a 20-year panel by Nicoletti et al. (2001). The cross-section data are restricted to product market and employment protection regulation; other types, such as environmental or occupational health and safety regulation, are ignored. The panel extension is restricted further to a small subset of seven non-manufacturing industries: gas, electricity, post, telecommunications, passenger air transport, railways and road freight. Types of regulations considered also are limited, with data availability varying by industry: barriers to entry (available for all industries), public ownership (all industries except road freight), vertical integration (gas, electricity and railways), market structure (gas, telecommunications and railways), and price controls (only road freight).

All measures of regulation including ours are aggregate indices. Our index is more highly aggregated than any of the others simply because it covers the full array of regulations, but all are aggregates. None simply reports a quantitative measure of the magnitude or effect of a single regulation. The OECD measure, for example, collects answers to about 1300 questions and combines them into an index through a multistep aggregation procedure. The methods of aggregation differ substantially across indices. Our method is to weight each regulation by its number of pages in the CFR, which captures at least partially the complexity of the regulation, as we discuss below. Many other indices are constructed as simple averages of basic data, with no attempt to weight by the importance or complexity of the regulations included. The OECD uses a multistep procedure, in which the OECD staff creates a collection of categorical sub-aggregates mostly as simple averages of basic data and then uses factor analysis to aggregate those into its final indices.

All measures except ours are based at least in part on survey data, typically obtained from questionnaires sent to government officials (OECD), market participants (Kaufman et al.), and/or lawyers (Djankov et al. 2002). Our measure is based solely on the page count of the CFR.

A3.2 Evaluation

Our measure is a pure time series covering a long time span. The earlier measures of regulation have short to non-existent time spans, the longest being 20 years and the shortest 1 year. Such data cannot be used to study regulation’s effects on macroeconomic dynamic adjustment paths, which requires following the evolution of variables through time. There is more hope of studying regulation’s effects on average growth rates by using the cross-sectional dimension of the data to overcome the inadequate time dimension, which is precisely what several of the previous studies do. However, growth is an intertemporal phenomenon, so it would be useful to have time series estimates of regulation’s effects on it, especially in light of Ventura’s (1997) demonstration that the interpretation of cross-country growth regressions is confounded by the effects of foreign trade. Our measure, with its comparatively long time dimension, allows us to study both the long-run growth and short-run dynamic adjustment effects of regulation. The earlier studies, with their strong cross-section element but weak intertemporal element, are better suited for cross-sectional issues.

Our measure also is more comprehensive than the earlier measures, none of which encompasses the total set of regulations in any country. Incomplete coverage leads to two problems: (1) omitted variables bias, and, in any time series study, (2) divergence between the time series behavior of subsets of regulation on the one hand and of total regulation on the other.

Table 1 shows that the page counts of the various titles of the CFR are highly correlated with one another, whether measured in levels or growth rates. The mean correlation among levels is 0.60, with an even higher median of 0.77. The maximum correlation in levels is 0.99, and the minimum correlation is \(-\)0.76. The correlations in growth rates are much lower, of course, with a mean of only 0.16 (median of 0.15), but there still are quite a few correlations of substantial magnitude, with the maximum and minimum being 0.74 and \(-0.63\), respectively.Footnote 42 Such high correlations show that including just one type of regulation in a statistical analysis is likely to be misleading because of multicollinearity and consequent omitted variables problems. The problem is even more severe when addressing issues of macroeconomic dynamics. The correlations in Table 1 are all contemporaneous; for analyzing time series behavior, we also want to know the dynamic relations among various types of regulations. Granger-causality tests show the intertemporal dependence of one series on another after accounting for the first series’s dependence on its own lagged values. Table 2 summarizes Granger-causality test results for two titles of the CFR related to the kinds of regulations studied in previous analyses—regulation of entry and regulation of labor markets. Title 16 of the CFR pertains to Commercial Practices, and Title 29 pertains to Labor. Table 2 shows that the page counts of those titles both Granger-cause and are Granger-caused by the page counts of other titles, some apparently quite unrelated in content to the subjects of titles 16 and 29. Similar results hold for most of the other titles of the CFR. These Granger-causality relations among CFR titles show that there are temporal orderings in the statistical relations among the types of regulation and provide strong evidence that a time series analysis restricted to a subset of regulations is likely to suffer from serious omitted variables bias.

