Evolutionary trends of industry variables

Transcription

1 Interntionl Journl of Industril Orgniztion 16 (1998) Evolutionry trends of industry vribles Rjshree Agrwl* Deprtment of Economics, University of Centrl Florid, P.O. Box Orlndo, FL , USA Abstrct The pper offers empiricl evidence on the evolutionry trends of industry vribles. It investigtes the time pth of ptents, product price nd quntity, nd number of firms from the first introduction of product to mturity of the mrket. The pper extends erlier work by Gort nd Klepper by expnding the dt on products through 1991, nd ugmenting it with new products nd new dt sources. The reexmintion of the ging ptterns revel the Gort nd Klepper results to be robust. The time trends follow the expected ptterns with remrkble consistency nd re strongly significnt for most products. New evidence on ptenting ctivity revel n eventul decline in technologicl ctivity over the lter yers of the product mrket life-cycle Elsevier Science B.V. Keywords: Evolution; Product life-cycle; Time-trends JEL clssifiction: O30; L0 1. Introduction Gort nd Klepper (198) investigted the life-cycle of product mrkets tht re born s result of mjor innovtion nd drew ttention to the empiricl regulrities observed in key industry vribles. Their study distinguished between five stges in the evolution of the product mrkets, nd compred the number of firms, prices, quntity nd ptents cross the five stges. In trcing the history of the diffusion of product innovtions, Gort nd Klepper introduced the time pth of events s being criticl determinnts of the ultimte structure of the product * Tel.: ; fx: ; e-mil: grwl@bus.ucf.edu / 98/ $ Elsevier Science B.V. All rights reserved. PII S (97)

2 51 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) mrkets. According to the Gort Klepper model, when mrket is first introduced, substntil informtion externl to the existing firms renders entry into the mrket ttrctive, firms enter rpidly to exploit their informtion, prices decline nd output increses. When product mrket is more estblished, the incumbent knowledge (type 1 informtion) exceeds the externl informtion (type informtion), hence net entry declines nd becomes negtive s inefficient firms re forced to exit. As product mrket evolves, the number of firms first increse nd then decrese, product price declines nd quntity increses. This evolutionry pproch hs spwned significnt reserch in the re. Nelson nd Winter (198); Jovnovic (198); Jovnovic nd Lch (1989); Hopenhyn (1994) nd Jovnovic nd McDonld (Jovnovic nd McDonld, 1994,b) hve developed theoreticl models explining the observed evolutionry phenomen, while Acs nd Audretsch (1991); Audretsch (1991) review cse studies nd present necdotl evidence. Given the importnce of the Gort Klepper findings to the subsequent literture, this pper extends the dt on the products through 1991, nd ugments it with some new products nd new dt sources to reexmine the ging ptterns of number of product mrkets. Rther thn using stges of the product life-cycle to discern between verge levels, the pper uses trend nlysis to cpture the reltionship between mrket ge nd the key industry vribles. The Gort Klepper ptterns re seen to be remrkbly robust, nd ptent sttistics provide dditionl evidence on the increse nd subsequent decline in the number of innovtions relevnt to the product mrket. Section provides description of the dt. Section 3 discusses the results of the empiricl nlysis nd Section 4 briefly discusses the economic forces tht re likely to generte the time trends, bsed on the theoreticl models developed to explin the phenomen. Finlly, Section 5 dels with the limittions of the present study nd venues of future explortion.. Dt Dt on the number of firms re gthered from the Thoms Register of Americn Mnufcturers. Thirty-one of the 33 product mrkets nlyzed in this 1 study re drwn from the pool of products nlyzed in the Gort Klepper study. 1 While the study drws from the sme pool of products s the Gort Klepper study, the dt re developed independently. Fifteen of the 46 products in the Gort Klepper study could not be used for new dt development for vrious resons. Some products, like nylon, telemeter, computers nd solr btteries hd breks in consistency either becuse the listing ws missing in the Thoms Register of Americn Mnufcturers, or due to substntil chnges in definition of product over the yers. Products like DDT nd cryogenic tnks were omitted since they were discontinued over the yers for which the nlysis ws extended (from 1973 to 1991). Other ctegories like streptomycin nd penicillin were discrded in fvor of broder product group ntibiotics. Finlly, few products were not included in the nlysis due to time limittions on the development of dt.

