Employment and Output Effects of Federal Regulations on Small Business

Transcription

1 Employment and Output Effects of Federal Regulatons on Small Busness Dustn Chambers * Salsbury Unversty and Jang-Tng Guo Unversty of Calforna, Rversde September 3, 208 Abstract Ths paper examnes the dsparate mpact of US federal regulatons on small busnesses. In the context of a two-sector dynamc general equlbrum macroeconomc model, we obtan three emprcally testable mplcatons of hgher regulaton: ) the total number of small frms s reduced, 2) the employment share of small frms shrnks, and 3) small frms share of total output declnes. Snce the frst of these testable hypotheses has already been confrmed n prevous studes, we focus our attenton on the latter two, and fnd strong emprcal support for both. Specfcally, we estmate that a ten percent ncrease n federal regulatons reduces the employment share of small frms by nearly 0.7%, and an equally large ncrease n federal regulatons decreases the output share of small frms by nearly.5%. JEL Classfcaton: C23, E23, L, L5 Keywords: regulaton, small busness, frm sze, ndustry concentraton, dynamc general equlbrum. * Correspondng Author. Department of Economcs and Fnance, Salsbury Unversty, 0 Camden Ave., Salsbury, MD 280, , Fax: , E-mal: DLChambers@salsbury.edu. Department of Economcs, 423 Sproul Hall, Unversty of Calforna, Rversde, CA, 9252, , Fax: , E-mal: guojt@ucr.edu.

2 . Introducton It s well known that tracng the economc mpact of federal regulatons on households and frms s a dffcult and contentous task. Well-desgned regulatons may enhance socal welfare by reducng negatve externaltes and correctng market falures, whle poorly conceved regulatons ( red tape ) may yeld few f any benefts. Nonetheless, busness regulatons mpose addtonal complance costs on frms. For example, Dawson and Seater (203) report that between 949 and 20, federal regulatons resulted n an accumulated loss of goods and servces totalng $38.8 trllon. More worryngly, there s strong reason to suspect that these hgh costs dsproportonately burden smaller frms. In another recent study, Cran and Cran (204) estmate that small busnesses (wth fewer than 50 employees) faced average complance costs of $,724 per employee as compared to $9,083 for large busnesses (wth more than 00 employees). Gven the mportance of small busness as a source of economc dynamsm, nnovaton, job growth and socal moblty, t s surprsng that few studes have nvestgated the outszed mpact of regulatons on these crtcally mportant busnesses. Moreover, the prevous research that has examned ths topc s prmarly emprcal, and does not postulate formal theoretcal models to motvate ts regresson models or results. Therefore, our paper seeks to fll ths gap n the lterature wth a two-sector general equlbrum model that generates emprcally testable predctons regardng the dsparate mpact of federal regulatons on small and large frms. Consstent wth our model, ths paper presents emprcal evdence that hgher regulatons reduce both small frms share of employment and output wthn the U.S. economy. Although t has been long understood that economes of scales n regulatory complance costs may gve larger frms an advantage over ther smaller compettors, a lack of ndustry-specfc regulaton data has hampered the emprcal examnaton of ths topc. Early research ether reled on crude proxes for the level of federal regulaton lke page counts n the Code of Federal Regulatons (CFR) (see for example Dawson and Seater [203]) or as noted by Ktchng et al. (205), potentally based feedback from surveys sent to small busness owners. Fortunately, we can emprcally test the effects of regulatons on small busnesses by utlzng a relatvely new database called RegData, whch was constructed usng machne learnng algorthms that mned the CFR for

