Food Marketing Policy Center

Transcription

1 Food Marketng Polcy Center Market Power and/or Effcency: An Applcaton to U.S. Food Processng by Rgoberto A. Lopez, Azzedne M. Azzam, and Carmen Lrón-España Food Marketng Polcy Center Research Report No. 60 July 2001 Research Report Seres Unversty of Connectcut Department of Agrcultural and Resource Economcs

2 Market Power and/or Effcency: An Applcaton to U.S. Food Processng by Rgoberto A. Lopez, Azzeddne M. Azzam, and Carmen Lrón-España Food Marketng Polcy Center Research Report No. 60 July 2001 Department of Agrcultural and Resource Economcs Unversty of Connectcut

3 Contents Tables... Acknowledgement... Afflaton... Preface Introducton The Model Emprcal Procedures Emprcal Results Conduct, Demand, and Economes of Sze Estmated Market Power and Effcency Effects Concludng Remarks... 5 References... 5 Appendx... 9 Tables Table 1. Selected Parameter Estmates, U.S. Food Processng Industres, Table 2. Impacts of Industral Concentraton on Olgopoly Power, Cost, and Output Prce... 8

4 Acknowledgement The authors are grateful to two anonymous referees as well as semnar partcpants at the Economcs Research Servce (USDA), Unversté Laval and Unversty of Basel. Fnancal support was provded by the USDA CRREES specal grant No , the Unversty of Connectcut and the Unversty of Nebraska. Ths s Scentfc Contrbuton No of the Storrs Agrcultural Experment Staton. A verson of ths report s forthcomng n Revew of Industral Organzaton. Afflaton The authors are professor at the Unversty of Connectcut, professor at the Unversty of Nebraska-Lncoln, and Ph.D. canddate at the Unversty of Connectcut, respectvely. Contact address: R. Lopez, Dept. of Agrcultural and Resource Economcs, Unversty of Connectcut, Storrs, CT , USA [tel , rlopez@canr.uconn.edu]. Preface Ths artcle separates olgopoly-power and cost-effcency effects of changes n ndustral concentraton and assesses ther mpact on output prces n 32 food-processng ndustres. Emprcal results ndcate that although concentraton nduces cost effcency n one-thrd of the ndustres, olgopoly-power effects ether domnate cost effcency or renforce neffcency, resultng n hgher output prces n most ndustres. The artcle also provdes fresh econometrc estmates of olgopoly power and economes of sze for the ndustres n queston. Key words: ndustral concentraton, economes of scale, ndustral organzaton, olgopoly power, food processng JEL Codes: L00, L11, L13, L66