3 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) To incorporte lter dy innovtions, nd to mintin representtive smple of commercil nd industril products, two product mrkets contct lenses nd video cssette recorders tht gined prominence in the lst few decdes re dded. Technicl products (high tech) re distinguished from non-technicl products (low tech) bsed on Hdlock, Hecker nd Gnnon s (Hdlock et l., 1991) clssifiction of high technology 3-digit Stndrd Industril Clssifiction (SIC) industries. Appendix A gives list of the products in the study, the yer they were first introduced, nd the technologicl index. Ptent sttistics, first used by Schmookler (1966), re widely used s n indictor of innovtive ctivity nd proxy for innovtions. Their use, however, hs been criticized for four mjor resons. First, ptent clssifiction by the U.S. Ptent Office (Clssifiction Definitions U.S. Ptent Office) is bsed on technologicl functionl rther thn industry-of-use principle, which leds to problem in reclssifiction of ptents corresponding to industries bsed on SIC codes. Second, ptent counts fil to cpture the importnce nd vrying impct of innovtions on the technologicl environment. Third, industries vry in their propensity to ptent, due in prt to the existing trde-off between obtining exclusive rights grnted by ptent nd the loss of secrecy thereof. Finlly, s Griliches (1989) points out, ptent dt my not be constnt yrdstick mesure due to constrints on the number of ptent officers nd budgetry lloctions of the ptent office. In spite of ll the ssocited problems, ptent dt provide t lest crude proxy for innovtive ctivity. Hll, Griliches nd Husmn (Hll et l., 1986) show tht ptents re strongly nd positively correlted with Reserch nd Development (R&D) ctivity. In ddition, the first nd third problems bove re voided in this nlysis. Since time series nlysis of ptents within the seprte products is considered, the problem of industry differences in the propensity to ptent is voided. Further, the nlysis is bsed on nrrow product level definition of the mrket rther thn on the brod four or five digit SIC industry level. The reclssifiction of ptent clss/ subclss into ctegory corresponding to the product is hence gretly simplified nd conducted by referring to the Index of U.S. Ptent Clssifiction System; Mnul of Clssifiction. The nnul ptent dt for the products re compiled from the corresponding clss/ subclss definitions of the US Ptent Office for 31 of the 33 products in the study. Dt on product price nd quntity re reltively difficult to obtin t the product level, nd re successfully compiled on only 17 of the 33 products. The price nd quntity dt used in the Gort Klepper study is ugmented by The clss/subclss definition for ech product is vilble on request. Pints nd contct lenses re the two exceptions. For some products, initil redefining of the product ctegory ws necessry. For exmple, ntibiotics hd to be broken down into the different drug clsses which compose the group. Expert dvice nd consulttion of technicl dictionries were undertken where necessry. Also, the ptents grnted for photogrph records could not be distinguished from ptents grnted for compct discs product clerly superior substitute for phonogrph records introduced within the lst couple of decdes. For this reson, the ptent series is only considered for the first 6 yers of the life-cycle of phonogrph records.

4 514 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) informtion on six dditionl products. Informtion from trde journls, the Census of Mnufctures (Bureu of the Census) nd the Producer Price Index (Bureu of Lbor Sttistics) re used to construct s complete historicl dt set s possible. Appendix B gives the list of products for which price nd quntity dt were obtined, the yers for which dt re vilble nd the sources used. Producer price indices (PPI) re used where vilble, nd implicit prices clculted by dividing the vlue of shipments (sles) by the production (quntity sold) re used elsewhere. The price series for ech product is then deflted by the PPI for ll goods nd services to correct for infltion nd generl productivity chnges (economy wide rther thn product specific). 3. Empiricl nlysis The present study bstrcts wy from the brekdown of the product life-cycle in stges used in the Gort Klepper nlysis. Using nnul dt rther thn differences cross stges, the study summrizes the time trends of the industry vribles by regressing the key industry vribles on product mrket ge. To cpture the increse nd subsequent decline in the number of firms nd ptenting ctivity within product mrket, these two industry vribles re regressed on qudrtic function of product ge. The ptterns of growth in quntity nd price nd the rte of chnge of growth re summrized by regressing the log of quntity nd the log of price on qudrtic function of ge. Eqs. (1) (4) give the qudrtic 3 regressions on product mrket ge for ech of the four industry vribles. Ptent 5 1 Age Age 1 u (1) 1 3 Numfirm 5 d 1 d Age d Age 1 v () 1 3 log(price) 5 b b Age 1 b Age 1 w (3) 1 3 log(quntity) 5 x 1 x Age x Age 1 z (4) 1 3 where Age represents the ge of the product mrket, nd the signs of the 3 Gross entry nd exit underlying the trend in the number of firms hve recently been studied in the context of stges in the evolutionry cycle by Agrwl nd Gort (1996), who extend the Gort Klepper nlysis. They show tht the ptterns of gross entry nd exit re consistent with the time trends hypothesized for the number of firms, with gross entry peking erly in stge nd gross exit peking in stge 4.