3 language consstent wth regulatons and probablstcally matched these regulatons to the NAICS-coded ndustres to whch they most lkely apply (see McLaughln and Sherouse [207] for detals). Indeed, several recent studes have used RegData to examne the general mpact of federal regulatons on entrepreneurshp, wth conflctng results. Baley and Thomas (207) fnd that greater ndustry-specfc regulatons are assocated wth a reducton n the entry of new frms, wth the greatest mpact affectng smaller frms. Smlarly, Chambers et al. (208) fnd that an ncrease n ndustry-specfc regulatons s assocated wth a reducton n both the number and employment of small frms, whereas large frms (wth 500 or more employees) are unaffected. In a notable departure from the above two papers, Goldschlag and Tabarrok (208) modfy ther dependent varables (new frm formaton and hrng) by way of the Davs-Haltwanger- Schuh (DHS) transformaton, whch s clamed to be a more robust measure of dynamsm. The resultng regresson models, despte utlzng a rght-hand-sde structure very smlar to Baley and Thomas (207) and Chambers et al. (208) and the same underlyng data sources (.e., RegData and the Census of US Busness), fal to fnd a statstcally sgnfcant assocaton between federal regulatons and the transformed measures of new frm formaton or hrng. Ths lack of consensus, whch clearly results from dfferences n the regresson models dependent varables, underscores the need for a theory to generate testable emprcal hypotheses, whch n turn provde gudance for approprate emprcal specfcatons. The analytcal framework that forms the bass for our emprcal work s Dhawan and Guo s (200) dynamc general equlbrum macroeconomc model wth two producton sectors whch are made up of large and small frms, respectvely. The two sectors dffer from each other wth regard to the level of fxed set-up costs and returns-toscale n ther producton functons. Each sector has an ntermedate-good segment n whch monopolstcally compettve frms operate wth fully moble captal and labor nputs. The number of these ntermedate frms s determned endogenously through the condton of free entry and ext. A fnal good s produced n each sector from the set of avalable ntermedate goods n a perfectly compettve envronment. These two fnal goods are then aggregated nto a sngle output (GDP) that can be consumed or nvested For more detals on the DHS transformaton, see Davs et al. (998). 2

4 by the representatve household. Ths parsmonous structure s motvated by the observaton that both large and small frms exst sde by sde wthn many ndustres, whle practcally producng the same commodty. In a calbrated verson of the above macroeconomy, Dhawan and Guo (200) numercally show that at the model s steady state, an ncrease n government regulatons wll yeld the followng emprcally testable outcomes: ) the total number of small frms s reduced, 2) the employment share of small frms shrnks, and 3) small frms share of total output declnes. The frst outcome s approprately tested and supported by Baley and Thomas (207) and Chambers et al. (208), but the latter two predctons have not been emprcally evaluated n the lterature. Therefore, our emprcal analyss tests for a reducton n both small frms share of employment and output n response to hgher federal regulatons. Wth regard to both predctons, we fnd strong evdence n favor of our theoretcal model. Specfcally, we fnd that a ten percent ncrease n federal regulatons reduces the employment share of small frms by nearly 0.7%, and that an equally large ncrease n federal regulatons decreases the output share of small frms by nearly.5%. The remander of ths paper s organzed as follows. Secton 2 presents the model economy and analyzes ts equlbrum condtons. Secton 3 descrbes our emprcal models and dscusses ther results. Secton 4 concludes. 2. The Model Ths secton sketches the two-sector dynamc general equlbrum macroeconomc model developed by Dhawan and Guo (200). The economy s nhabted by a unt measure of dentcal nfntely-lved households, together wth two types of heterogeneous frms under ncreasng returns and monopolstc competton. In partcular, large and small frms produce the same fnal good usng technologes that exhbt dentcal factor ntenstes, but subjected to dfferent fxed set-up costs and returns-to-scale n producton. 2. Frms The macroeconomy s producton sde s comprsed of two sectors ndexed by =, 2, where sector s populated by large frms and sector 2 conssts of small frms. Snce frms are solvng a statc proft maxmzaton problem, the tme-subscrpts wll be 3