5 Market Power and/or Effcency 1. Introducton In hs revew of the new emprcal ndustral organzaton (NEIO) lterature, Bresnahan (1989) concludes that although NEIO models are useful n measurng market power, they are not as useful n gudng polcy when market structure rather than conduct s the polcy target. Frst, he argues that the narrow focus of NEIO studes on sngle and often heavly concentrated ndustres s too lmted to be useful n mappng structure nto conduct and performance. Cross-ndustry studes, on the other hand, provde nformaton over a wder range of ndustres and, despte ther well-known problems, contnue to nfluence polcy (Salnger, 1990). Second, Bresnahan also argues that snce market power s mputed rather than observed n NEIO models, ts relatonshp to observable structural and behavoral varables n often unclear to polcy makers. Thus, to make NEIO fndngs more polcyrelevant, studes should consder a wder range of ndustres and ncorporate observable structural measures of nterest to polcy makers, such as ndustral concentraton. Ths artcle develops an NEIO model that formally ncorporates measures of ndustral concentraton and separates out the olgopoly-power from the costeffcency effects of concentraton on output prces. 1 The model s the olgopoly analogue of Azzam's (1997) olgopsony model, whch extends Appelbaum's (1982) model to formally nclude ndustral concentraton. The separaton of the two effects s not only of academc nterest but s also of publc polcy concern because the basc problem facng anttrust authortes s that of a tradeoff between effcency and market power (Wllamson, 1968; Perry, 1984; Ban and McFetrdge, 2000). That s, whether or not concentraton s n the publc nterest depends crtcally on whether or not the cost-effcency gans through concentraton offset the welfare losses from greater market power. The model s appled to 4-dgt SIC data on 32 U.S. manufacturng ndustres over the perod. The econometrc results provde fresh estmates of olgopoly 1 A large number of studes have tested the relatonshp between effcency proxes and prce-cost margns (e.g., Demsetz, 1973; Martn, 1988; Rosenbaum, 1994) whle others have tested more explctly the relatonshp between costs and concentraton (Peltzman, 1977; Dckson, 1994) or the ad hoc relatonshp between prce and market structure (e.g., Cotterll, 1986). A few recent studes have separated olgopsony power from cost effcency usng structural models (Azzam, 1997; Azzam and Schroeter, 1995) but studes of ths type focusng on olgopoly power are lackng. Food Marketng Polcy Center Research Report No. 60 Lopez, Azzam, and Lrón-España power and economes of sze n these ndustres and reveal that Cournot behavor s wdespread. The emprcal fndngs also ndcate that although costeffcency effects from concentraton are mportant n one-thrd of the ndustres, n nearly every case the olgopoly-power effects domnate or renforce cost neffcences, resultng n hgher output prces. The few exceptons where concentraton s benefcal to buyers are n the fat-and-ol sector, whch s characterzed by strong economes of sze and product homogenety. Fnally, ths analyss shows that NEIO models that brdge the gap between conduct and market structure can be useful for polcy makng decsons, especally those targetng ndustres based on effcency vs. the consumer's nterest. 2. The Model The startng pont s an ndustry of N frms producng a homogeneous good Q requrng factors x r for r = 1K,, k and facng a derved market demand curve Q = f ( p, z), (1) where p s output prce and z s a vector of demand shfters. Proft maxmzaton by the th frm yelds the supply relaton where s s C ( q p = (1 + φ) + η q, w, t), (2) = q Q s the th frm's market share, / η = ( dq / dp)(1/ Q) s the sem-elastcty of demand ( η < 0), φ n = d q / dq s the th frm's conectural varaton, C (.) s the cost functon, w s a vector of factor prces, and t s the state of technology. By Shephard's Lemma, the derved th demand for the r factor by the th frm s gven by x r C ( q, w, t) = for r = 1,2, K, k. (3) w 1