5 Tble 1 Qudrtic regression of ptents on product ge Product Adjusted R Constnt ( p-vlue) Age ( p-vlue) Age ( p-vlue) Pek Technicl products Antibiotics (0.0001) 0.14 (0.0001) 0.00 (0.0001) 9 Betry guges (0.5058) 0.11 (0.01) (0.009) 16 Cthode ry tubes (0.0001) 0.03 (0.0) (0.47) 38 Electrocrdiogrphs (0.8745) (0.66) (0.05) no pek Gs turbines (0.0001) 0.1 (0.0001) 0.00 (0.0001) 17 Guided missiles (0.098) 0.14 (0.0001) (0.0003) 16 Gyroscopes (0.0001) 0.10 (0.0001) (0.0001) 5 Jet engines (0.0001) 0.13 (0.0001) 0.00 (0.0001) 15 Microfilm reders (0.0001) 0.06 (0.001) (0.04) 34 Nucler rectors (0.0001) 0.13 (0.0001) (0.000) 33 Outbord motors (0.0001) (0.74) (0.01) no pek Phonogrph records (0.0001) 0.04 (0.08) (0.003) 9 Photocopying mchines (0.0001) 0.09 (0.0001) (0.8) no pek Piezoelectric crystls (0.0001) 0.06 (0.67) (0.09) 48 Rdr ntenn ssemblies (0.0001) 0.01 (0.37) (0.33) no pek Rdition meters (0.0001) 0.09 (0.0001) (0.07) no pek Recording tpes (0.960) 0.15 (0.01) 0.00 (0.06) 35 Rocket engines (0.0001) 0.06 (0.05) (0.57) 9 Styrene (0.0001) 0.11 (0.0001) 0.00 (0.0001) 35 Video cssette recorders (0.0001) 0.16 (0.01) (0.30) 15 Non-technicl products Artificil christms trees (0.0067) 0.04 (0.11) (0.4) 37 Bll-point pens (0.0001) 0.01 (0.61) (0.5) 14 Combintion locks (0.0001) 0.08 (0.0001) (0.0001) 11 Electric blnkets (0.6651) 0.05 (0.0004) (0.03) 35 Electric shvers (0.070) 0.06 (0.) (0.6) 14 Freezers (0.0001) 0.08 (0.004) 0.00 (0.00) 17 Freon compressors (0.0001) 0.01 (0.4) (0.11) 3 Het pumps (0.0001) 0.01 (0.48) (0.55) 31 Oxygen tents (0.443) 0.01 (0.10) (0.03) no pek Polriscopes (0.770) 0.03 (0.06) (0.05) 43 Rdint heting bsebords (0.0190) (0.85) (0.67) no pek Source: Dt compiled from the US Ptent Office. Technicl nd non-technicl products bsed on Hdlock, Hecker nd Gnnon (Hdlock et l., 1991) clssifiction of 3-digit SIC industries s technologicl by rtio of R&D personnel to sles. Product ge t which the number of ptents re t the mximum. R. Agrwl / Int. J. Ind. Orgn. 16 (1998)