5 suppressed for notatonal convenence n ths subsecton. The fnal good n each sector Y s produced from combnng a contnuum of ntermedate nputs X j, j[ 0, ], wth M the followng constant returns-to-scale technology: Y Xjdj, 0, where M 0 represents the endogenously determned measure of ntermedate nputs that are utlzed n sector. The fnal-good segment of each sector s assumed to be perfectly compettve, and we denote P j as the prce of the j th ntermedate nput relatve to the fnal good n sector. The fnal-good producers' proft maxmzaton condton yelds the demand j j functon: X P Y, where the prce elastcty of demand s M. Ths n turn mples that s equal to the markup rato of prce over margnal cost or the degree of monopoly power. Each ntermedate good s produced by a monopolst, who mplements a producton functon that allows for ncreasng returns-to scale: () X K L Z, 0,, Z 0, j j j where K j and L j are captal and labor nputs employed by the j th ntermedate producer n sector. In addton, Z represents a constant amount of ntermedate goods that must be expended n sector as fxed costs for settng up producton facltes before any sale s made. The presence of such costs mples that the ntermedate-good technology (3) exhbts ncreasng returns n producton. Next, n accordance wth emprcal fndngs, reported by Mlls and Schumann s (985) and Fegenbaum and Karnan (99), that small/large frms are more/less flexble n handlng market fluctuatons snce they rely more on varable/fxed factors of producton, we postulate that Z > Z 2. On the other hand, addtonal ncreasng returns-to-scale wll be present n (3) when > because of rsng margnal productvty. We also note that from a large sample of publcly traded U.S. frms, Dhawan (996) fnds that the degree of returns-to-scale n producton s hgher for large frms; therefore γ > γ 2 s mposed. Under the assumpton that factor markets are perfectly compettve wthn each sector, the frst-order condtons for ntermedate frm j s proft maxmzaton problem are gven by 4

6 (2) ( ) (X Z )P (X Z )P w and r, j j j j Lj Kj where w s the real wage rate and r be the real rental rate of captal n sector. For analytcal smplcty, we further postulate that () both captal and labor nputs are fully moble across the two producton sectors; and that () frms can enter and ext the ntermedate-good segment of each sector freely, hence they wll not make any proft. Usng ths zero-proft condton and equaton (2) yelds the equlbrum sze of ntermedate frm j: where 0. In what follows, our analyss s Xj Z, restrcted to a symmetrc equlbrum at whch all ntermedate-good produces make the same decsons wthn each sector: K P P X X K M and L L j, j, j, j, for all j [ 0, M ], where K and L represent M the total captal stock and labor hours employed n sector. It s then straghtforward to show that the equlbrum number of ntermedate frms n sector s (3) M K L, Z where M M + M 2 ; and that the sectoral producton functon for the fnal good s (4) Y M X A K L, where A. Z Fnally, the total output (GDP) for the economy Y s generated through the followng CES aggregator functon: (5) Y Y Y,, 0 and, where the elastcty of substtuton between Y and Y 2 s. 2 It follows that the 2 When =, GDP s smply the sum of sectoral outputs,.e. Y = Y + Y 2. In ths case, there exsts a generc corner soluton n whch only large frms wll produce. Ths possblty s ruled out snce t s not consstent wth the emprcal evdence. 5