The foregoing remarks have greatest force when applied to attempts to study the economic effects of a particular type of regulation. If one is interested in the impact of total regulation, the high correlations among the different types might actually be considered good news because they suggest that a subset of regulations may capture the behavior of the whole. Indeed, Nicoletti et al. (2001), who have perhaps the most restricted measure of all, interpret their indicators as “a proxy for the overall regulatory policies followed by OECD countries over the sample period (p. 43).”

Unfortunately, examination of the data shows this hope to be ill-founded. Nicoletti et al.’s (2001) measure spans 1978-98 and shows a 66 % decline over that period. Subsets of CFR titles corresponding to Nicoletti et al.’s measure behave similarly. For example, titles 23 (Highways), 46 (Shipping), and 49 (Transportation) of the CFR encompass regulation of air transport, railways, and road freight, one Nicoletti et al.’s regulation groups. The page count of titles 23, 46, and 49 drops from a total of 8400 in 1978 to 8261 in 1998, which is qualitatively the same behavior as Nicoletti et al.’s measure. Nevertheless, the page count of the total CFR displays the opposite behavior, rising 47 % over 1978-98. The inescapable implication is that subsets of regulation are not reliable proxies for total regulation.

Our measure of regulation is the only measure constructed by a completely objective method. Our measure consists of the page counts in the CFR, a number requiring no judgement to obtain. Subjectivity enters all other measures in two ways. First, as remarked above, all other measures are based at least in part on survey data. As Nicoletti et al. (2000) note, the people completing the surveys have some latitude in interpreting the survey questions and may respond idiosyncratically. Second, the survey must be designed and the responses must be combined, processes that involve the judgement of the investigator. For example, the OECD index begins with responses to about 1300 survey questions. The responses usually are Yes or No. Groups of these responses are combined by averaging to create categorical variables with values from 0 to 6. The procedure used for measuring the scope of public enterprise illustrates the issues. Respondents are asked if there are “national, state or provincial government controls in at least one of” 24 industries chosen by the OECD. Some of the industries chosen are 2-digit ISIC (e.g., wholesale trade, financial institutions), some are 3-digit (e.g., tobacco manufactures), and some are 4-digit (e.g., electricity, motion picture distribution and projection). Despite the differences in size and importance, all industries have been assigned the same weight of 1 in the construction of the categorical variables. See Table A2.1.1 in (Nicoletti et al. (2000), p. 60), for details.Footnote 43 Other measures of regulation are generally more subjective than the OECD’s. An oddity that results from subjectively deciding what is and is not regulation is that two of the published measures of regulation contain elements that have nothing to do with regulation. The OECD measure includes data on publicly-owned enterprises, a form of government intervention but not regulation. Loayza et al.’s (2005) measure includes data on spending and taxation and on the fraction of the workforce that is unionized, neither of which pertains to regulation. These measures thus confound regulation with other government and even non-government activities.

Some indices of regulation attempt to measure the burden imposed by the component regulations by including quantitative and/or qualitative data pertaining to the regulations included. Examples include the number of procedures a new firm must go through to start operation (Djankov et al. 2002) and regulatory complexity (OECD; Nicoletti et al. 2000). Our measure contains no such direct measures but nonetheless controls for regulatory burden to some extent. It seems reasonable to suppose that, on average at least, the more complex a regulation, the more pages it will require. Indeed, the OECD measures complexity by the presence or absence of a long list of regulatory requirements. The larger the number of requirements, the more the pages of regulations necessary to describe them, which is precisely what our measure captures. Indeed, our approach may give a more complete picture of regulatory burden than the OECD’s measure because page counts indicate not only the presence or absence of particular provisions but also their complexity.

A3.3 Summary

Our page count measure of the extent of regulation compares well with other measures. Although limited to a single country, it has a much longer time span than any other measure. It is unique in being totally comprehensive.Footnote 44 It also is unique in being 100 % objective both in the data underlying it and in the method of constructing the index. It offers indicators at two levels of aggregation—one final index for total regulation and many sub-indices for the different classes (“titles”) of regulation. It is easily replicated and easily updated. Finally, CFR page counts are a more precise measure of regulation than page counts of other federal publications, such as the Federal Register or the U. S. Code, suggested by others.Footnote 45 The Federal Register contains proposed regulations and other irrelevant material; the U. S. Code contains all federal laws, not just regulations.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Dawson, J.W., Seater, J.J. Federal regulation and aggregate economic growth. J Econ Growth 18, 137–177 (2013). https://doi.org/10.1007/s10887-013-9088-y

Download citation

Keywords

  • Regulation
  • Macroeconomic performance
  • Economic growth
  • Productivity slowdown

JEL classification

  • E20
  • L50
  • O40