6 516 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) coefficients of ech of the time terms ensure the shpe of the trends (see Perry (1954); Ching (1974)). Tble 1 gives the time trend regression nlysis of ptents for 31 of the 33 products. The qudrtic specifiction of product ge for ptents is designed to cpture the incresing nd then declining trend in ptents. As seen in Tble 1, fifteen of the twenty technicl products hve the expected signs for the coefficients of product ge. The coefficients re significnt t the 0.01 level for eight products, while four more re significnt t the 0.1 level. The reltion between ptents nd product ge is little weker for non-technicl products. Six of the eleven products hve the expected signs, of which three re significnt t the 0.1 level. A reson for the weker reltion could be the much smller nnul rte of ptenting when 5 compred to the technicl products. As the results show, while only 55 percent of non-technicl products show the qudrtic trends, 75 percent of the technicl products exhibit qudrtic reltion of ptenting ctivity with time. Tble 1 lso shows the ge of the product t which the ptent dt reches its pek; in ll but five products the pek is reched well before the end of the smple period. For some of non-technicl products, the nnul count of ptents is no more thn three or four, nd no cler mode is present. The observed decline in the ptent trend is deprture from the results in the Gort Klepper nlysis. The erlier study showed the ptenting rte to increse over the stges nd could not cpture the pek nd subsequent decline in ptenting ctivity. Since the pek in ptents occurs in most products in the lter stges, the decline in ptenting ctivity could not be discerned by the erlier study which focused on differences over stges of men levels of ptenting rther thn the nnul trend of ptents. Tble presents the results of qudrtic regression of the number of firms on product ge. The results indicte tht 5 of the 33 products hve the hypothesized signs, showing n initil increse in the number of firms followed by decline. The lst column in Tble gives the product ge t which the number of firms pek, nd revels tht 6 products show distinct pek in the number of firms. The number of firms in electric shvers peks very erly in its product cycle, which is the reson tht the trend fils to cpture the phenomenon. A comprison of the peks observed in the trends for number of firms nd ptents is lso of interest. Among the twenty technicl products, eleven products show tht the pek in ptent dt occurs in the period of contrction of the number of firms, while three products experience the pek in ptenting nd number of 4 4 For ptents nd number of firms, positive coefficient of ge nd negtive coefficient of Age will yield the incresing nd then declining trend. The functionl forms for quntity nd price llow the growth rte nd its rte of chnge to hve the sme or different signs. The men growth rte cn be positive or negtive, nd increse or decrese with ge. 5 The verge nnul number of ptents in the technicl products ctegory is 57, s opposed to 3 in the non-technicl ctegory.

7 Tble Time trend of number of firms: qudrtic regression of number of firms on product ge Product Adjusted R Constnt ( p-vlue) Age ( p-vlue) Age ( p-vlue) Pek Antibiotics (0.0168) (0.0001) 0.04 (0.0001) 0 Artificil christms trees (0.170) (0.0001) (0.0001) 9 Bll-point pens (0.0104) 1.95 (0.0001) (0.0001) no pek Betry guges (0.1055) (0.0001) 0.04 (0.0001) 16 Cthode ry tubes (0.037) 1.88 (0.0001) 0.01 (0.0001) 6 Combintion locks (0.0001) 0.00 (0.0001) (0.0001) no pek Contct lenses (0.0001) 1.0 (0.0001) (0.0001) no pek Electric blnkets (0.1814) (0.0001) (0.0001) 48 Electric shvers (0.0001) (0.0001) 0.00 (0.0001) 5 Electrocrdiogrphs (0.993) 0.97 (0.0001) (0.0001) 4 Freezers (0.0001) 1.90 (0.007) (0.0001) 10 Freon compressors (0.0084) (0.0001) (0.0001) 5 Gs turbines (0.030) (0.0068) (0.0479) no pek Guided missiles (0.351) 1.56 (0.0001) (0.0001) 14 Gyroscopes (0.0047) (0.0001) (0.051) 53 Het pumps (0.1810) (0.038) 0.00 (0.0118) no pek Jet engines (0.045) (0.0001) 0.08 (0.0001) 18 Microfilm reders (0.306) (0.0001) (0.0106) no pek Nucler rectors (0.0073).767 (0.0001) (0.0001) 11 Outbord motors (0.0001) 0.40 (0.0001) (0.0005) 53 Oxygen tents (0.1037) (0.0001) 0.05 (0.0001) 3 Pints (0.0001) 7.7 (0.0001) (0.0001) 34 Phonogrph records (0.0036) (0.0001) (0.1138) 17 Photocopying mchines (0.5491) 1.56 (0.0001) 0.07 (0.0001) 6 Piezoelectric crystls (0.0001) (0.0001) 0.00 (0.0001) 4 Polriscopes (0.030) (0.0001) (0.0001) 31 Rdr ntenn ssemblies (0.087).80 (0.0001) (0.0001) 14 Rdint heting bsebords (0.007).157 (0.0001) (0.0001) 3 Rdition meters (0.0493) 1.17 (0.0001) 0.03 (0.0001) 18 Recording tpes (0.477) (0.0001) (0.0001) 35 Rocket engines (0.0001) (0.0001) (0.0001) 7 Styrene (0.01) 0.98 (0.0001) (0.0531) 44 Video cssette recorders (0.006) (0.8479) (0.000) no pek Source: Bsed on dt compiled from the Thoms Register of Americn Mnufcturers. Product ge t which the number of ptents re t the mximum. R. Agrwl / Int. J. Ind. Orgn. 16 (1998)

