Productivity Growth in the Canadian Broadcasting and Telecommunications Industry: Evidence from Micro Data

Transcription

1 Catalogue no. 11F0027M No. 089 ISSN ISBN Research Paper Economc Analyss (EA) Research Paper Seres Productvty Growth n the Canadan Broadcastng and Telecommuncatons Industry: Evdence from Mcro Data by Wulong Gu and Améle Lafrance Economc Analyss Dvson

2 How to obtan more nformaton For nformaton about ths product or the wde range of servces and data avalable from Statstcs Canada, vst our webste, You can also contact us by emal at telephone, from Monday to Frday, 8:30 a.m. to 4:30 p.m., at the followng toll-free numbers: Statstcal Informaton Servce Natonal telecommuncatons devce for the hearng mpared Fax lne Depostory Servces Program Inqures lne Fax lne To access ths product Ths product, Catalogue no. 11F0027M, s avalable free n electronc format. To obtan a sngle ssue, vst our webste, and browse by Key resource > Publcatons. Standards of servce to the publc Statstcs Canada s commtted to servng ts clents n a prompt, relable and courteous manner. To ths end, Statstcs Canada has developed standards of servce that ts employees observe. To obtan a copy of these servce standards, please contact Statstcs Canada toll-free at The servce standards are also publshed on under About us > The agency > Provdng servces to Canadans. Publshed by authorty of the Mnster responsble for Statstcs Canada Mnster of Industry, 2014 All rghts reserved. Use of ths publcaton s governed by the Statstcs Canada Open Lcence Agreement ( statcan.gc.ca/reference/copyrght-drot-auteur-eng.htm). Cette publcaton est auss dsponble en franças. Note of apprecaton Canada owes the success of ts statstcal system to a long standng partnershp between Statstcs Canada, the ctzens of Canada, ts busnesses, governments and other nsttutons. Accurate and tmely statstcal nformaton could not be produced wthout ther contnued co operaton and goodwll. Standard symbols The followng symbols are used n Statstcs Canada publcatons:. not avalable for any reference perod.. not avalable for a specfc reference perod... not applcable 0 true zero or a value rounded to zero 0 s value rounded to 0 (zero) where there s a meanngful dstncton between true zero and the value that was rounded p prelmnary r revsed x suppressed to meet the confdentalty requrements of the Statstcs Act E use wth cauton F too unrelable to be publshed * sgnfcantly dfferent from reference category (p < 0.05)

3 Productvty Growth n the Canadan Broadcastng and Telecommuncatons Industry: Evdence from Mcro Data by Wulong Gu and Améle Lafrance Economc Analyss Dvson, Statstcs Canada 11F0027M No. 089 ISSN ISBN February 2014 Authors names are lsted alphabetcally. Economc Analyss Research Paper Seres The Economc Analyss Research Paper Seres publshes research conducted by the staff of Natonal Accounts and Analytcal Studes, vstng Fellows, and academc assocates. The Seres s meant to stmulate dscusson on a range of topcs, ncludng the mpact of the New Economy, productvty, frm proftablty and growth, technology use, deprecaton functons, satellte accounts, savngs, leasng, hedonc estmatons, dversfcaton, nvestment, and purchasng power. The authors welcome readers comments and suggestons. The prmary dstrbuton medum for the papers s the Internet. These papers can be accessed for free at All papers n the Seres undergo nsttutonal and peer revew to ensure that they conform to Statstcs Canada's mandate as a government statstcal agency and adhere to generally accepted professonal standards. The papers n the Seres often contan results of multvarate analyss and other statstcal technques. It should be recognzed that the estmates derved from these analyses are subject to uncertanty. The level of uncertanty depends on several factors: the functonal form used n the multvarate analyss; the econometrc technque employed; the approprateness of the statstcal assumptons embedded n the model or technque; the comprehensveness of the varables ncluded n the analyss; and the accuracy of the data that are utlzed. The peer revew process s meant to mnmze problems n these areas. Publcatons Revew Commttee Analytcal Studes Branch, Statstcs Canada 18th Floor, R.H. Coats Buldng Ottawa, Ontaro K1A 0T6 Acknowledgements We thank John Baldwn, Isabelle Amano, Erwn Dewert, Perre Therren and semnar partcpants at Statstcs Canada for ther helpful comments.

4 Table of contents Abstract... 5 Executve summary Introducton An overvew of the Canadan broadcastng and telecommuncatons sectors Methodology Producton possblty fronter approach Drect aggregaton across frms Data Emprcal results Results from the producton possblty fronter approach Results from drect aggregaton across frms Conclusons References Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

5 Abstract Ths study examnes the dynamcs of the rapd productvty growth that occurred n the Canadan broadcastng and telecommuncatons ndustry snce the md-1980s. Growth wthn frms was the man contrbutor to aggregate productvty growth. Both techncal progress and scale economes contrbuted to aggregate productvty growth. Durng the 1984-to-1998 perod, the compettve process assocated wth frm entry and ext made an mportant contrbuton to aggregate labour productvty growth. Durng the 2000-to-2008 perod, frm entry and ext made lttle contrbuton to aggregate labour and multfactor productvty growth. Growth n aggregate labour productvty fell between these two perods, manly because of a declne n the contrbuton of captal ntensty, especally that from entry and ext. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

