ipm 23 Alexopoulos Cohen.fm The Effects of Computer Technologies on the Canadian Economy: Evidence from New Direct Measures Michelle Alexopoulos and Jon Cohen1 University of Toronto ABSTRACT New indicators of technical change in the field of computers based on new titles held by Canadian libraries are presented, and are used to demonstrate that a positive computer technology shock in Canada increases hours worked, output, and productivity in the short run. These measures indicate, first, that advances in the implementation of computer technology in Canada are largely influenced by innovations in the United States; and second, when compared to a United States-based indicator, that a gap emerged between United States and Canadian-held titles around the time that the productivity gap emerged between the two countries. Given that a strong, causal relationship is found to exist between the new indicators and total factor productivity, this evidence provides additional support for the hypothesis that crosss-border differences in the development and use of new computer technologies play a key role in explaining Canada’s productivity gap with the United States. RÉSUMÉ De nouveaux indicateurs du changement technique dans le domaine de l'informatique, basés sur les nouveaux titres des bibliothèques canadiennes, sont présentés et montrent qu'un choc technologique positif dans le domaine de l'informatique au Canada accroît à court terme le nombre d'heures travaillées, les résultats et la productivité. Ces mesures indiquent, premièrement, que les progrès de l'application de la technologie informatique au Canada sont en grande partie influencés par les innovations aux États-Unis et, deuxièmement, que lorsqu'elles sont comparées à un indicateur américaine, un écart apparaît entre les titres dans les bibliothèques canadiennes et américaines à peu près au moment où l'écart entre la productivité des deux pays se fait jour. Compte tenu de l'existence d'une forte relation de cause à effet entre les nouveaux indicateurs et la productivité totale des facteurs, ces observations renforcent l'hypothèse selon laquelle les différences entre le développement et l'utilisation des nouvelles technologies informatiques dans les deux pays jouent un rôle essentiel pour expliquer l'écart de productivité du Canada. 1 Michelle Alexopoulos is Associate Professor and Jon Cohen is Professor Emeritus in the Department of Econom- ics at the University of Toronto. The authors thank the referees for useful comments. Email: malex@chass.utoronto.ca. I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 17 C O M P U T E R T E C H N O L O G I E S A R E O F T E N viewed as a key contributor to productivity growth in advanced industrial countries such as the United States and Canada.2 It follows that cross-country differences in the development and use of these technologies may at least par- tially account for cross-country productivity differentials. This is particularly relevant for Canada and the United States because it is fre- quently argued that the growth in the produc- t i v i t y g a p b e t w e e n t h e s e t w o c o u n t r i e s i s attri butabl e to the mo re ra pid a doption of information technologies in the latter than in former.3 As compelling as this argument may appear, the lack of direct measures of technical change in this area has made it difficult to pro- vide a quantitative assessment of either the impact of computers on economic activity in the two countries or the causal link between differences in their adoption rates and produc- tivity differentials. In short, if we want to quan- t i f y t h e i m p a c t o n t h e s e e c o n o m i e s o f innovations in this area, we must be able to measure them. As it happens, that is precisely what we propose to do in this article. We first present new direct measures of technological change in the field of computers in Canada for the 1950-2005 period, based on the number of new computer-related titles held by Canadian libraries4 and then use them to show that: (1) the United States is the principal source of advances in computer tech- nology in Canada; (2) the rate of adoption of new computer technologies in the United States began to surpass that of Canada in the 1970s; and (3) a strong relationship exists between computer innovations and productivity, GDP, and hours worked.5 Together, these findings provide empir- ical support for the hypothesis that cross-border differences in the commercialization and rate of adoption of new computer technologies have played a key role in the widening of the Canada- U.S. productivity gap. The article is divided into four sections. The first section discusses the indicators and their properties. Section two reports the results of our regressions. Section three discusses the poten- tial link to the Canadian and U.S. productivity