8 Tble 3 Time trend for price: qudrtic regression of log(price) on product ge Product Adjusted R Constnt ( p-vlue) Age ( p-vlue) Age ( p-vlue) Antibiotics (0.0001) 0.69 (0.0001) (0.0001) Bll-point pens (0.0001) (0.0001) (0.0001) Cthode ry tubes b (0.0659) (0.0974) (0.649) c (0.0001) 0.54 (0.0001) (0.0001) Electric blnkets (0.0001) (0.0001) (0.0001) Electric shvers (0.0061) (0.0001) (0.1161) Electrocrdiogrphs (0.69) (0.0939) (0.5864) Freezers (0.0001) (0.0001) (0.006) Freon compressors (0.0001) (0.0001) (0.0039) Het pumps (0.0001) (0.0001) (0.0551) Jet engines (0.0001) (0.0046) (0.0001) Outbord motors (0.0001) (0.0001) (0.0001) Phonogrph records (0.0001).0183 (0.000) (0.3103) Piezoelectric crystls (0.3067) (0.03) (0.6799) Rdint heting bsebords (0.0954) (0.007) (0.090) Recording tpes d (0.005).10 (0.095) (0.9749) e (0.0314) 0.43 (0.0005) (0.036) Styrene (.0001) (0.0001) (0.0001) Video cssette recorders (0.0001) 0.75 (0.0001) (0.0017) Refer to Appendix B for source of dt on prticulr products. Price nd quntity dt on these products were not used in the Gort Klepper study. bc The time trend of prices of cthode ry tubes is ffected by sudden increse in prices t the time corresponding to the introduction of color television. In order to b correct for this fctor, seprte regression nlysis is conducted for period c blck nd white CRT, nd predominntly color CRT. de The time trend of recording tpes is discontinuous t Age 7 due to chnge in the units of mesurement of the product (from mil. sq. ft. to mil. units). Hence d e seprte regression nlysis is considered for ech of the periods before nd fter Age R. Agrwl / Int. J. Ind. Orgn. 16 (1998)

9 Tble 4 Time trend for log(quntity): qudrtic regression of log(quntity) on product ge Product Adjusted R Constnt ( p-vlue) Age ( p-vlue) Age ( p-vlue) Antibiotics (0.0001) (0.0001) (0.0001) Bll-point pens (0.0001) 0.49 (.0001) (0.0001) Cthode ry tubes (0.1563) (0.0004) (0.0031) Electric blnkets (0.0001) (0.0001) (0.0001) Electric shvers (0.0019) (0.0001) (0.0001) Electrocrdiogrphs (0.646) 0.5 (0.0105) (0.013) Freezers (0.0001) (0.0001) (0.0001) Freon compressors (0.0004) (0.0001) (0.000) Het pumps (0.0001) (0.039) (0.5831) Jet engines (0.0001) (0.0001) (0.0001) Outbord motors (0.0146) 0.10 (0.0001) (0.0001) Phonogrph records (0.0001) (0.0001) (0.000) Piezoelectric crystls (0.48) (0.0030) (0.0078) Rdint heting bsebords (0.0001) (0.0001) (0.0001) Recording tpes b (0.0106) (0.0396) (0.005) c (0.6079) (0.0494) (0.9149) Styrene (0.000) (0.0001) (0.0001) Video cssette recorders (0.0039) (0.0001) (0.0001) Refer to Appendix B for source of dt on prticulr products. Price nd quntity dt on these products were not used in the Gort Klepper study. bc The time trend of recording tpes is discontinuous t Age 7 due to chnge in the units of mesurement of the product (from mil. sq. ft. to mil. units). Hence b c seprte regression nlysis is considered for ech of the periods before nd fter Age 7. R. Agrwl / Int. J. Ind. Orgn. 16 (1998)