6 Executve summary The Canadan broadcastng and telecommuncatons ndustry experenced strong output growth after the md-1980s. From 1984 to 2008, real gross output of ths sector ncreased at an annual rate of 5.5%. The ndustry also had among the most rapd labour and multfactor productvty (MFP) growth rates durng that perod. Ths paper examnes two aspects of productvty growth n Canada s broadcastng and telecommuncatons ndustry. The frst s the extent to whch aggregate MFP growth n the sector came from scale economes as opposed to techncal progress. The second s the extent to whch aggregate labour productvty growth and MFP growth came from wthn-frm growth, and from the effect of reallocaton due to frm entry and ext and wthn ncumbents the dynamc forces assocated wth compettve change. Based on data from Statstcs Canada s T2-LEAP longtudnal database, the analyss uncovers evdence of ncreasng returns to scale n broadcastng and telecommuncatons snce Techncal progress contrbuted between 1.3 and 1.4 percentage ponts to overall growth n aggregate MFP n the 1984-to-1998 and 2000-to-2008 perods, but scale economes were also mportant, accountng for about 0.5 percentage ponts, or 30% to 40%, of overall MFP growth n those two perods. Growth wthn contnung frms was the most mportant factor n aggregate labour productvty growth and MFP growth. Frm entry and ext made an mportant contrbuton to aggregate labour productvty growth before 2000, but not thereafter. Durng the 1984-to-1998 perod, when entrants had much hgher labour productvty levels than dd exts, frm entry and ext accounted for 1.2 percentage ponts, or about a quarter, of aggregate labour productvty growth. Durng the 2000-to-2008 perod, frm entry and ext contrbuted 0.2 percentage ponts. Annual aggregate MFP growth rose from 1.4% n the 1984-to-1998 perod to 1.9% n the 2000-to perod. The ncrease after 2000 was attrbutable to techncal progress wthn frms, capacty utlzaton, between-frm reallocaton, and net entry, each of whch made a small but postve contrbuton. Despte the ncrease n annual aggregate MFP growth after 2000, aggregate labour productvty growth slowed, manly because of a decrease n the contrbuton of captal deepenng, especally from frm entry and ext. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

7 1 Introducton The Canadan broadcastng and telecommuncatons ndustry experenced rapd output growth after the md-1980s. From 1984 to 2008, real gross output n ths ndustry ncreased at an average annual rate of 5.5%. As well, labour productvty (gross output per hour worked) rose by 3.9% per year, and multfactor productvty (MFP) by 1.2% per year 1 rates that were among the hghest of Canadan ndustres. However, between the 1984-to-1998 and 2000-to-2008 perods, annual labour productvty growth n the ndustry slowed from 4.9% to 2.9%. As well, productvty growth n Canadan broadcastng and telecommuncatons has been low compared wth the Unted States (Baldwn and Gu, 2008). Snce 2000, the ndustry has been one of the man contrbutors to lower overall productvty growth n the Canadan busness sector relatve to the U.S. The fnal report from the Telecommuncatons Polcy Revew Panel concluded that Canada s telecommuncatons ndustry s losng ts compettve edge, partcularly n the two most mportant and fastest-growng sectors: wreless and broadband servces (Snclar et al., 2006). Canada has the hghest wreless voce and data rates among the OECD countres (L and Nna-Moses, 2010), and s laggng behnd Japan, South Korea and the U.S. n rollng out fber optc cable and n the development of next-generaton networks (Snclar et al., 2006). The lack of competton and the small scale of the ndustry n Canada have been cted as major factors n the lower productvty growth. Canada restrcts foregn control of telecommuncatons provders (Snclar et al., 2006), and the relatvely small market compared wth the U.S. may have a negatve effect on the productvty of Canadan telecommuncatons, an ndustry where scale economes are mportant. Ths paper quantfes the sources of aggregate productvty growth n Canadan broadcastng and telecommuncatons snce the md-1980s. Two questons are posed. Frst, to what extent dd aggregate MFP growth n the ndustry come from scale economes rather than techncal progress? Second, to what extent dd aggregate labour productvty growth and MFP growth come from growth wthn frms and from reallocaton due to frm entry and ext and shfts n share across ncumbents the dynamc forces assocated wth compettve change. Measures of labour productvty and MFP are examned. Labour productvty ndcates how effcently labour s transformed n the producton process. It s relevant because of ts close connecton wth growth n real wage rates. MFP s a more comprehensve weghted average of both labour and captal productvty. It s calculated as the dfference between output growth and the growth n output that would be expected from the applcaton of labour and captal nputs based on assumptons about producton technology that s, how much output addtonal unts of labour and captal mght be expected to have yelded. MFP s relevant because t s nterpreted as capturng all unmeasured factors, ncludng dsemboded techncal progress. More mportantly, labour productvty growth can be decomposed nto MFP growth and a term nvolvng the growth n captal/labour ntensty or captal deepenng. Thus, underlyng changes n the economy that affect labour productvty wll be found n those that affect MFP growth and n those that affect changes n captal deepenng. The broadcastng and telecommuncatons ndustry n Canada conssts of two sub-sectors: broadcastng (sub-sector 515 accordng to North Amercan Industry classfcaton System (NAICS), 2007) and telecommuncatons (NAICS 517). Telecommuncatons s domnant, accountng for more than 90% of total sales n the ndustry snce However, because of 1. Statstcs Canada Productvty Program, CANSIM table The share s calculated usng the data from T2-LEAP. The data from the nput/output tables showed a smlar share. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