6 Market Power and/or Effcency Followng Olson and Sheh (1989), and Baffes and Vasavada (1989), the cost functon s assumed to take the modfed generalzed Leontef form C ( q, w) = q + q t γ w + q Food Marketng Polcy Center Research Report No α β w w 1/ 2 w 1/ 2, (4) where α, γ I, and β I are parameters. Multplyng through equatons (2) and (3) by s, usng (4), and summng across the ndustry yelds, respectvely, the ndustrywde analogue of the supply relaton H p = ( 1 + Φ) 1/ 2 1/ 2 + α w w + t η and factor demand where + 2HQ β w 1/ 2 γ w (5) X w r = α Q w + tγ + HQβ (6) for r=1,2, k, H = s the Herfndahl ndex, Φ s the ndustry (weghted) conectural varaton and X r = s total ndustry employment of the r th factor. By vrtue of the expresson for the sem-elastcty of demand, the demand functon (1) takes the sem-logarthmc form s x 2 r ln Q δ + ηp + δ z, (7) = 0 where η, δ 0 and δ are parameters. The frst term on the rght-hand sde of the supply relaton n (5) s the markup over margnal cost. Its magntude depends on the level of concentraton (H), the sem-elastcty of demand (η), and the type of market conduct (Φ ). If conduct s compettve, then Φ = 1 and the markup s zero. For Cournot, Φ = 0 and the markup s H / η. For Lopez, Azzam, and Lrón-España conduct that s less compettve than Cournot, 0 < Φ (1/ H ) 1 and the upper bound on the markup s 1/ η. However, for noncompettve conduct, concentraton affects both the markup and margnal cost. Note that the commonly-used conectural varaton elastcty (Appelbaum, 1982) can be defned as * Φ = (1 + Φ)H whch ranges between 0 and 1, the prce elastcty of demand s gven by η * = ηp, and the ndustry olgopoly power s defned by L = Φ * / η *. Followng Azzam (1997), Φ s treated as constant. 2 Thus, dfferentatng (5) wth respect to H yelds the decomposed effects of concentraton on output prce: dp (1 + Φ) = + 2Q w, dh η β (8) where the frst term on the rght-hand sde s the olgopoly-power effect and the second s the costeffcency effect (or the effect of a rse n concentraton on margnal cost). A measure for the cost elastcty wth respect to output s gven by the rato of ndustry margnal cost to average cost where and Α + 2HQB e cy =, (9) Α + HQB Α = α w w + t 1/ 2 1/ 2 B = β w. Note that e cy depcts economes of sze and s the nverse of the degree of returns to scale. If B=0, constant returns exst, and the only effect of rsng concentraton on prce s through olgopoly power. If B>0, dseconomes of scale exst, and a rse n concentraton rases prces through a rse n both olgopoly power and costs. When 2 Followng the work of Stgler (1964) and Stalhammar (1991), two addtonal specfcatons of Φ were tested. One was to let Φ = θ 0 + θ 1 ln H n order to allow ndustry conduct to vary wth concentraton. Another test was the ncluson of mports and exports. Although these extensons provded some nterestng nsghts n some cases, they deterorated the results of nterest. Gven the focus and scope of ths artcle, Φ was therefore treated as a constant. γ w 2