10 50 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) firms t pproximtely the sme time. For non-technicl products the pek in ptenting occurs just s often before nd fter the pek in the number of firms. The mximum ptenting ctivity occurs, prticulrly for technicl products, predominntly in the period of contrction in the number of firms in the industry. This cn be interpreted in two wys. First, it my be reflection of the increse in competitive intensity experienced by firms in lter stges of the product mrket. Alterntively, it my simply indicte tht s mjor technologicl venues re exhusted, there re mny more minor enhncements tht re ptented. Further nlysis clerly requires better dt thn re currently vilble. Tble 3 gives the results of the qudrtic regression of log(price) on product mrket ge. The hypothesis tht price decreses (b 1b,0) t decresing rte 3 (b 3.0) over the product life-cycle is strongly supported by the trend nlysis. The high djusted R show the low vrince of price from the trend element. Price is 6 seen to decline significntly with product ge for ll products, nd for thirteen of the seventeen products, the decline in price occurs t significntly decresing rte. The initil rte of decline in price rnges from bout one percent nnully for electrocrdiogrphs to 6 percent decline seen in products like ntibiotics nd video cssette recorders. The men nnul decline rte in the first yer of the products in the smple is lmost 10 percent. After 30 yers, the men decline rte of the smple is less thn one percent, showing stbility in prices s the product 7 mrket mtures. Tble 4 displys the results of the qudrtic trend in log(quntity). The hypothesis tht quntity increses (x1x 3.0) t decresing rte (x 3,0) over the product life-cycle is lso strongly supported by the trend nlysis, with high djusted R for most products in the nlysis. The growth rte of quntity is significntly positive for ll the products, nd the decline in growth is significnt for ll but one product. The initil nnul growth in quntity rnges from modest seven percent increse for freezers to 90 percent increse for video cssette recorders. The men growth rte in quntity in the first yer for the products in the smple is lmost 8 percent, but 30 yers lter, the growth rte slows down to less 8 thn five percent. 4. Economic forces underlying the observed trends The ging ptterns in key industry vribles reveled by the bove trend nlysis re remrkbly consistent over wide vriety of products. In the 6 The ltter period of recording tpes is the only exception. 7 The coefficients of Age in electrocrdiogrphs, jet engines nd outbord motors revel prices rising t Age 30. Price increses in electrocrdiogrphs could be ttributble to rising helth cre costs nd significnt chnges in the technology of the product. 8 Products like bll point pens, jet engines nd rdint heting bsebords show modest decline in quntity fter 30 yers.

11 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) Gort Klepper nlysis, these ptterns re ttributed to chnge in informtion source in the industry from externl forces to incumbents s the product mrket ges. A similr explntion is given in the concepts of technologicl regimes, developed by Winter (1984) nd lter elborted by Audretsch (1991). They distinguish between routinized regime nd n entrepreneuril regime. A routinized regime, like the Gort nd Klepper type 1 informtion, ids incumbents nd is more likely to be observed in the mture period of the product mrket, while n entrepreneuril regime, more likely to be observed in n infnt product mrket, encourges entry. The theoreticl model developed by Jovnovic nd McDonld (1994b) focuses on the life-cycle of competitive industry nd distinguishes between firms tht employ different levels of technology in production. The nlysis is simplified by ssuming two mjor innovtions in the mrket. The first innovtion results in the introduction of the mrket, nd the subsequent innovtion is refinement of the technology. Low-tech firms (firms tht use the first type of technology) hve higher cost of production thn high-tech firms (firms tht dopt the refinement). As the industry evolves, the proportion of high-tech firms in the mrket increses. As result, industry output increses, nd price declines. The model shows n exit of low-tech firms (ctstrophic or grdul) following the introduction of the refinement in the shke-out period. Stylized fcts developed in the evolutionry literture by the bove reserchers nd Jovnovic nd Lch (1989), Dosi, Mrsilli, Orsenigo nd Slvtore, nd Hopenhyn (1994) suggest tht technologicl chnge is the chief economic force generting the observed ptterns. Firms enter new product mrket introduced due to mjor product innovtion, nd competition is initilly technology oriented. As the product mrket grows, there is n increse in innovtive ctivity, number of firms, nd product quntity. Product price declines due to increses in both competitiveness nd productivity. A shke-out of firms occurs when dominnt technology emerges, nd inefficient firms re forced to exit the mrket. In the lter stges of the mrket evolution, technologicl venues re exhusted nd product design is stndrdized, resulting in reltively constnt price nd quntity levels. 5. Conclusion, limittions nd venues of future reserch The present study extends the Gort Klepper nlysis to reexmine the ging ptterns of product mrket. The period of study is extended to 1991, n lmost twenty yer updte. The time period extension represents not only longer time series on ll products, but more importntly, seventeen of the products currently under nlysis hd not reched their finl stge in the Gort Klepper study. There is new informtion on two recent product mrkets, nd price nd quntity dt re constructed successfully for six more products (See Appendix B). Finlly, the nlysis of the ptent dt updtes the erlier work, nd by strtifying the products