8 growng ntegraton of the two sub-sectors, t s dffcult to classfy frms nto separate broadcastng or telecommuncatons categores. Therefore, ths paper examnes the combned broadcastng and telecommuncatons ndustry. 3 Gven that the ndustry s characterzed by ncreasng returns to scale and mperfect competton, the growth accountng framework that was developed under the assumpton of perfect competton and constant returns to scale s extended here to take those features nto account (Solow 1957; Jorgenson and Grlches, 1967; Dewert 1976). The extended framework was developed by Denny et al. (1981), Dewert (1991), Hall (1988, 1990), Basu and Fernald (2001, 2002), Petrn and Levnsohn (2010), and Dewert et al. (2011). Baldwn et al. (2012) adapted that framework for use wth frm-level data and derved a method of decomposton that shows the extent to whch the growth n aggregate labour productvty comes from a wthn-frm growth effect and from the effect of between-frm reallocaton among ncumbents and between enterng and extng frms. The wthn-frm growth component can be further decomposed nto the effect of captal and ntermedate nput deepenng, technologcal progress, scale economes, and nput utlzaton. The between-frm component reflects the effect of reallocaton of nputs and outputs across frms on aggregate captal and ntermedate nput deepenng and aggregate MFP growth. Baldwn et al. (2012) also decomposed aggregate MFP growth nto a wthn-frm growth effect that ncludes technologcal progress, scale economes and nput utlzaton, and a between-frm reallocaton effect that captures the effect of nput reallocatons across frms on aggregate MFP growth. Both decompostons are used here to examne whether the mportance of the underlyng components dffers for each productvty measure. The paper s organzed as follows. Secton 2 provdes background nformaton about the regulatory framework of Canadan broadcastng and telecommuncatons. It also summarzes prevous studes of productvty growth n that sector. Secton 3, borrowng heavly from Baldwn et al. (2012), descrbes the method used to examne productvty growth n the ndustry. Secton 4 presents the data, and Secton 5 descrbes the emprcal evdence. Secton 6 concludes. 2 An overvew of the Canadan broadcastng and telecommuncatons sectors The telecommuncatons ndustry conssts of wred or wrelne (the largest segment) and wreless telecommuncatons carrers, as well as satellte telecommuncatons. The broadcastng ndustry comprses rado and televson broadcastng, and pay and specalty televson. The penetraton of telephone servces n Canada s among the hghest n the world; however, mgraton from tradtonal fxed-lne to wreless servces contnues. Canada also has a hgh rate of broadband Internet access. Whle these potental opportuntes have attracted entrants, both ndustres have barrers to entry. In broadcastng, the avalablty of lcenses s lmted, and telecommuncatons entry requres substantal captal outlay. A number of studes have examned productvty growth n Canadan telecommuncatons. Denny et al. (1981) showed that, n the absence of perfect competton and constant returns to scale, MFP growth from the tradtonal accountng framework s no longer equal to techncal progress. Rather, MFP growth s the sum of three components: techncal progress, scale economes, and non-margnal cost-prcng that arses from regulaton of the rate of return. When they appled that methodology to Bell Canada, the company s MFP was estmated to have ncreased 3.4% per 3. Investment and captal stock data are only avalable for the combned broadcastng and telecommuncatons ndustres. The output and ntermedate nputs from the Natonal Accounts are publcly avalable for the combned ndustres only. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

9 year durng the 1956-to-1976 perod: 64% of the ncrease represented scale economes; 20%, the effect of techncal progress; and 16%, the result of non-margnal cost prcng. Other research has demonstrated the mportance of scale economes. Fuss and Waverman (2002) concluded that most emprcal studes on scale economes n the telecommuncatons ndustry report evdence of ncreasng returns. For example, a number of Canadan studes found that from the 1950s to the md-1980s, the estmated returns to scale for Bell Canada ranged from 1.6 to 2.0. Prevous studes have also examned the mult-product nature of the telecommuncatons ndustry and the non-margnal cost prcng of outputs. When an ndustry produces multple products, and because of regulaton (rate-of-return or prce-cap regulaton), the prce s not proportonal to margnal costs, the correct weght to use to aggregate ndvdual outputs s the share of cost elastctes, rather than revenue share, whch would be approprate n the case of perfect competton. Fuss (1994) found that for Bell Canada, the prce was lower than margnal costs for local phone servces, but hgher for toll servces. The bas ntroduced by usng ncorrect weghts was found to be substantal n Bell Canada s case. The telecommuncatons ndustry s captal ntensve and entals consderable nvestment. The tradtonal growth accountng framework often requres the assumpton of long-run equlbrum, wth captal beng fully adjusted to demand condtons. However, n the short run, captal s not fully adjusted, whch may result n excess capacty. Because a drect observaton of the rate of utlzaton of captal nput s usually not avalable, earler studes employed a proxy for capacty utlzaton to adjust the captal used n calculatons of MFP growth. 4 Berndt and Fuss (1982) developed a framework to take changes n capacty utlzaton nto account. They argued that an adjustment should be made to the prce of captal nput n the tradtonal growth accountng framework. They showed that the ex post return on captal captures the effect of changes n capacty utlzaton. When the ex post return on captal s used to measure the contrbuton of captal to output growth, MFP growth calculated as the dfference between output growth and the contrbutons of captal and other nputs s adjusted for changes n capacty utlzaton. Nonetheless, the procedure does not completely remove cyclcal fluctuatons n the MFP growth that are assocated wth varable capacty utlzaton (Basu and Fernard 2001, Hulten 2010). Gu and Wang (2013) argued that the ex post return to captal should be used to adjust the quantty of captal nput, not the prce of captal nput as n Berndt and Fuss (2012). They showed that when the rato of ex post to ex ante return to captal s used to adjust the quantty of captal nput, the measure of MFP growth takes the rate of captal utlzaton nto account. Ths paper examnes scale economes, technologcal progress and capacty utlzaton n Canada s broadcastng and telecommuncatons ndustry. The analyss extends prevous studes to examne the effect of the compettve process n the ndustry assocated wth the reallocaton of nputs and outputs across ncumbents and between entrants and exts. The paper does not examne the mult-product nature of and non-margnal cost prcng n broadcastng and telecommuncatons; such analyss requres nformaton that s not avalable n the data used n ths analyss. 4. Some studes provded drect evdence on the rate of captal utlzaton usng survey data (the Insttut natonal de la statstque et des études économques (INSEE) has such survey). Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