7 Market Power and/or Effcency economes of scale are present (B<0), the effect of a rse n concentraton on prce can be postve, negatve, or zero, dependng on whether the olgopoly-power effect s larger than, smaller than, or the same as the costeffcency effect Emprcal Procedures The model s operatonalzed wth data at the 4-dgt SIC (1987 defntons) level for the perod. The econometrc model s based on equatons (5), (6), and (7) depctng prcng behavor, nput demand equatons, and the output demand equaton. Although (5) s the man equaton of nterest, the nput and output demand equatons mpose stronger theoretcal restrctons and assst n dentfyng the correspondng parameters n the prcng equatons. The model assumes three varable nputs: materals (X K ), labor (X L ), and captal (X K ). 4 Thus, the estmatng model conssts of fve equatons: the prcng equaton, three nput demand equatons, and the output demand equaton. The latter s assumed to be a functon of output prce (P), ncome (y) and a trend varable (t), where prce and ncome are deflated by the consumer prce ndex (d). The endogenous varables are Q, P, X K, X L, and X K. The exogenous varables are W K, W L, W K, y, t (=1 for 1972 through 21 for 1992), d, and H. The parameters to be estmated are Φ,η, the α 's, the β 's, the γ 's, d 0, and the δ 0 and δ 's. 5 3 Note that these effects of concentraton are for a constant level of output and that hgher concentraton leads to lower, hgher, or no change n costs f there are ncreasng, decreasng, or constant returns to scale, respectvely. By fxng output, a rse n the Herfndahl ndex mples a change n the dstrbuton n output across frms, wth more output beng produced by the larger frms leadng to lower ndustry cost n the presence of economes of scale. Note that the econometrc model measures economes of scale on a market shareweghted ndustry cost functon and that techncal change (t) s assumed to affect the ndustry margnal cost ntercept, not the slope. 4 See Bartelsman and Gray (1996) for precse defntons of these varables. Followng the U.S. Department of Labor (1997), captal servces are assumed to be proportonal to captal stocks. The dervaton of the prce of captal s dscussed below. The materal nputs nclude raw materals (agrcultural, packagng, and certan servces) as well as energy. The latter accounted for less than 2% of varable costs n all cases. 5 The structural model contaned 17 coeffcents, whch we estmated wth 21 observatons. Snce the sample s small and the standard errors for nonlnear models are only approxmately correct for small samples, the statstcal Food Marketng Polcy Center Research Report No. 60 Lopez, Azzam, and Lrón-España The man data source for prces and quanttes of outputs and nputs was the onlne Natonal Bureau of Economc Research database of Bartelsman, Becker and Gray (2000) on U.S. manufacturng ndustres. Followng the U.S. Department of Labor (1997, p. 107), we defne the rental prce of captal n dollars per unt of real captal stock. Due to lack of data on the prce of captal at the 4-dgt SIC level, all ndustres are assumed to face the same rental prces but each to have dfferent levels of captal stock. Therefore, the rental prce of captal was computed by dvdng the cost of captal servces (provded electroncally by the Bureau of Labor Statstcs) dvded by total captal assets at the 2- dgt SIC level. Due to data lmtatons, we use an nstrumental varable for H for the years when t was not avalable (see Appendx for detals). Gven the endogenety of output quantty and the prce of output, the system of fve equatons s estmated wth non-lnear 3SLS usng the SHAZAM 7.0 software. The results for 32 food ndustres are presented below. 4. Emprcal Results 4.1 Conduct, Demand, and Economes of Sze Table 1 presents selected estmated parameters: Φ, η, L and ε cy. The null hypothess for conduct s Φ = 1 (compettve behavor) and for the elastcty of cost wth respect to output ε = 1 (constant returns to scale). An alternatve conduct hypothess tested s Cournot behavor ( Φ = 0), gven ts common use n emprcal analyss. As usual, one, two and three astersks next to the estmated coeffcents ndcate sgnfcance at the one, fve and 10 percent levels, respectvely. Note that the ndcated statstcal sgnfcance for Φ an ε n Table 1 are relatve to 1 and 1 (rather than the usual null hypothess of 0), respectvely. The results n Table 1 ndcate that the estmated Φ s ranged from for SIC 2062 (very close to the lower lmt of 1) to for SIC 2097 (well below ts maxmum lmt of 63 gven by H 1 1). Twenty of 32 ndustres (63%) had non-compettve ndustry conduct at the 5 percent level. It should be noted that the estmated Φ s for 20 of the ndustres also reected the null hypothess of Cournot behavor ( Φ = 0) at the 5% level. The derved conectural varaton elastctes ranged from for SIC 2062 to for The estmated olgopoly Lerner ndexes (evaluated at sgnfcance of the coeffcents should be nterpreted wth cauton. cy cy 3