gap. The fourth and final section concludes. The Indicators Most would agree that a good direct measure of technical change should, at a minimum: (1) be available at least on an annual basis over a long time horizon; (2) be objectively determined; (3) weight different technologies according to their importance; and (4) tap into the full range of new advances. Moreover, for many purposes we would add a fifth requirement – that the indicator cap- ture innovations at the moment of their commer- cialization. This is important for two reasons. First, much of the impact on output, productivity and employment occurs through the adoption of new technologies. Second, unanticipated tech- nology shocks, an important feature of many eco- nomic models, are generally identified not by the timing of the invention or even the patenting of a new technique but by its use. Our new book-based indicators possess all of these features. They are: (1) objective because they are determined by cataloguing criteria established and followed consistently by librari- ans; (2) quantifiable because they are based on the number of new titles; (3) weighted because 2 See, for example, Alexopoulos (2011), Alexopoulos and Cohen (2011), Oliner et al. (2007), Khan and Santos (2002), Stiroh (2002), Van Ark et al. (2003), and Sharpe (2006), and the citations within. 3 See Basu et al. (2003) for an interesting study of differences in productivity growth between the United States and the United Kingdom and the relationship to IT technologies. 4 The approach, developed by Alexopoulos (2011), was applied initially to the United States. 5 In what follows we use the terms book-based and publication-based interchangeably. The indicators are primarily based on new manuscripts. However, pamphlets that are catalogued are included. Serial publi- cations and continuing resources are, by and large, excluded from the counts. 1 8 N U M B E R 2 3 , S P R I N G 2 0 1 2 more titles are published on important advances than on lesser ones; and (4) broad-based because new titles appear on all innovations of any sig- nificance.6 Finally, new publications, for good economic reasons, are timed to coincide with the commercialization of new products or pro- c es s e s . T h e i n no v a t i n g c o m p a n i e s w a nt t o spread the word about their new devices – what they are, how to use and maintain them, – while publishers and their writers want to profit from the market demand for information about these new technologies. In all cases timing is critical – too early and there is no market, too late, and the market is fully served. Although there are clearly other means aside from print to convey information about new technologies, our find- ings (Alexopoulos (2011) and Alexopoulos and Cohen (2009, 2011)) suggest that these book- based indicators provide a compelling way to quantify technical change and to evaluate its impact on the economy.7 Description of the Indicators Although we focused in our previous work on the United States, it is possible to use a similar methodology (developed in Alexopoulos (2011)) to create comparable technology indicators for Canada. This is because Canadian libraries also use MAchine Readable Cataloging (MARC) records to run their online catalogues.8 As is well-known, computer-related innovations in Canada are an amalgam of home-grown and imported advances, much more so than, for example, in the United States. Moreover, Can- ada does not have a library of the size and scope of the Library of Congress in the United States. For these reasons, we reshaped our approach to ensure that we capture the full range of foreign and domestic developed computer technologies commercialized in Canada. In particular, we used information from the catalogues of 1,062 Canadian libraries covered in the WorldCat database of the Online Computer Library Cen- tre (OCLC) on the number of new computer titles published between 1950 and 2005, without regard for country of publication.9 While not all Canadian libraries are members of OCLC, the membership includes the National Library of Canada, the country’s largest public libraries (e.g. those in Toronto, Montreal and Vancou- ver), and all major university libraries. As such, the combined MARC records of the member libraries provide a comprehensive list of all major new computer-related titles available to the Canadian public. Even though the MARC records were designed to serve the online cataloguing needs of librari- ans, it turns out , because of the large amounts of data buried in them, that they are also a poten- tially powerful research device for, among others, economists. For example, each MARC record contains information on the type of book (for example, a new title, a new edition of an existing one, a reprint, or a translation), the country and language of the publication, the publisher, the Library of Congress and/or the Dewey Decimal Classification Code, and a list of major subjects treated in the book. These data enable us to com- 6 Alexopoulos (2011), and Alexopoulos and Cohen (2009, 2011) also present evidence that the book publication measures are related to traditional measures of technical change such as R&D, patents, major innovations, and journal article counts in the United States. 