10 3 Methodology Ths secton descrbes the methodology used to relate changes n aggregate labour productvty and MFP growth n broadcastng and telecommuncatons to what occurs at the frm level, followng the approach of Baldwn et al. (2012). Aggregate labour productvty and MFP growth can be decomposed nto two man components: 1) wthn-frm growth and 2) between-frm reallocaton arsng from reallocaton among ncumbents and reallocaton from frm entry and ext (Bartelsman et. al, 2005; Foster, Haltwanger, and Krzan 2001; Grlches and Regev, 1995). The decomposton method used by Baldwn et al. (2012) bulds on work by Jorgenson (1966) and Jorgenson et al. (2005) that decomposes aggregate productvty growth nto ts ndustry components, but apples t at the level of the frm. It also ntroduces non-neoclasscal features of the frm-level economc envronment, such as mperfect competton and scale economes, whereas the orgnal Jorgenson decomposton was developed under the assumpton of perfect competton and constant returns to scale. Jorgenson (1966) and Jorgenson et al. (2005) derved a decomposton of aggregate MFP growth nto a wthn-ndustry contrbuton and a between-ndustry reallocaton by comparng two approaches to estmatng aggregate MFP growth: a producton possblty fronter approach and a method that drectly aggregates across ndustres, sometmes called the top-down and bottom-up approaches (Dewert and Yu, 2012; Gu, 2012; and Schreyer 2012). Unlke the orgnal decomposton by Jorgensen (1966) and Jorgenson et al. (2005), whch uses ndustrylevel data, the present study uses frm-level data Producton possblty fronter approach The producton possblty fronter approach assumes that captal, labour and ntermedate nputs receve the same prce n all frms, but frms have dfferent producton functons that relate gross output (V) to captal, labour and ntermedate nputs at the frm level, and the prce of gross output dffers across frms. 6 Under these assumptons, aggregate gross output can be expressed as a functon of aggregate captal, aggregate labour, aggregate ntermedate nput, and a tme varable that proxes technology (T), whereas the aggregate gross output can be wrtten as a Tornqvst aggregaton of gross output across frms: V= FKLMT (,,, ), and lnv = w lnv, (1) where ln denotes the change between perods t 1 and t n logarthm, and w s the share of frm n aggregate nomnal gross output, averaged over the two perods. 5. Basu and Fernald (2002) and Petrn and Levnsohn (2010) proposed a smlar decomposton. The contrbuton of the present study s to show that the decomposton s related to the work of Jorgenson (1966) and Jorgenson et al. (2005) who decomposed aggregate MFP growth nto contrbutons of ndvdual ndustres. 6. Frms wll pay the same prces for ntermedate nputs f they purchase the same mx of output from other frms for ntermedate nput use. The formulaton n ths secton s smlar under the alternatve assumpton that gross output functon s the same across frms, and the prces of all nputs and output are the same across frms. Ths ensures the exstence of aggregate gross output producton functon at the ndustry level. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

11 Aggregate labour productvty growth, defned as the dfference between growth n aggregate gross output and growth n aggregate labour nput, can be wrtten as: ln P= lnv ln L, (2) = w ln P + w ln L ln L where ln P = lnv ln L s labour productvty growth at frm, defned as the dfference between output growth lnv and labour nput growth ln L. The growth n aggregate labour productvty n equaton 2 s decomposed nto two components: a wthn-frm effect, holdng frms shares constant, and a between-frm reallocaton effect. The wthn-frm effect s postve when labour productvty ncreases wthn frms; the between-frm reallocaton effect s postve when labour shfts toward frms wth hgher labour productvty. When the product and factor markets are compettve and the producton functon s characterzed by constant returns to scale, aggregate MFP growth can be expressed as the dfference between aggregate labour productvty growth and the effect of captal and ntermedate nput deepenng, usng the standard growth accountng framework: v = ln P α ln( K / L) α ln( M / L), (3) T K M where v T s MFP growth, and α K and α M are the shares of captal and ntermedate nputs n nomnal gross output, averaged over the two perods. Aggregate labour productvty growth, aggregate MFP growth and aggregate captal and ntermedate nput deepenng effects from the top-down approach can be traced to what occurs at the frm level. Ths s done from drect aggregaton across frms, as shown n the followng secton. 3.2 Drect aggregaton across frms The alternatve approach for estmatng aggregate labour productvty and MFP s drect aggregaton across frms (Jorgenson et al 1987, 2005). Ths method relaxes the assumpton n the producton possblty fronter approach that all nputs receve the same prce across all frms. Instead, t assumes that the prces of captal, labour and ntermedate nputs dffer across frms. For ths analyss, the drect aggregaton approach s extended to take nto account nonneoclasscal features of the economc envronment n whch frms operate. Specfcally, t s assumed that the producton functon of ndvdual frms s characterzed by ncreasng returns to scale, and that there s mperfect competton n the product market. Frm s assumed to have a producton functon that expresses gross output ( V ) as a functon of captal ( K ), labour ( L ), ntermedate nputs ( M ) and technology ( T ): V = F ( e K, e L, e M, T), (4) K L M where ek, el, e M denote the unobserved utlzaton of captal, labour and ntermedate nputs, and T ndexes technology. The producton functon exhbts ncreasng returns to scale γ. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