8 Market Power and/or Effcency mean values of P) ranged from n SIC 2075 to n SIC Twenty-seven of the 32 ndustres (84%) have mark-ups statstcally dfferent from perfect competton at the 5% level. The number of ndustres wth non-compettve mark-ups s much larger than those ndcatng noncollusve conduct from only estmated Φ s. As shown by Nevo (2000) n the breakfast cereal ndustry, t s possble to have non-collusve behavor and yet have a strong degree of olgopoly power and level of mark-up. As a case n pont, the soybean ol ndustry (SIC 2075) was the most collusve, based on the conectural varaton elastctes; however, ths ndustry had the lowest mark-up as ndcated by the Lerner ndex, suggestng that n some cases concentraton s a way to survve when ndustres are operatng on small margns. Nonetheless, the results are consstent wth a myrad of studes that have found that most food processng ndustres are non-compettve. 6 In terms of economes of sze (calculated at mean values of data), the results reveal that approxmately 12 ndustres (38%) have sgnfcant economes of sze, 9 (28%) have sgnfcant dseconomes of sze, and the remanng 11 (34%) dd not reect the constant returns to sze hypothess at the 5% level. The economes of sze parameters, whch are crtcal for the evaluaton of cost effcency effects, appear to be reasonable and consstent wth prevous fndngs Estmated Market Power and Effcency Effects Table 2 presents the results based on equaton (8) for the separate effects of changes n the Herfndahl ndex on olgopoly-power, cost effcency, and output prce. These effects were calculated and tested at mean values of the data. In terms of olgopoly power, at the 10% level, the results ndcate that 26 of the 32 ndustres (81%) sgnfcantly ncrease olgopoly-power as concentraton ncreases. For the remanng sx ndustres, ncreases n 6 See, for example, Bhuyan and Lopez (1997). In partcular, the estmated Lerner ndex for the meatpackng ndustry (average=0.099) s between the ones estmated by Schroeter (1988) for beef (0.04) and by Schroeter and Azzam (1990) for beef and pork (0.14). Lkewse, the mark-ups estmated by Morrson-Paul (2001) correspond to Lerner ndexes between 0.05 and 0.20 between 1971 and The average estmate of Morrson-Paul (2001) for long run economes of sze n meatpackng s (cf n Table 1). The average degree of economes of sze n Table 1 s 0.997, whch s nearly CRS, somewhat above the estmated by Bhuyan and Lopez (1997). Food Marketng Polcy Center Research Report No. 60 Lopez, Azzam, and Lrón-España concentraton do not result n sgnfcant ncreases n olgopoly power. 8 In terms of cost effects, the results ndcate that 11 ndustres (34%) show sgnfcant gans (at the 10% level) n cost effcency wth concentraton, whle eght show a postve and sgnfcant ncrease n cost as concentraton rses. The remanng 13 ndustres do not show a sgnfcant mpact of concentraton on cost effcency. 9 The queston then remans how much of these potental cost savngs s beng passed on and how much s beng pocketed by the ndustres. Combnng olgopoly power and cost effects, concentraton results n sgnfcant decreases n output prces n only three ndustres (9.4%), wth 22 ndustres (68.7%) showng a net and sgnfcant ncrease n prce as concentraton rses, whle seven (21.9%) show nsgnfcant output prce effects at the 10% level. The notable cases where concentraton has a bengn effect on prces are n the fat and ol ndustres: 2074, 2076 and A number of ndustres show a strong trade-off between olgopoly-power and effcency effects, where both effects are statstcally sgnfcant and ther magntudes consderable. In ths category fall SICs 2011 (meatpackng), 2022 (cheese), 2026 (flud mlk), 2044 (rce mllng), 2074 (soybean ol), 2076 (vegetable ol), 2077 (anmal fats and ol), 2097 (manufactured ce), 2098 (macaron and spaghett), and food preparatons (2099). 5. Concludng Remarks The purpose of ths artcle s two-fold. Frst, t provdes a theory-based emprcal model that separates the olgopoly-power and cost-effcency effects of ndustral concentraton on output prces. The econometrc model adapts the olgopsony model of Azzam (1997) to the olgopoly-power case. A second contrbuton s that by applyng the model to tme-seres data across U.S. food processng ndustres, nformaton 8 The mpacts are partcularly large and sgnfcant for sausages and prepared meats (2013), poultry and egg processng (2015), flud mlk (2026), candy and confectonary (2064), prepared feeds (2048), anmal and marne fats and ols (2077), canned fruts and vegetables (2033), manufactured ce (2097), macaron and spaghett (2098), and food preparatons (2099). 9 In descendng order, the cost effcency effects that are statstcally sgnfcant at the 10% level are partcularly pronounced n the followng ndustres: anmal and marne fats (2077), flud mlk (2026), cottonseed ol (2074), manufactured ce (2097), food preparatons (2099), rce mllng (2044), vegetable ol (2076), and meat packng (2011). 4