7 Although some may be concerned that changes in the number of titles is driven by ups and downs in the publishing industry, our findings in the papers cited indicate that the patterns, on the whole, appear to be dictated by changes in innovations. Finally, although cataloguing and keyword assignment are poten- tially subject to error, there is no reason to believe that misclassification is a problem. 8 See Appendix A for an example of a MARC record. 9 The data used for this research were based on a snapshot of the OCLC’s WorldCat database as of the mid- dle of 2010. We took 2005 as our cut-off date to avoid any biases created by the backlog of uncatalogued titles. I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 19 0 20 40 60 80 100 120 19 50 19 56 Chart 1 Indicators of C Chart 2 Fraction of Eng in Canada Pub 0 1000 2000 3000 4000 5000 6000 7000 All new co 0 1000 2000 3000 4000 5000 6000 7000 New Engli held in Ca 0 1000 2000 3000 4000 5000 6000 New US pu computer 19 50 19 56 19 50 19 56 19 50 19 56 19 62 19 68 19 74 19 80 19 86 19 92 19 98 20 04 pile a list of new titles published each year on computers and computer science that are held by Canadian libraries in the sample between 1950 and 2005.10 To ensure that our titles actually rep- resent the appearance of new technologies, we eliminate from the sample all books that include history as a descriptor since they, almost by defi- nition, focus on the past not the present. Thanks to the richness of the MARC records, we are able to create, as can be seen in Chart 1, three slightly different indicators for the purposes of this analy- sis: the first includes all books held in the field by the Canadian libraries regardless of the language or country of publication; the second excludes non-English language titles; while the third is limited to English language titles published in the United States. In all cases, the indicator includes manuals and books that deal with new computer technologies that describe their nature and function, how they work, and how to use or repair them. Some of the titles are published or sponsored by the innovator or the company that developed it, while others are written by third parties who hope to profit from sales of the book. As noted earlier, all groups have an economic incentive to ensure that the publications appear as close as possible to the commercialization date of the new technologies. At the same time, libraries, seeking to serve their market, will then purchase the books they believe will be demanded and used by their patrons. For this reason we would expect to observe a close chronological coincidence between the copyright date of the first book on a new technology that appears in a library (as captured by the WorldCat database) and its commercialization date as reported in other source material. The results in Table 1, based on the dates for a sample of com- puter innovations commercialized in Canada and the United States, confirms this timing. 10 See Appendix B for a description of the Dewey Decimal Classifications and Library of Congress Classifications associated with the computer and computer science classifications included in the counts. omputer-Related Titles by Copyright Date lish-Language Held Computer Books lished in the United States (per cent) mputer titles held in Canada sh-language computer titles nada blished English-language titles held in Canada 19 62 19 68 19 74 19 80 19 86 19 92 19 98 20 04 19 62 19 68 19 74 19 80 19 86 19 92 19 98 20 04 19 62 19 68 19 74 19 80 19 86 19 92 19 98 20 04 2 0 N U M B E R 2 3 , S P R I N G 2 0 1 2 Table 1 also demonstrates the fluidity with which computer technologies flow across bor- ders. It appears that the technologies developed in the United States and the United Kingdom were quickly embraced in Canada. For example, the commercialization dates in Canada for com- puter software programs such as Lotus 1-2-3, and Windows were virtually identical to those in Table 1 Comparison of Dates for Selected Computer Innovations Notes: First book dates correspond to the copyright date of the first book held by a library in the OCLC WorldCat data- base. See Alexopoulos and Cohen (2011) for sources of the US innovation and commercialization dates for Windows, Lotus, Apple II+, Macintosh, Lisa, IBM PC, IBM PC/AT and Commodore 64, and the footnotes for information used for the Canadian commercialization dates and the dating of