12 Followng Hall (1990) and Basu and Fernald (2001, 2002), output growth can be wrtten as 7 : lnv = µ ln X + a ln e + v, (5) T, where ln X s a weghted sum of nput growth usng the share of nput costs n nomnal gross output as weghts: ln X = ( α ln K + α ln L + α ln M ), (6) K L M and ln e s a weghted sum of the changes n nput utlzaton: ln e = α ln e + α ln e + α ln e. (7) K K L L M M α K, α L, and α M are the average cost shares of captal, labour and ntermedate nputs n nomnal gross output. The sum of those nput cost shares n gross output s less than one f there s economc proft. v T, s MFP growth. µ s the mark-up over margnal cost. The mark-up s related to the returns to scale γ,and the rato of economc profts to total revenue s π, by the followng equaton: P AC P µ = = = γ / (1 sπ ). (8) MC MC AC The frst equalty n equaton (8) follows from the defnton of mark-up as the rato of output prce ( P ) to margnal cost MC. The last equalty follows from an mplcaton of cost mnmzaton. The rato of average cost ( AC ) to margnal cost equals the extent of returns to scale ( γ ) under the assumpton of cost mnmzaton. In the analyss that follows, economc profts wll be assumed to be zero. Ths wll be the case f the ndustry s characterzed by monopolstc competton. 8 When economc profts are zero, mark-up s equal to returns to scale, and the sum of nput costs share n nomnal gross output s equal to one. Subtractng labour nput growth from both sdes of equaton (5) yelds the followng equaton showng the source of growth n labour productvty at frm : ln P = ( µ 1) ln X + α ln( K / L) + α ln( M / L) + a ln e + v. (9) K M T, The equaton decomposes growth n frm labour productvty nto ts varous components, ncludng scale economes, captal deepenng, ntermedate nput deepenng, varable nput utlzaton, and technologcal progress. 7. It s assumed that mark-ups do not change over a perod. When they do change, the average mark-up over the perod should be used n the equaton. 8. The decomposton can be extended to a more general framework when economc profts are not zero. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

13 The growth n frm labour productvty can be aggregated usng equaton (2) to derve aggregate labour productvty growth, whch s then substtuted n equaton (3) to obtan a decomposton of aggregate MFP growth: where v = w ( µ 1) ln X + w a ln e + w v + REALL T T, J J PJ J REALLJ = α J wj ln J ln J, wj =, J = K, L, M, PJ J w J s the share of frm n the cost of nput J averaged over two perods; that nput J receves at a frm; and approach. 9 (10) P J s the prce P J s the prce of nput J n the producton possblty fronter Aggregate MFP growth s decomposed nto a wthn-frm growth component and a between-frm reallocaton component. The wthn-frm component as shown n the frst three terms captures the effect of changes at ndvdual frms, holdng ther output share constant, that ncludes the effect of scale economes, varable nput utlzaton, and techncal progress. The last term of the decomposton s the between-frm component that measures the effect of reallocaton of captal, labour and ntermedate nputs on aggregate MFP growth. Reallocaton of an nput contrbutes postvely to aggregate MFP growth f the nput s shfted toward frms wth a hgher nput prce and a hgher margnal product. Under the assumptons of constant returns to scale, perfect competton and no excess capacty, the MFP decomposton (10) smplfes to the more standard decomposton as shown n the last two terms that expresses aggregate MFP growth as a sum of wthn-frm and between-frm reallocaton effects (Jorgenson, et al., 2005). Aggregaton of labour productvty growth gven n equaton (9) across frms usng equaton (2) yelds a decomposton of aggregate labour productvty growth: where, ln P= w ln P + w ln L ln L, (11) w ln P ( µ ) α ( ), = w 1 ln X + w ln J / L + wa ln e + wv, J T J= KM, w ln L ln L = REALLJ + α J ln( J / L) wj ln( J / L ) J= K, L, M J= K, M (12) (13) Aggregate labour productvty growth s decomposed nto a between-frm and a wthn-frm component n equaton (11). The wthn-frm effect s further decomposed nto a scale effect, a captal and ntermedate nput deepenng effect, a varable nput utlzaton effect, and techncal progress, as shown n equaton (12). The between-frm reallocaton effect s decomposed nto the 9. When frm output s based on gross output and aggregate output s a value-added concept, the correct weghts for aggregatng MFP growth across frms are Domar weghts, whch are equal to the rato of frms gross output to aggregate nomnal value-added (Domar, 1961). Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

14 effect of the reallocaton of nputs on aggregate MFP growth as shown n the frst term of equaton 13, and the effect of reallocaton on aggregate captal deepenng and aggregate ntermedate nput deepenng as shown n the second term of equaton 13. To complete the decomposton, t can be shown that the aggregate captal and ntermedate nput deepenng effect from the producton possblty fronter approach can be decomposed nto the effect of captal and ntermedate nput deepenng at the frm level, and the effect of nput reallocaton across frms on the aggregate captal deepenng effect. The decomposton of aggregate MFP growth and aggregate labour productvty growth needs to be extended to estmate the mpact of frm entry and ext. For enterng frms, nputs and outputs are observed only at the end of the perod; for extng frms, nputs and outputs are observed only at the start of the perod. As such, the growth rates of nputs, outputs and productvty over a perod cannot be calculated for entrants and exts. For ths reason, recent emprcal studes often focused on contnung frms and gnored the effect of frm entry and ext (Basu and Fernald, 2002; Petrn and Levnsohn, 2010). To extend the decomposton to estmate the effect of frm entry and ext, t s assumed that a hypothetcal frm exsts whose nputs and outputs at the start of the perod are set equal to those of exts, and whose nputs and outputs at the end of the perod are set equal to those of entrants at the end of the perod. The contrbuton of frm entry and ext to aggregate MFP growth, labour productvty growth and captal deepenng can be measured as the drect contrbuton of the hypothetcal frm to the wthn-frm component n the decomposton. For example, the contrbuton of frm entry and ext to aggregate labour productvty s estmated as the dfference between the average labour productvty of the entry cohort at the end of a perod and that of the ext cohort at the start of the perod, multpled by ther average shares n aggregate output. Ths approach to estmatng the effect of frm entry and ext s consstent wth that of Baldwn (1995), who argued that f entrants essentally dsplace exts, the effect of frm entry and ext should be evaluated by comparng entrants and exts. 4 Data Ths paper examnes labour productvty growth and MFP growth n the Canadan broadcastng and telecommuncatons ndustry from 1984 to The frm-level data are from Statstcs Canada s T2-LEAP longtudnal database, whch covers all ncorporated frms wth employees. It contans detaled nformaton from frms fnancal statements, ncludng balance sheets and ncome statements, and a longtudnal frm dentfer that can be used to examne frm entry and ext over tme. The 1984 to 1998 data are from the T2-LEAP fle, whch s based on the 1980 Standard Industry Classfcaton (SIC); the 1999 to 2008 data are from the T2-LEAP fle, whch s based on the 2007 North Amercan Industry Classfcaton System (NAICS). The broadcastng and telecommuncatons ndustry examned n ths paper conssts of telecommuncaton broadcastng (SIC 481), telecommuncaton carrers (SIC 482), and other communcaton ndustres (SIC 483) for 1984 to 1998; t conssts of broadcastng (NAICS 515) and telecommuncatons (NAICS 517) for 1999 to Data on gross output, and on captal, labour and ntermedate nputs are requred n current and constant dollars. Gross output n current dollars s measured by total sales. Gross nput n current dollars s the sum of the costs of captal, labour and ntermedate nputs. The cost of captal s measured as net ncome, 10 and the cost of labour s measured by total payroll. The cost of ntermedate nput s calculated resdually as the dfference between sales and the sum of net ncome and total payroll. 10. The varable s called net ncome for tax purposes n the 1984-to-1998 T2-LEAP fle, and net non-farm ncome n the 1999-to-2008 T2-LEAP fle. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