12 5 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) ccording to technologicl ctivity, the pper exmines whether the hypothesized pttern holds more strongly for high-tech products thn low-tech products. The Gort Klepper results re seen to be remrkbly robust. The new evidence supports their findings over longer period of time, other product mrkets, nd for products introduced in lter spn of time. The longer period of nlysis (from ) revels no chnge in the min results from the Gort Klepper nlysis. For the seventeen products tht did not rech the finl stges in the erlier study nd the two products tht gined prominence in the lst two decdes, the current study replictes the Gort Klepper findings with remrkble consistency. While ptent dt cnnot cpture chnges in the importnce of innovtions over the product life-cycle, counts of innovtive ctivity still show n eventul decline in technologicl ctivity for most of the products in the nlyses, lending support to the stylized fct tht the rte of technologicl chnge slows down in the ltter stges of the life-cycle. The pttern is more strongly observed in product mrkets which re generlly more technicl in nture. While only 55 percent of nontechnicl products exhibited the hypothesized pttern, 75 percent of the technicl products in the smple showed n initil increse in ptenting followed by decline in lter yers. Similrly, the number of firms in the product mrket follows the hypothesized pttern, with 75 percent of the products in the study showing n increse, followed by decline in the number of firms. A comprison of the peks of the ptent nd the number of firms trends revel tht, prticulrly for technicl products, the pek in ptenting ctivity occurs fter the pek in the number of firms. Product price declines t declining rte over the life-cycle, nd product quntity grows t declining rte over time. The price nd quntity trends re remrkbly consistent for lmost ll the products in the study, with highly significnt coefficients for the trend terms nd low mounts of unexplined vrition. While the dt yield consistent time trends simultneously for brod rnge of products, they suffer from certin limittions. The study uses only count of the number of ptents nd number of firms. As stted before, ptents fil to cpture the degree of importnce of the innovtion, hence they cnnot be used to distinguish mjor nd minor innovtions. If the ptent informtion could hve been weighted by informtion on the possible impct of the prticulr innovtion, one could lso hve proceeded to test the hypothesis tht innovtions decline in importnce over time. However, to the extent tht the price trend shows decresing slope, one cn infer tht productivity increses decline over time. The price nd quntity series used re clerly limited insofr s they do not cpture the effect of qulity chnges. As the qulity of the product increses over the product life-cycle (n dditionl innovtion hs t lest non-decresing effect on qulity), the specifictions of the product chnge even though the inherent product definition is still mintined (current video cssette recorders re fr superior to the first introduction, however the product still serves the sme bsic use). The present series understte the decrese in prices nd increse in quntity

13 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) per unit qulity. This implies tht the bis cused by qulity, if removed, would yield stronger results thn those obtined in the study. The count of the number of firms implicitly ssumes tht the firms re homogenous nd hve the sme ttributes over time. It would be interesting to explore how the firm ttributes, such s size, product diversity, initil endowments nd lerning by doing chnge over the evolutionry cycle of the product. Acknowledgements I would like to cknowledge Michel Gort, Mrk Roberts nd two nonymous referees for their helpful comments. All remining errors re mine. Appendix A Products in study, yer of introduction, corresponding SIC code nd technologicl index Product nme Yer of commercil b SIC code Technologicl c introduction Antibiotics Artificil christms trees Bll-point pens Betry guges 1956 n.. 1 Cthode ry tubes Combintion locks 191 n.. 0 Contct lenses Electric blnkets Electric shvers Electrocrdiogrphs Freezers Freon compressors Gs turbines Guided missiles Gyroscopes Het pumps Jet engines Microfilm reders Nucler rectors Outbord motors Oxygen tents Pints Phonogrph records Photocopying mchines index