9 Market Power and/or Effcency s provded on the potental mpact of further concentraton on olgopoly power and cost effcency wthn each ndustry and across ndustres, whch can be useful to polcy makers concerned wth market structure. From the emprcal results, we see some systematc patterns of the mpact of ndustral concentraton n the U.S. food ndustres. Specfcally, we fnd that further ncreases n concentraton would: (1) sgnfcantly ncrease olgopoly power; (2) result n cost effcency n one-thrd of the ndustres; and (3) ncrease output prce n nearly every case. These patterns rase some nterestng questons. What s and what determnes, for example, the crtcal level of concentraton beyond whch concentraton nduces net neffcency? There are other ssues that we do not address here but that are worthy of further attenton. Among these are the role of foregn trade and possble smultaneous olgopsony effects on factor prces n certan markets. In spte of the pattern of concentraton, there are strong dfferences among ndustres. It s tmely and relevant to examne the sources and consequences of these dfferences more fully. Extendng the analyss to ncorporate any of these ssues or applyng t to other ndustres wll provde further nsghts nto the market power and cost-effcency effects of changes n concentraton. Fnally, ths artcle shows that NEIO analyses can be made more polcy-relevant f they cover market structure and a wder range of ndustres, as suggested by Bresnahan (1989). References Appelbaum, E The Estmaton of the Degree of Olgopoly Power. Journal of Econometrcs 19: Azzam, A.M Measurng Market Power and Cost- Effcency Effects of Industral Concentraton. Journal of Industral Economcs 45: Azzam, A. and J. Schroeter Tradeoffs Between Olgopsony Power and Cost-Effcency from Horzontal Consoldaton: An Example from Beef Packng. Amercan Journal of Agrcultural Economcs 77: Baffes, J. and U. Vasavada On the Choce of Functonal Forms n Agrcultural Producton Analyss. Appled Economcs 21: Bartelsman, E.J. and W. Gray The NBER Manufacturng Productvty Database. Natonal Bureau of Economc Research, Techncal Workng Paper 205. Bartelsman, E.J., R.A. Becker, and W.B. Gray NBER- CES Manufacturng Industry Database, ( Bhuyan, S. and R.A. Lopez Olgopoly Power n the Food and Tobacco Industres. Amercan Journal of Agrcultural Economcs 79: Lopez, Azzam, and Lrón-España Ban, L. and D.G. McFetrdge The Effcences Defense n Merger Cases: Implcatons of Alternatve Standards. Canadan Journal of Economcs 33: Bresnahan, T.F Emprcal Studes of Industres wth Market Power. In Handbook of Industral Organzaton, Vol. 2, R. Schmalensee and R. Wllg, eds., North- Holland, Amsterdam. Chow, G.C. and A.I. Ln Best Lnear Unbased Interpolaton, Dstrbuton, and Extrapolaton of Tme Seres by Related Seres. Revew of Economcs and Statstcs 53: Cotterll, R.W Market Power n the Retal Food Industry: Evdence from Vermont. Revew of Economcs and Statstcs 68: Demsetz, H Industry Structure, Market Rvalry, and Publc Polcy. Journal of Law and Economcs 16:1-9. Dckson, V Aggregate Industry Cost Functons and the Herfndahl Index. Southern Journal of Economcs 61: Gsser, M 'Welfare Implcatons of Olgopoly n U.S. Food Manufacturng. Amercan Journal of Agrcultural Economcs 63: Golan, A., G. Judge, and J.M. Perloff Estmatng the Sze Dstrbuton of Frms Usng Government Summary Statstcs. Journal of Industral Economcs 44: Martn, S Market Power and/or Effcency? Revew of Economcs and Statstcs 60: Morrson-Paul, C.J Cost Economes and Market Power: The Case of the U.S. Meat Packng Industry. Revew of Economcs and Statstcs 83, n press. Nevo, A Measurng Market Power n the Ready-to-Eat Cereal Industry. Econometrca 69(2): Olson, D.O. and Y.N. Sheh Estmatng Functonal Forms n Cost Prces. European Economc Revew 33: Peltzman, S The Gans and Losses from Industral Concentraton. Journal of Law and Economcs 20: Perry, M.K Scale Economes, Imperfect Competton, and Publc Polcy. Journal of Industral Economcs 32: Rosenbaum, D.I Effcency v. Colluson: Evdence Cast n Cement. Revew of Industral Organzaton 9: Salnger, M.A The Concentraton-Margn Relatonshp Revsted. Brookngs Papers on Economc Actvty: Mcroeconomcs. Washngton, D.C.: Brookngs Insttuton. Schroeter, J.R Estmatng the Degree of Market Power n the Beef Packng Industry. Revew of Economcs and Statstcs 70: Schroeter, J.R. and A. M. Azzam Measurng Market Power n Mult-Product Olgopoles: The U.S. Meat Industry. Appled Economcs 22: Stalhammar, N.O Domestc Market Power and Foregn Trade. Internatonal Journal of Industral Organzaton 9: Stgler, G A Theory of Olgopoly. Journal of Poltcal Economy 72: Food Marketng Polcy Center Research Report No. 60 5