7 shadow prce of Y relatve to the aggregate output Y s gven by Y Y SP ; Y Y and that Y = SP *Y + SP 2 *Y 2 because (5) dsplays constant returns-to-scale. Under the assumpton of full factor moblty, wage and rental rates wll be equalzed across the two sectors: SP *w = SP 2 *w 2 = w and SP *r = SP 2 *r 2 = r. 2.2 Households The economy s also populated by a unt measure of dentcal nfntely-lved households, each endowed wth one unt of tme, and maxmzes a dscounted sum of lfetme utlty t L t (6) log Ct B, 0, B 0, t0 where s the dscount factor, B s a preference parameter, and denotes the nverse of the ntertemporal elastcty of substtuton n labor supply. In addton, C t and L t are the representatve household s consumpton and labor hours at tme t, respectvely. The budget constrant that t faces s gven by (7) C K K w L rk, K 0 gven, t t t t t t t 0 where K t s the household s captal stock, and 0, denotes the captal deprecaton rate. The frst-order condtons for the household s dynamc optmzaton problem are (8) BCtL t w t, C C (9) (0) t t K t t lm 0, t Ct r, t where (8) s an ntra-temporal condton that equates the household s margnal rate of substtuton between consumpton and lesure to the real wage rate; equaton (9) s the standard Euler equaton for ntertemporal consumpton choces; and (0) s the transversalty condton. 2.3 Symmetrc Equlbrum and Steady State We focus on the model s symmetrc equlbrum n whch producers of fnal and ntermedate goods maxmze profts; households maxmze utltes; and the marketclearng condtons n captal and labor markets wll hold: 6

8 K K K, and L L L. It can be shown that the equlbrum fractons of t t 2t t t 2t aggregate captal stock and labor hours used n sector, denoted as Kt and Lt, are equal to the same constant for all t, 2 () Z Kt Lt, where. Z 2 Substtutng () nto (4) and (5) yelds that the total output of the economy s gven by t t t 2 2 (2) Y AK L, where A A A, where < to rule out the possblty of sustaned endogenous growth. Next, usng the model s equlbrum condtons, t s straghtforward to show that there exsts a unque nteror steady state at whch the real rental rate, hours worked, and captal stock (expressed as bar varables) are (3) r r r ( ), L, and K L. B( r ) A Wth equaton (3), the correspondng steady-state expressons of all remanng endogenous varables can be easly derved. 3. Emprcal Results In a calbrated verson of the above macroeconomy, Dhawan and Guo (200; Table 3, Experment #2, p. 659) numercally show that ncreased government regulatons, represented by lowerng the markup-rato parameter λ, wll result n three emprcally testable outcomes on the model s steady state: ) the total number of small frms M 2 falls; 2) the employment share of small frms SP 2 *Y2 L2 L shrnks; and 3) the output share of small frms declnes. Intutvely, a hgher level of monopoly power decreases Y the ndvdual sze and total number of small frms (X 2 and M 2 ). As a result, the employment as well as the output shares of small frms wll fall. 7

9 As t turns out, the frst emprcal outcome has already been confrmed by Baley and Thomas (207) and Chambers et al. (208). Usng a panel of frm brths and deaths at the 4-dgt NAICS ndustry level from the Census of US Busness (SUSB) and ndustry-level federal regulaton data from RegData, Baley and Thomas (207) fnd robust evdence that hgher federal regulatons wthn a gven ndustry reduces the overall number of small frms wthn that ndustry. Specfcally, a one percent ncrease n federal regulatons s assocated wth 0.047% reducton n small frm formaton (.e. brths) n the subsequent year. Utlzng a smlar panel of net annual changes n total small frms by ndustry derved from the SUSB database matched wth changes n ndustry-level federal regulatons from RegData, Chambers et al. (208) fnd robust evdence that a one percent ncrease n federal regulatons s assocated wth a contemporaneous 0.042% declne n total small frms. Despte modelng dfferences, both papers yeld nearly dentcal results: a 0% ncrease n federal regulatons s assocated wth a nearly 0.5% reducton n the number of small frms. To verfy the remanng two testable mplcatons from our dynamc general equlbrum model, the emprcal analyss below uses regresson models smlar to Baley and Thomas (207) and Chambers et al. (208) to test for a reducton n both small frms share of employment and output n response to hgher federal regulatons. 3. Employment Share of Small Frms To test whether the employment share of small frms shrnks n response to rsng regulaton, we collect employment estmates by ndustry and frm sze from the Census Bureau s SUSB dataset and match 5-dgt NAICS code ndustres wth federal regulatons from RegData. 3 Unlke gross recepts (see Secton 3.2), employment estmates are avalable annually, thus our resultng panel spans 8 tme perods ( ) and 248 ndustres. Usng ths data, we estmate the fxed effects panel model of Baley and Thomas (207), replacng ther dependent varable (small frm brths and deaths) wth the natural log of the employment share of small frms: 4 3 Followng the US Small Busness Admnstraton, we classfy frms wth 500 or more employees as large busnesses and frms wth fewer employees as small busnesses. 4 Baley and Thomas (207) conduct extensve dentfcaton testng whch strongly supports ths regresson model specfcaton.e. the log dependent varable regressed on tme and ndustry fxed effects and the log of regulatons, hence we adopt t as our baselne model for testng purposes. 8