PAT, Corel Draw and the Sinclair ZX80. Innovation Date of innovation Country of Invention Commercialization Date in the United States First American Book Date Commercialization Date in Canada First Canadian Book Date Windows Nov. 1983 US Nov. 1985 1985 Nov 19851 1. http://www.guidebookgallery.org/ads/magazines/windows/win10-powerwindows-8 1986 Lotus Nov. 1982 US Jan. 1983 1983 Feb 19832 2. Michael Kieran, "Programs for micros enter new generation," The Globe and Mail, pp. R.12-R.12, Feb 28, 1983 1983 Apple II+ 1978 US June 1979 1979 19793 3. See comment on http://www.facebook.com/torontostar/posts/227813607276004 by J. Lyng, a resident of Toronto, and blog post on http://taoofnews.com/2011/01/25/thirty-years-in-new-media/ 1979 Macintosh Jan. 1984 US First Quarter 1984 1983 Jan 19844 4. Jonathan Chevreau, "Xerox Canada will carry Lisa, Macintosh machines," The Globe and Mail, B.14, May 11, 1984 1984 Lisa 1978 US Jan. 1983 1983 April 19835 5. Michael Kieran, "Programs for micros enter new generation," The Globe and Mail, pp. R.12-R.12, Feb 28, 1983 1984 IBM PC July 1980 US Aug. 1981 19826 6. WorldCat points to a scanned book captured by Google that is entitled Technical specifications under the series: IBM Personal Computer. Hardware reference library published by IBM. Aug 19817 7. Jonathan Chevreau, "Computerland chief expects IBM entry to add credibility to personal market," The Globe and Mail, pp. B.9-B.9, Aug 27, 1981 1982 IBM PC/AT Aug. 1984 US Fall 1984 19858 8. While not physically held in a library, OCLC records point to a technical publication by IBM for this computer with a copyright date 1984. Fall 19849 9. Jonathan Chevreau, "IBM launches new 20-megabyte PC unit," The Globe and Mail, pp. B.1-B.1, Aug 15, 1984 1985 Commodore 64 Jan. 1982 US Nov. 1982 1982 Sept 198210 10. Jonathan Chevreau.”Price war is expected in personal computers” The Globe and Mail, pp. B 15, Sept. 8, 1982 1982 PAT (OpenText)11 11. http://www.opentext.com/2/global/company/company-history.htm#ecml 1989 Canada 1992 1993 by 1991 1992 Corel Draw12 12. http://www.fundinguniverse.com/company-histories/Corel-Corporation-Company-History.html and American review of the technology by S. Rosenberg, “Corel Draw shows great promise” Byte Magazine, June 1, p. 213. 1987-1989 Canada 1989 1988 Jan 1989 1988 Sinclair ZX8013 13. http://en.wikipedia.org/wiki/Sinclair_ZX80 and http://maben.homeip.net/static/S100/sinclair/brochure/Sinclair%20ZX80%20Jan%2081%20Byte%20review.pdf 1980 UK Fall 1980 1981 Late 1980 1980 I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 21 http://www.guidebookgallery.org/ads/magazines/windows/win10-powerwindows-8 http://www.facebook.com/torontostar/posts/227813607276004 http://www.facebook.com/torontostar/posts/227813607276004 http://www.fundinguniverse.com/company-histories/Corel-Corporation-Company-History.html http://www.fundinguniverse.com/company-histories/Corel-Corporation-Company-History.html http://en.wikipedia.org/wiki/Sinclair_ZX80 http://maben.homeip.net/static/S100/sinclair/brochure/Sinclair%20ZX80%20Jan%2081%20Byte%20review.pdf the United States. The same is true for com- puter hardware such as the Apple II+, Commo- dore 64, IBM PCs, Sinclair, and the Macintosh all of which appear almost simultaneously in the two countries. As it happens, the flow was not uni - dir ec ti on al . C an ad ian i n n ovat io n s li ke C o re l D r aw, de v el o p ed by C o re l , a n d PAT, developed by Open Text Corporation, were adopted by American firms soon after their Canadian release dates. A close relationship can also be observed in Table 1 between first commercialization dates and copyright dates (book dates in the table), independent of the location of the innovations. For all of these cases, there is never more than a year’s difference between the copyright date and the year of its adoption in either the United States or Canada. In other words, the appear- ance of a computer-related book in a Canadian library provides a good indicator of the initial arrival (commercialization) of the new technol- ogy in the country. Moreover, as we have shown elsewhere (Alexopoulos (2011), and Alexopoulos and Cohen (2011)), new titles are associated with the introduction of new processes or prod- ucts and not their diffusion. Our indicators, in short, should provide a good measure of com- puter innovations in Canada. We turn now to the central question of this article – what impact did these new technologies have on economic activity in Canada? Output, Productivity, and Technical Change We have used similar indicators in other papers (Alexopoulos (2008, 2011), and Alex- opoulos and Cohen (2009 and 2011)) to explore the relationship between innovative activity in a v ar i et y o f f i el d s, ou t p u t , p ro du c t i vi t y, a nd employment in the United States, drawing on the MARC records of the