10 Market Power and/or Effcency Lopez, Azzam, and Lrón-España U.S. Department of Labor BLS Handbook of Methods, Bureau of Labor Statstcs, Bulletn Wllamson, O Economes as an Anttrust Defense: The Welfare Trade-Offs. Amercan Economc Revew 58: Food Marketng Polcy Center Research Report No. 60 6

11 Market Power and/or Effcency Lopez, Azzam, and Lrón-España Table 1: Selected Parameter Estmates, U.S. Food Processng Industres, SIC Industry Φ s.e. η s.e. L s.e. ε cy s.e Meat Packng *** *** *** *** Saus. & Prep. Meats ** *** Poultry & Egg Proc *** Cheese *** Dry/Cond.& Ev.Mlk *** *** Ice Cream & Fr. Desserts *** ** Flud Mlk ** *** *** Canned Specaltes *** *** *** *** Canned Frut & Veg ** *** Dred Frut & Veg *** *** *** Pckles, Sauces, etc *** *** Cereal Breakfast Foods *** *** ** *** Rce Mllng *** *** ** Prep. Flour & Doughs *** * *** Wet Corn Mllng *** *** ** Prep. Feeds * *** *** Cane Sugar *** *** ** Cane Sugar Refnng *** Beet Sugar *** *** *** * Candy & Conf. Prods *** *** Chocolate & Cocoa *** *** Cottonseed Ol Mll *** ** Soybean Ol Mll *** Vegetable Ol Mll *** *** *** *** An./Mar. Fats & Ols *** *** *** Malt Beverages *** *** *** *** Wnes & Brandy *** Extracts & Syrups *** *** *** *** Roasted Coffee *** *** Manuf. Ice *** *** *** *** Macaron & Spaghett *** *** *** Food Preparatons *** *** *** * Notes: Levels of statstcal sgnfcance are represented by *(10%), **(5%) and ***(1%). The standard errors (s.e.) are ndcated next to the estmated coeffcents. The null hypothess for Φ s -1 (perfect competton), whle the null hypothess for ε cy s 1 (CRS). These results are based on the ont estmaton of equatons (5), (6) and (7). Food Marketng Polcy Center Research Report No. 60 7

12 Market Power and/or Effcency Lopez, Azzam, and Lrón-España Table 2: Impacts of Industral Concentraton on Olgopoly Power, Cost and Output Prce. Impact of H on Olgopoly Cost Output SIC Industry Power s.e. Effcency s.e. Prce s.e Meat Packng *** *** Saus. & Prep. Meats *** *** Poultry & Egg Proc *** *** Cheese *** *** Dry Cond. & Evap. Mlk *** *** Ice Cream & Fr. Desserts *** ** *** Flud Mlk *** *** *** Canned Specaltes *** *** *** Canned Frut & Veg *** *** *** Dred Frut & Veg *** *** *** Pckles, Sauces, etc *** *** *** Cereal Breakfast Foods *** *** ** Rce Mllng ** Prep. Flour & Doughs *** Wet Corn Mllng ** *** Prep. Feeds *** *** Cookes & Crackers * *** *** Cane Sugar ** Can Sugar Refnng Beet Sugar *** * Candy & Conf. Prods *** *** Chocolate & Cocoa *** *** Cottonseed Ol Mll *** *** *** Soybean Ol Mll Vegetable Ol Mll ** *** * An./Mar. Fats & Ols *** *** ** Malt Beverages *** *** *** Wnes & Brandy *** *** Fd. Extracts & Syrups *** *** *** Roasted Coffee ** Manuf. Ice *** *** *** Macaron & Spaghett *** *** *** Food Preparatons *** * *** Notes: Levels of statstcal sgnfcance are represented by * (10%), ** (5%) and *** (1%). The standard errors (s.e.) are ndcated next to the estmated coeffcents. These results are based on equaton (8). Food Marketng Polcy Center Research Report No. 60 8