10 (4) employmentt t regt u t, where employment t s the natural log of the share of total ndustry s employment by small frms n year t, α s the ndustry fxed effect, δ t s a perod fxed effect, reg t s the natural log of total federal regulatory restrctons applcable to ndustry n year t, and u t s a mean zero error term. The ndustry fxed effects capture any dfferences n average employment shares of small frms across ndustres; whle the perod fxed effects capture common shocks to employment shares across ndustres, ncludng busness cycles and changes n government polcy (whch may be correlated wth varaton n ndustry-level regulatons). Gven the double-log specfcaton of the model, the coeffcent on ndustry regulatons can be nterpreted as an elastcty measure. Estmates of Equaton (4) are provded n Table. In our preferred specfcaton (see column of Table ), the regulaton coeffcent s negatve and statstcally sgnfcant at the % level, mplyng that a one percent ncrease n federal regulatons reduces the employment share of small frms by nearly 0.07%. Whle ths coeffcent may seem small n magntude, t mples that a 0% ncrease n federal regulatons reduces the employment share of small frms wthn an affected ndustry by almost 0.7%. Gven that nearly 59 mllon workers were employed by U.S. small busnesses n 205, the volume of affected workers s qute large. 5 Overall, the goodness of ft of ths model s qute hgh, explanng just over 96% of the varaton n the employment share of small frms. To demonstrate the robustness of these results, columns 2 through 4 of Table provde estmaton results of varants of Equaton (4). In column 2, the perod effects, whch could be correlated wth changes n federal polcy emboded n regulatory statutes, are removed. The resultng regulaton coeffcent s nearly dentcal (-0.07) and retans ts % level of statstcal sgnfcance, strongly suggestng that unobserved common shocks (whch ncludes common economc condtons) are uncorrelated wth the regulaton seres, as ther omsson would otherwse bas the regulaton coeffcent. In column 3, we follow Chambers et al. (208) and replace the perod fxed effects wth drect measures of the busness cycle. The coeffcent on unemployment s postve and statstcally sgnfcant at the % level, 5 Based on SUSB data, Chambers et al. (208) reports that 58,938,47 workers were employed by frms wth fewer than 500 employees n