Library of Congress, Amazon.com’s booklists, and R.R. Bowker’s publishers’ lists. In particular, we have found, first, that new computer technologies have been a n i m p o r t a n t d e t e r m i n a n t o f p r o d u c t i v i t y growth in the United States during the post WWII period, and second, that computer- driven technology shocks have led to short run increases in productivity, employment, and out- put. We repeat the analysis using the new Cana- dian indicators and ask: do we observe the same relationships in Canada? To answer this question, we estimate the fol- lowing bi-variate VARs:11 Yt = α+γt+ρYt-1 +εt (1) where Yt = [ln(Zt), ln(Xt)]’, with Zt being our measure of aggregate output or total factor productivity (TFP), and Xt being the number of new computer titles.12 As in Alexopoulos (2011) and Shea (1998), our computer indica- tor is ordered last in the VAR and a computer technology shock is identified by assuming that it affects the Z variables with a one year time lag. Our measure of aggregate TFP is from Madsen (2007) while hours worked and real GDP are based on data from Maddison (2010), the Historical Statistics of Canada, and CANSIM. Chart 3 displays the impulse responses to a one standard deviation com- puter technology shock (as identified by our indicator), and 90 per cent confidence inter- vals. Table 2 reports the Granger-causality 11 A vector autoregression (VAR) is a popular statistical model used to capture the linear interdependencies among multiple time series. As above, each of the variables in the model is represented by an equation linking the variable’s current value to lags of its own values, the lags of all the other variables in the model and other deterministic series such as a time trend. 12 Our choice of this specification is driven by two main considerations. First, it is comparable to the spec- ification used in our earlier work focusing on the United States. Second, Gospodinov, Maynard and Pesav- ento (2011) highlight problems associated with choosing a specification based on univariate unit root tests and demonstrate that severe biases can be introduced by removing low frequency movements by estimating VARs in first differences. 2 2 N U M B E R 2 3 , S P R I N G 2 0 1 2 http://en.wikipedia.org/wiki/Linear_dependence http://en.wikipedia.org/wiki/Time_series nglish Language uter Titles 15 20 15 20 15 20 nglish Language uter Titles nglish Language uter Titles and Xt is the value Do Hours Grange-Cause Computer Technologies? 0.789 0.757 0.850 Response of GDP to a positive shock to: All Canadian Held Computer Titles All Canadian Held English Language Computer Titles All Canadian Held E US Published Comp Response of Hours to a positive shock to: Response of TFP to a positive shock to: 0 5 10 15 20 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 5 10 15 20 0 5 10 15 20 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0 5 10 15 20 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0 5 10 0 5 10 15 20 0 5 10 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0 5 10 15 20 0 5 10 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035 0.040 All Canadian Held Computer Titles All Canadian Held English Language Computer Titles All Canadian Held E US Published Comp All Canadian Held Computer Titles All Canadian Held English Language Computer Titles All Canadian Held E US Published Comp tests, and Table 3 reports the variance decom- positions for the bi-variate cases. The results echo those for the United States reported in Alexopoulos (2011) and Alexopoulos Chart 3 Impulse Response Functions- Bi-Variate VAR Table 2 P-values of Granger Causality Tests Notes: For all cases Yt = α+γt+ρYt-1 +εt , where Yt = [ln(GDPt), ln(Xt)]’, Yt = [ln(TFPt), ln(Xt)]’ or Yt = [ln(Hourst), ln(Xt)]’ of the indicator at time t. Technology Indicator Do Computer Technologies Granger-Cause GDP? Does GDP Granger-Cause Computer Technologies? Do Computer Technologies Granger-Cause TFP? Does TFP Granger-Cause Computer Technologies? Do Computer Technologies Granger-Cause Hours? All Canadian held computer books (COMPALL) 0.071 0.795 0.059 0.914 0.046 All Canadian held computer books in English (COMPENG) 0.019 0.518 0.018 0.856 0.040 All Canadian held computer books in English published in the United States(COMPUS) 0.111 0.168 0.064 0.315 0.061 I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 23 and Cohen (2011). We find, first, that com- puter-related technical change, as measured by our indicators, had a significant impact on out- put, hours worked, and TFP in post-WWII Canada. Second, our Granger-causality tests indicate that causality runs from computer- based innovations to output, hours, and TFP and not the other way round. And, third, of the three series (all computer-related books, all English language titles, and English language, U.S.