15 Gross output n constant dollars s derved by deflatng total sales by a gross output deflator at the ndustry level taken from Statstcs Canada s Productvty Accounts (Baldwn et al. 2007). Captal nput n constant dollars s measured by total assets deflated by a captal stock deflator at the ndustry level from the Productvty Accounts. The cost of ntermedate nputs n constant dollars s derved by deflatng the cost of ntermedate nputs by an ntermedate nput deflator at the ndustry level from the Productvty Accounts. Labour nput n T2-LEAP s measured by average labour unts (ALU), calculated as the rato of total payroll of a frm to the average annual wages of the workers n that frm s ndustry, sze class and provnce. Chart 1 dsplays the number of frms, output, and captal, labour and ntermedate nputs from T2-LEAP for the broadcastng and telecommuncatons ndustry from 1984 through A break appears n the seres from 1999 to The data for 1999 from the NAICS T2-LEAP fle are, therefore, not used. Instead, the analyss focuses on the 15-year perod from 1984 to 1998 from the SIC T2-LEAP fle, and the nne-year perod from 2000 to 2008 from the NAICS T2-LEAP fle. Chart 1 Output, nputs and number of frms, broadcastng and telecommuncatons, 1984 to 2008 ndex = 100 for 1984, except for number of frms Number of frms Gross output Captal stock Labour Intermedate nputs Notes: The data for 1984 to 1998 are from the T2-LEAP fle (form T2 (Corporaton Income Tax Return) lnked to Longtudnal Employment Analyss Program) based on the Standard Industral Classfcaton for that perod; the data for 1999 to 2008 are from the T2-LEAP fle based on the North Amercan Industry Classfcaton System for that perod. Source: Statstcs Canada, authors' tabulatons from T2-LEAP. Table 1 compares the annual growth rates of output, nputs and labour productvty n broadcastng and telecommuncatons estmated from T2-LEAP wth the aggregate statstcs from the Captal, Labour, Energy, Materals and Servces (KLEMS) database 12. For the 1984-to-1998 perod, the two sources show smlar growth rates n gross output, labour and labour productvty. In T2-LEAP, the growth of captal s somewhat hgher, and the growth of ntermedate nput, somewhat lower. The dfferences n the growth rates of captal may reflect the dfferences n the 11 Ths break s manly due to a change n the source of the T2 data. 12 The KLEMS database provdes tme seres data for multfactor productvty, output and nputs ncludng captal, labour, energy, materals and purchased servces n the NAICS back to Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

16 valuaton of captal n the two sources captal n T2-LEAP represents the book value, whereas captal n the KLEMS database s valued at replacement costs. Table 1 Annual growth rates of output and nputs, broadcastng and telecommuncatons, KLEMS versus T2-LEAP, 1984 to 1998 and 2000 to to to to 2008 less 1984 to 1998 percent KLEMS Gross output Labour Captal stock Intermedate nput Labour productvty T2-LEAP Gross output Labour Captal stock Intermedate nput Labour productvty Notes: Labour s measured by total hours worked n the KLEMS (Captal, Labour, Energy, Materals and Servces) database, and by average labour unts n T2-LEAP (Form T2 (Corporaton Income Tax Return) lnked to Longtudnal Employment Analyss Program). Captal s measured by net captal stock n the KLEMS database, and by the book value of total assets deflated by captal stock prce deflators n T2-LEAP. Sources: Statstcs Canada, authors' tabulatons from KLEMS and T2-LEAP databases. For the 2000-to-2008 perod, the annual growth rates of output, nputs and labour productvty from T2-LEAP are generally lower than those from the KLEMS database. Both sources show declnes n the growth rates of output, nputs and labour productvty, but the declnes are more pronounced n T2-LEAP. Table 2 compares output, captal, labour and ntermedate nputs n nomnal values estmated from T2-LEAP wth the aggregate statstcs from the KLEMS database. The share of labour cost n gross output s smlar n the two sources 30% of gross output durng the 1984-to-1998 perod, and 20% to 25% durng the 2000-to-2008 perod. The share of captal cost n gross output s lower n T2-LEAP (about 20%) than n the KLEMS (about 40%), manly because the captal cost n T2-LEAP excludes the amortzaton of tangble assets, whereas t s ncluded n the KLEMS. 13 The share of ntermedate nputs s much hgher n T2-LEAP than n the KLEMS, because the cost of ntermedate nputs s calculated resdually n T2-LEAP. 13. Amortzaton of tangble assets s avalable, but often mssng, on the NAICS T2-LEAP fle for 1999 to 2008, and t s not on the SIC T2-LEAP fle for 1984 to Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