13 Market Power and/or Effcency Lopez, Azzam, and Lrón-España Appendx: Herfndahl Indexes The Herfndahl-Hrschmann ndex has only been publshed for 1982, 1987 and Addtonally, the partal ndustral concentraton ratos (CR4, CR8, CR20, and CR50) have been publshed for (then Census years ncludng) 1972 and Gven ths paucty of data, H had to be estmated for each food ndustry for much of the sample perod. Ths process nvolved two steps: (1) the estmaton of the 1972 and 1977 H ndexes from partal concentraton ratos; (2) nterpolaton of the H ndexes for the nter-census years. The frst step nvolved the applcaton of entropy to estmate the market shares of the top 50 frms. Usng the technque presented by Golan, Judge, and Perloff (1996), the market shares were forecasted for all frms encompassed n CR4, CR8, CR20 and CR50. It turns out that the maxmum entropy soluton for the sze dstrbuton of frms yelds equal market shares of frms wthn ntervals of these concentraton ratos. After forecastng the market share of each frm n 1972 and 1977, we restrcted the estmated dstrbuton of market shares to a thrd-degree polynomal functon n order to estmate the most probable market share for each frm. The average R-square of the estmated polynomal functons was 95%. From the ndvdual market share, we then estmated the H ndex as the sum of the squares of market shares resultng from the polynomal dstrbuton. The second step nvolved regressng the Census-year H ndexes on a set of nstrumental varables that were avalable for the entre sample perod (Chow and Ln, 1971) such as total number of employees, sales, payroll, and value of shpments per employee for whch a complete tme-seres data were avalable. These regressons yelded an average R-square of 85%. Then the H ndexes were estmated wth the predcted values from these regressons. The forecasted values were qute smooth n relaton to the H ndexes for the Census years. The use of splne functons and exponental smoothng (usng the CurveExpert software) dd not sgnfcantly altered the results. Fnally, t should be noted that foodprocessng ndustres for whch the 1982 Herfndahl ndexes were not avalable (due to reclassfcaton of the SIC codes n 1987) were excluded from the sample. Food Marketng Polcy Center Research Report No. 60 9

14 FOOD MARKETING POLICY CENTER RESEARCH REPORT SERIES Ths seres ncludes fnal reports for contract research conducted by Polcy Center Staff. The seres also contans research drecton and polcy analyss papers. Some of these reports have been commssoned by the Center and are authored by especally qualfed ndvduals from other nsttutons. (A lst of prevous reports n the seres s gven on the nsde back cover.) Other publcatons dstrbuted by the Polcy Center are the Workng Paper Seres, Journal Reprnt Seres for Regonal Research Proect NE-165: Prvate Strateges, Publc Polces, and Food System Performance, and the Food Marketng Issue Paper Seres. Food Marketng Polcy Center staff contrbute to these seres. Indvduals may receve a lst of publcatons n these seres and paper copes of older Research Reports are avalable for $20.00 each, $5.00 for students. Call or mal your request at the number or address below. Please make all checks payable to the Unversty of Connectcut. Research Reports can be downloaded free of charge from our webste gven below. Food Marketng Polcy Center 1376 Storrs Road, U-21 Unversty of Connectcut Storrs, CT Tel: (860) FAX: (860) emal: fmpc@uconn.edu