11 albet economcally nsgnfcant wth a coeffcent value of In other words, a one percentage-pont ncrease n the unemployment rate (e.g. an ncrease from 4% to 5% unemployment) reduces the employment share of small frms by just under 0.0%. On the other hand, the GDP Gap turns out to be statstcally nsgnfcant. Overall, the magntude of the estmated regulaton coeffcent ncreases slghtly ( ), mplyng that a % ncrease n federal regulatons reduces the employment share of small frms by just over 0.09%. Fnally, column 4 elmnates both ndustry and tme perod fxed effects. The regulaton coeffcent s nearly dentcal to that of column 3, n whch the perod fxed effects are replaced by the busness cycle covarates. Taken together, these results mply that () most of the varaton n the employment share panel s cross sectonal (not temporal); () omtted varable bas does not appear to be a pressng problem; and () ncreasng regulaton exerts a negatve and statstcally sgnfcant mpact on the employment share of small frms. 3.2 Output Share of Small Frms To assess the fnal mplcaton of our model, namely that greater federal regulatons reduce the output share of small frms, we match gross recepts (.e. output) by ndustry and frm sze from the Census Bureau s SUSB dataset wth federal regulatons from RegData. Unfortunately, the monetary value of output s only avalable durng Economc Census years (.e. years endng n 2 or 7), whch lmts NAICS-based recept data to three tme perods: 2002, 2007 and 202. Usng ths short panel, we estmate a log-log fxed effects model very smlar to Baley and Thomas (207) 6 : (5) outputt t regt u t, where output t s the natural log of the share of total ndustry s output produced by small frms n year (t), α s the ndustry fxed effect, δ t s a perod fxed effect, reg t s the natural log of total federal regulatory restrctons applcable to ndustry n year t, and u t s a mean zero error term. The ndustry fxed effects capture any dfferences n average 6 Baley and Thomas (207) separately regress the natural log of frm brths and deaths onto an ndustryspecfc fxed effect, tme perod effect, and a one-year lag of the natural log of federal regulatons. Because our output panel contans only three tme perods whch are spaced fve years apart, we follow Chambers et al. (208) and regress our dependent varable on the natural log of contemporaneous regulaton. 0

12 output shares of small frms across ndustres, whle perod fxed effects capture common shocks to output shares across ndustres. As wth Equaton (4), the coeffcent on ndustry regulatons has an elastcty nterpretaton. The estmates of Equaton (5) are provded n Table 2. In our preferred specfcaton (see column of Table 2), the regulaton coeffcent s negatve and statstcally sgnfcant at the 5% level, mplyng that a ten percent ncrease n federal regulatons reduces the output share of small frms by nearly.5%. 7 As wth the employment share model, hgher regulatons appear to beneft large frms at the expense of ther smaller compettors. The goodness of ft of ths smple model s also qute hgh, explanng nearly 97% of the varaton n share of small frm output. To verfy the robustness of these results, we employ the same estmaton strategy used n Secton 3.. As t turns out, the results are smlar to those from the employment share model n Secton 3., namely: () most of the varaton n the output share panel s cross sectonal (not temporal); () our model does not suffer from any omtted varable bas; and () ncreasng regulaton has a negatve and statstcally sgnfcant mpact on the output share of small frms. In column 2, the perod effects, whch could be correlated wth changes n federal polcy emboded n regulatory statutes, are removed. The resultng regulaton coeffcent retans ts 5% level of statstcal sgnfcance and negatve sgn, declnng slghtly n magntude to Ths mples that a ten percent ncrease n ndustry-specfc federal regulatons reduces small frms share of output by just less than %. In column 3, we replace the perod fxed effects wth drect measures of the busness cycle. Although both busness cycle covarates are statstcally nsgnfcant, the regulaton coeffcent s dentcal n magntude to our preferred specfcaton n column. Ths strongly suggests that the perod effects are capturng busness cycle varatons; but gven the unchangng goodness of ft across columns to 3, the busness cycle plays no role n drvng changes n small frms output shares over tme. Fnally, column 4 elmnates both ndustry and tme perod fxed effects. The regulaton coeffcent s nearly dentcal to that of column 2, n whch the perod fxed effects are omtted. 7 Due to the lmted number of tme perods (3), we cannot estmate robust standard errors clustered by ndustry. Instead, we report ordnary standard errors.

13 4. Concluson Usng the two-sector dynamc general equlbrum macroeconomc model developed by Dhawan and Guo (200), we obtan three emprcally testable hypothess regardng the dsparate mpact of federal regulatons on small and large frms: ) a reducton n the total number of small frms, 2) the employment share of small frms shrnks, and 3) small frms share of total output declnes. Utlzng the relatvely new RegData database, the frst of these mplcatons has already been confrmed by Baley and Thomas (207) and Chambers et al. (208), whle ths paper s the frst (to our knowledge) to provde emprcal evdence n support of the latter two mplcatons. Specfcally, we fnd that a ten percent ncrease n federal regulatons reduces the employment share of small frms by nearly 0.7%, and an equally large ncrease n federal regulatons (0%) reduces the output share of small frms by nearly.5%. Although these effects may seem small on the margn, these mpacts are economcally sgnfcant vs-à -vs the sheer number of small busnesses n the U.S. In addton, each regulaton may operate lke tossng an ndvdually small pebble nto a runnng stream, the cumulatve long-run mpact of many pebbles may dam the rver. Therefore, when polcy makers consder draftng new government regulatons, they should exercse cauton and carefully wegh any estmated benefts aganst the costs borne dsproportonately by small busnesses. 2