17 Table 2 Average shares of nputs n nomnal gross output, broadcastng and telecommuncatons, KLEMS versus T2-LEAP, 1984 to 1998 and 2000 to to to 2008 percent percent KLEMS Labour cost Captal cost Intermedate nput costs T2-LEAP Labour cost Captal cost Intermedate nput costs Notes: May not add up to 100% because of roundng. KLEMS stands for "Captal, Labour, Energy, Materals and Servces"; T2 -LEAP stands for "Form T2 (Corporaton Income Tax Return) lnked to Employment Analyss Program". Sources: Statstcs Canada, authors tabulatons from KLEMS database and T2-LEAP. May not add up to 100% due to roundng. For the decomposton, the T2-LEAP data are benchmarked to the aggregate data from the KLEMS database. Ths ensures that the book values of assets n T2-LEAP are adjusted to reflect the current costs of captal stock, and that the captal costs n T2-LEAP are adjusted to nclude the amortzaton of tangble assets, both of whch are more approprate for productvty analyss. 14 Table 3 presents frm entry and ext rates n broadcastng and telecommuncatons for the two perods. For 1984 to 1998, entrants accounted for 74% of the total number of frms, and exts, 58%. The correspondng fgures for 2000 to 2008 were 60% and 51%. Table 3 Number, share of gross output and share of employment of enterng, extng and contnung frms, broadcastng and telecommuncatons, 1984 to 1998 and 2000 to 2008 Frms Share of of frms Share of gross output Share of employment number percent 1984 to 1998 Entrants Exts Contnuers to 2008 Entrants Exts Contnuers 628 not applcable Notes: Gross output and employment of contnuers over a perod are measured as the average of the start and end years values. Source: Statstcs Canada, authors tabulatons from T2-LEAP (Form T2 (Corporaton Income Tax Return) lnked to Longtudnal Employment Analyss Program). 14. Decomposton results usng the orgnal T2-LEAP data are smlar, except for the relatve mportance of captal and ntermedate nput deepenng for labour productvty growth. The captal deepenng effect s lower and the ntermedate nput deepenng effect s hgher from the orgnal T2-LEAP fle, compared wth those estmated from the benchmarked fle. Ths s a result of relatvely lower shares of captal costs and hgher shares of ntermedate nput costs n the orgnal fle, whch are used to estmate the captal and ntermedate nput deepenng effects. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

18 Whle entrants and exts made up a large share of the number of broadcastng and telecommuncatons frms, they accounted for small shares of gross output and employment. 15 For example, durng the 1984-to-1998 perod, entrants accounted for 24% of gross output and 17% of employment, and exters, 5% and 8%, respectvely Emprcal results Ths secton reports parameter estmates of scale economes and the effect of capacty utlzaton, whch wll be used to decompose aggregate productvty growth. It then presents the decomposton results for aggregate labour productvty growth and MFP growth for 1984 to 1998 and for 2000 to Two approaches to estmatng the scale economes and the effect of capacty utlzaton are dscussed. The frst estmates equaton (5) that expresses log output growth as a functon of log growth n combned nputs and a capacty utlzaton varable. The coeffcent µ on the log growth n combned nputs provdes an estmate of scale economes. The second approach estmates the recprocal of the scale economes 1/ µ by estmatng an equaton that expresses the log growth n combned nputs as a functon of log output growth and a capacty utlzaton varable: ln X = α + α lnv + β ln e + ε, α = 1/ µ, ε..d. errors (14) t ο t t t t The scale economes term s just the recprocal of the estmated coeffcent 1/α, and the effect of capacty utlzaton on output growth s estmated as β / α. Dewert and Fox (2008) consdered these two approaches for estmatng scale economes plus a number of others. Hall (1990) argued that f the pror expectaton s that there are ncreasng scale economes, t s more nformatve to estmate the recprocal of the scale economes and adopt the second approach, whch Dewert and Fox (2008) consdered to be more pragmatc. 17 Ths paper follows Hall and estmates equaton (14) to obtan an estmate of scale economes. The drect measure of capacty utlzaton s not avalable from T2-LEAP. The rato of ex post return to captal to ex ante return to captal can be used as a measure of capacty utlzaton, but ths requres an estmate of ex ante return to captal for each frm. Ths paper follows Baldwn et al. (2012) and uses the rato of ex post captal ncome to gross output n logarthm as a measure of capacty utlzaton. The sample for estmatng the frst dfference equaton (14) conssts of a pooled sample of contnung frms over eght three-year perods: 1984-to-1987, 1987-to-1990, 1990-to-1993, 1994-to-1996, 1996-to-1998, 2000-to-2003, 2003-to-2006, and 2006-to The data for the frst fve perods are from the SIC T2-LEAP fle; the data for the last three perods are from the NAICS T2-LEAP fle. The results of usng data for a pooled sample of contnuers over perods of varous lengths (such one, two, four or fve years) are smlar. 15. The frm entry and ext rates n terms of gross output and employment are hgher here than those reported by Baldwn and Lafrance (2011), partly because of dfferences n the perod examned. Moreover, mputatons for sales and employment were made n the 2000-to-2007 T2-LEAP fle, whch was used n Baldwn and Lafrance (2011). 16. These frm entry and ext rates also suggest that entrants are more productve, and exts are less productve, than contnuers n broadcastng and telecommuncatons. Furthermore, entrants are more productve and larger than exts. Smlar evdence on the relatve sze and productvty of entrants, exts and contnuers has been found n many other ndustres (Baldwn and Lafrance 2011). 17. Dewert and Fox (2008) also show that equaton (14) s exact when there are multple outputs and multple nputs, and mark-up s equal across varous outputs. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