14 54 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) Piezoelectric crystls Polriscopes 198 n.. 0 Rdr ntenn ssemblies Rdint heting bsebords Rdition meters Recording tpes Rocket engines Styrene Video cssette recorders n..: not vilble. Bsed on the Thoms Register of Americn Mnufcturers. b SIC codes obtined from the Alphbeticl list of SIC codes, Census of Mnufctures 1987 Mnul (Bureu of the Census) nd from Predicsts Bsebook. c Technologicl index bsed on Hdlock, Hecker nd Gnnon (Hdlock et l., 1991) clssifiction of 3-digit SIC industries s technologicl by rtio of R&D personnel to sles. Appendix B Sources nd period for dt on price nd quntity by product Product Period Source(s) Antibiotics ITC Bll-point pens WIMA, BLS Cthode ry tubes EMDB, Predicsts, BLS Electric blnkets DM Electric shvers DM Electrocrdiogrphs Predicsts Freezers DM, BLS Freon compressors Predicsts Het pumps Predicsts Jet engines AIAA Outbord motors BC, Predicsts, BLS Phonogrph records BC, BLS Piezoelectric crystls Predicsts Rdint heting Predicsts bsebords Recording tpes Predicsts Styrene ITC Video cssette recorders DM Price nd quntity dt on these products were not used in the Gort Klepper study. AIAA: Aerospce Industries Assocition of Americ, Aerospce Fcts nd Figures. BC: Bureu of the Census, Census of Mnufctures nd Annul Survey of Mnufctures. BLS: Bureu of Lbor Sttistics, Producer Price Index. (Previous nme: Wholesle Price Index).

15 R. Agrwl / Int. J. Ind. Orgn. 16 (1998) DM: Delerscope Merchndising (Previous Nmes: Merchndising, Merchndising Week). ITC: U.S. Interntionl Trde Commission, Synthetic Orgnic Chemicls: Production nd Sles. Predicsts: Predicsts Bsebook WIMA: Writing Instruments Mnufctures Assocition, Mechnicl Hndwriting Instruments Industry References Acs, Z.J., Audretsch, D.B., Innovtion s Mens of Entry: An Overview, in: P.A. Geroski nd J. Schwlbch, eds., Entry nd Mrket Contestbility: An Interntionl Comprison. Blckwell Press, Cmbridge, pp. 43. Aerospce Industries Assocition of Americ, Aerospce Fcts nd Figures. Agrwl, R., Gort, M., The evolution of mrkets nd entry, exit nd survivl of firms. Review of Economics nd Sttistics 78 (3), Audretsch, D.B., New firm survivl nd the technologicl regime. Review of Economics nd Sttistics 73, Bureu of the Census, Census of Mnufctures, Government Printing Office; Bureu of the Census, Annul Survey of Mnufctures, Government Printing Office. Bureu of Lbor Sttistics, Producer Price Index. (Wholesle Price Index) Government Printing Office. Ching, A.C., Fundmentl methods of mthemticl economics. McGrw Hill, New York. Clssifiction Definitions U.S. Ptent Office. Delerscope Merchndising (North Americn Publishing Co., Phildelphi). Gort, M., Klepper, S., 198. Time pths in the diffusion of product innovtions. Economic Journl 9 (367), Griliches, Z., Ptents: Recent Trends nd Puzzles. NBER Working Pper No. 9. Hdlock, P., Hecker, D., Gnnon, J., High technology employment nother view. Monthly Lbor Review 114 (7), Hll, B.H., Griliches, Z., Husmn, J.A., Ptents nd R&D: Is there lg?. Interntionl Economic Review 7 (), Hopenhyn, Author, plese supply. Index to the U.S. Ptent Clssifiction System, U.S. Ptent Office. Jovnovic, B., 198. Selection nd evolution of industry. Econometric 50, Jovnovic, B., Lch, S., Entry, exit nd diffusion with lerning-by-doing. Americn Economic Review 79, Jovnovic, B., McDonld, G., Competitive diffusion. Journl of Politicl Economy 10(1), 4 5. Jovnovic, B., McDonld, G., 1994b. The life-cycle of competitive industry. Journl of Politicl Economy 10(), Mnul of Clssifiction, U.S. Ptent Office. Nelson, R.R., Winter, S., 198. An evolutionry theory of economic chnge. Hrvrd University Press, Cmbridge. Perry, J.H., Chemicl business hndbook. McGrw Hill, New York. Predicsts Bsebook. Predicsts Inc., Clevelnd. Schmookler, J., Invention nd economic growth. Hrvrd University Press, Cmbridge. Thoms Register of Americn Mnufcturers. Thoms Publishing Co, New York. U.S. Interntionl Trde Commission, Synthetic Orgnic Chemicls: Production nd Sles, Government Printing Office. Winter, S.G., Schumpeterin competition in lterntive technologicl regimes. Journl of Economic Behvior nd Orgniztion 5, Writing Instruments Mnufctures Assocition, Mechnicl Hndwriting Instruments Industry.