14 References Baley, J. B., and D. W. Thomas (207), Regulatng Away Competton: The Effect of Regulaton on Entrepreneurshp and Employment, Journal of Regulatory Economcs 52, Chambers, D., McLaughln, P. A., and T. Rchards (208), Regulaton, Entrepreneurshp, and Frm Sze, Mercatus Workng Paper, Aprl 208. Cran, W. M., and N. V. Cran (204), The Cost of Federal Regulaton to the US economy, Manufacturng and Small Busness, Natonal Assocaton of Manufacturers, Washngton, D.C., September. Davs, S. J., Haltwanger, J. C., and S. Schuh (998), Job Creaton and Destructon, MIT Press, Cambrdge, MA. Dawson, J. W. and J. J. Seater (203), Federal Regulaton and Aggregate Economc Growth, Journal of Economc Growth 8, Dhawan, R. (996), Frm Sze, Fnancal Intermedaton and Busness Cycles, Commonwealth Publshers, New Delh, Inda. Dhawan, R. and J.-T. Guo (200), Declnng Share of Small Frms n U.S. Output: Causes and Consequences, Economc Inqury 39, Fegenbaum, A. and A. Karnan (99), Output Flexblty -- A Compettve Advantage For Small Frms, Strategc Management Journal, 2, 0-4. Goldschlag, N., and A. Tabarrok (208), Is Regulaton to Blame for the Declne n Amercan Entrepreneurshp? Economc Polcy 33, Ktchng, J., Hart, M. and N. Wlson (205), Burden or Beneft? Regulaton as a Dynamc Influence on Small Busness Performance, Internatonal Small Busness Journal 33, McLaughln, P. A., and O. Sherouse (207), QuantGov - A Polcy Analytcs Platform. Workng paper, Dec. 20, Mlls, D. and L. Schumann (985), Industry Structure wth Fluctuatng Demand, Amercan Economc Revew 75,

15 Table Employment Share Panel Estmates Varables () (2) (3) (4) Log Regulatons (lagged) *** 0.07*** *** 0.030*** (0.076) (0.0269) (0.0208) (0.0049) Unemployment *** (0.002) GDP Gap (0.0040) Industry Fxed Effects? Yes Yes Yes No Tme Perod Fxed Effects? Yes No No No Observatons 3,598 3,598 3,598 3,598 Goodness of Ft Notes: ) Dependent varable s the natural log of the share of total 5 dgt NAICS ndustry output produced by small frms n a gven year 2) Intercept ncluded by not reported 3) Whte robust standard errors clustered by ndustry n parenthess 4) ***, **, and * denote %, 5%, and 0% statstcal sgnfcance 4

16 Table 2 Output Share Panel Estmates Varables () (2) (3) (4) Log Regulatons 0.494** ** 0.494** *** (0.0675) (0.0443) (0.0675) (0.0277) Unemployment (0.0095) GDP Gap (0.0093) Industry Fxed Effects? Yes Yes Yes No Tme Perod Fxed Effects? Yes No No No Observatons Goodness of Ft Notes: ) Dependent varable s the natural log of the share of total 5 dgt NAICS ndustry output produced by small frms n a gven year 2) Intercept ncluded by not reported 3) Standard errors n parenthess 4) ***, **, and * denote %, 5%, and 0% statstcal sgnfcance 5