19 Frms wth no sales and frms whose captal costs or ntermedate nput costs are zero or negatve are excluded from the sample. 18 The estmates of the returns to scale and the effect of capacty utlzaton are presented n Table 4. A quantle regresson s used n the frst two columns to take potental outlers n the data nto account. The ordnary least squares (OLS) results are reported n the last column for comparson. 19 Table 4 Estmates of scale economes n broadcastng and telecommuncatons, 1984 to 2008 Quantle regresson Quantle regresson Ordnary least squares (OLS) Model 1 Model 2 Model 3 estmate p-value estmate p-value estmate p-value Alpha Beta Estmate of scale economes (1/Alpha) Effect of capacty utlzaton (mnus Beta/Alpha) not applcable Note: All regressons nclude perod fxed effects. Source: Statstcs Canada, authors' tabulatons from T2-LEAP (Form T2 (Corporaton Income Tax Return) lnked to Longtudnal Employment Analyss Program). In broadcastng and telecommuncatons, the estmated return to scale from quantle regresson s about 1.25, ndcatng, on average, ncreasng returns to scale. Ths s consstent wth estmates for the ndustry reported n prevous studes (Fuss and Waverman, 2002). The coeffcent on the captal utlzaton varable has the expected sgn and s statstcally sgnfcant. The scale economes estmate from the OLS s slghtly lower than the estmates from the quantle regresson, sgnalng outlers n the sample. For the decomposton of aggregate labour productvty growth and MFP growth, the estmates based on quantle regresson n Column (2) are used, as they are robust to outlers. The decomposton uses the sample of all frms wth postve sales, whch ncludes frms wth postve captal and ntermedate nput costs and those wth negatve captal and ntermedate nput costs. 20 As noted, some frms reported zero or negatve captal and ntermedate costs n T2-LEAP. Such frms accounted for about 10% of total sales n both the 1984-to-1998 and to-2008 perods. Negatve captal and ntermedate nputs present an ssue for aggregaton across frms. In the decomposton framework presented above, the share of captal and ntermedate nput costs n total gross output s postvely related to the margnal product of captal and ntermedate nputs, whch are all postve. 21 For frms wth zero or negatve captal and ntermedate nput costs, captal costs are set equal to the total assets of the frms tmes the ndustry average rato of net ncome to total assets, and the ntermedate nput costs are set equal to nomnal gross output of 18. About 10% of frm-year observatons n the SIC T2-LEAP fle report no sales for the 1984-to-1998 perod; about 30% of frm-year observatons report negatve or zero ntermedate nputs or negatve or zero captal ncome, and those frms account for about 10% of total sales. Smlar shares of frms wth no sales, negatve captal and ntermedate nput costs are found n the NAICS T2LEAP fle for the 2000-to-2008 perod. 19. The robust regresson rreg n STATA programmng software s also used, and the results are smlar. 20. The results are smlar when a subsample of frms wth postve captal and ntermedate nput costs s used for the decomposton, except for the contrbuton of net entry to aggregate labour productvty growth. Ths suggests that frms wth negatve captal costs or negatve ntermedate nputs, ncludng entrants and exts, are relatvely small. 21. The share of captal and ntermedate nput costs n gross output s equal to the captal and ntermedate nput elastcty dvded by the return to scale. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089

20 the frms tmes the ndustry average share of ntermedate nputs n gross output. 22 Essentally, t s assumed that the user costs of frms wth negatve captal costs are the same as the average user costs of captal n the ndustry. 5.1 Results from the producton possblty fronter approach Table 5 presents the decomposton results for annual aggregate labour productvty growth n broadcastng and telecommuncatons from the producton possblty fronter (top-down) approach. Aggregate labour productvty growth s decomposed nto ts man sources: aggregate captal and ntermedate nput deepenng, and aggregate MFP growth. Table 5 Decomposton of annual aggregate labour productvty growth, broadcastng and telecommuncatons, 1984 to 1998 and 2000 to 2008 (producton possblty fronter approach) 1984 to to to 2008 less 1984 to 1998 percent Aggregate labour productvty growth Contrbuton from Captal deepenng Intermedate nput deepenng Multfactor productvty growth Note: Percentages may not add up to roundng. Results are benchmarked to KLEMS (Captal, Labour, Energy, Materals and Servces. Source: Statstcs Canada, authors tabulatons from T2-LEAP (Form T2 (Corporaton Income Tax Return) lnked to Longtudnal Employment Analyss Program). Aggregate labour productvty n broadcastng and telecommuncatons grew steadly snce the md-1980s, although the pace slowed from 4.9% per year n the 1984-to-1998 perod to 2.9% n the 2000-to-2008 perod, a 2.1-percentage-pont declne. Ths slowdown was due to declnes n captal deepenng (nvestment) and n ntermedate nput deepenng. The declne n captal deepenng accounted for 1.8 percentage ponts of the overall decrease, and the declne n ntermedate nput deepenng, 0.8 percentage ponts. By contrast, aggregate MFP growth rose from 1.4% per year n the 1984-to-1998 perod to 1.9% per year n the 2000-to-2008 perod. 5.2 Results from drect aggregaton across frms In the next three tables, annual aggregate labour productvty growth and MFP growth and the nput deepenng effects from the top-down approach are decomposed nto a wthn-frm growth effect, a between-frm reallocaton effect, and the effect of net frm entry. The wthn-frm component captures the effect of growth wthn contnuers; the between-frm reallocaton, the reallocaton across contnuers; and the net entry component, the drect contrbuton of entry and ext cohorts. 22. Alternatvely, captal and ntermedate nput costs for frms wth negatve or zero captal and ntermedate nput costs can be mputed as sales tmes the ndustry average share of captal and ntermedate nput costs n total sales; the decomposton results of ths mputaton are smlar. They can also be mputed usng the ratos or shares n frms n the bottom quantle of the dstrbuton. Ths s not expected to change the results, because frms wth zero or negatve captal and ntermedate nput costs account for a small share of gross output. Economc Analyss Research Paper Seres Statstcs Canada Catalogue no.11f0027m, no. 089