- based publications), it is the second that has the strongest influence on output, hours, and TFP. The impulse response functions associated with our VARs can be seen in Chart 3. As the first panel in Chart 3 shows, GDP significantly rises above trend for approximately 25 years fol- lowing a positive shock to computer technolo- gies (as identified by our indicators) with the peak effect occurring after approximately seven years. Panels 2 and 3 demonstrate that at least part of the increase in output is attributable to rises in hours worked and TFP – both of whose responses are similar to that of GDP.13 Each of these variables significantly rise for 15 to 25 years with their peak effect occurring between years 5 and 7. Of equal interest, the effects for all of the variables are roughly the same for all three indicators. The variance decompositions are reported in Table 3. We find, first, that in the initial years the impact of technical change in computers on our three variables is relatively weak. To be more pre- cise, in year three, the indicators accounted for 3.4–6.9 per cent of the variation in GDP, 2.8–5.5 per cent of the variation in TFP and 2.2–5.1 per cent of the variation in hours. By year 6, however, the effect has changed quite noticeably: technical advances in computers now account for 9.0–22.0 per cent of the variation in GDP, 11.8–21.3 per 13 While the analyses in Alexopoulos (2011) and Alexopoulos and Cohen (2011) are based on a slightly different time period, 1955-1997 and 1980-2008, their findings suggest the peak impacts for a computer innovation occur earlier in the United States. Table 3 Per cent of Variation Due to Technology in Two Variable VARs Notes: These decompositions are based on bi-variate VARs where ln(GDP), ln(TFP) and ln(L) are ordered first. For the cases of using the new book measures and patents the VAR takes the form Yt = α+γt+ρYt-1 +εt where Yt = [ln(GDPt), ln(Xt)]’, Yt = [ln(TFPt), ln(Xt)]’ or Yt = [ln(Lt), ln(Xt)]’ and Xt is the value of the indicator at time t. Years ln(GDP) ln(TFP) ln(Hours) All Canadian held computer books (COMPALL) 3 3.379 2.821 2.158 6 12.593 11.791 9.124 9 21.117 21.433 16.531 12 27.353 29.247 22.293 All Canadian held computer books in English (COMPENG) 3 6.888 5.528 2.732 6 22.359 21.275 11.474 9 33.719 35.360 20.510 12 40.779 45.048 27.339 All Canadian held computer books in English published in the US (COMPUS) 3 3.803 5.193 5.121 6 8.987 13.650 14.785 9 11.715 18.697 20.775 12 13.135 21.506 23.788 2 4 N U M B E R 2 3 , S P R I N G 2 0 1 2 cent of the variation in TFP and 9.1–14.8 per cent of the variation in hours. By year 12, the levels have jumped again: 13.1–40.8 per cent of the vari- ation in GDP, 21.5–45.0 per cent of the variation in TFP and 22.3–27.3 per cent of the variation in hours.14 In general, the impact of computer tech- nologies on the three variables is largest at medium-run horizons. Second, the indicators based on new English language computer titles account, on the whole, for a much larger percent- age of the variance in our three variables than do the other two. Ta b l e 4 a n d C h a r t 4 r e p o r t t h e v a r i a n c e decompositions and impulse responses related to the tri-variate VAR: Yt = α+γt+ρYt-1 +εt (1) where Yt = [ln(TFPt), ln(Hourst), ln(Xt)]’. As above, the technology indicators are ordered 14 The variation in GDP and TFP attributable to computers reported by Alexopoulos (2011) and Alexopoulos and Cohen (2011) are of similar magnitude. However, for the United States, the computer innovations tend to explain a larger share of the variance in years 3-6. Chart 4 Impulse Response Function: Tri-Variate VAR Positive Shock To All Canadian Held Computer Books Response of TFP Response of Hours Positive Shock To All Canadian Held English Language Computer Books Positive Shock To All Canadian Held English Language US Published Computer Books -0.005 0.000 0.005 0.010 0.015 0.020 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.015 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.010 0 5 10 15 20 0 5 10 15 200 5 10 15 20 0 5 10 15 200 5 10 15 20 -0.015 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 -0.010 -0.005 0.000 0.005 0.010 0.015 0.020 0.025 0 5 10 15 20 Response of TFP Response of Hours Response of TFP Response of Hours I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 25 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 Canadi 19 50 19 56 Table 4 Per cent of Variat in the Tri-variate Notes: For all case ln(Xt)]’ and Xt All Canadian held c (LNCOMPA) All Canadian held c in English (LNCOMPE) All Canadian held c in English Published in the US (LNCOMPUE) Chart 5 Total Factor Pr 1950-2005 (1 Source: The TFP me an TFP American TFP 19 62 19 68 19 74 19 80 19 86 19 92 19 98 20 04 last. Again, we find evidence that a positive com- puter technology shock significantly increases productivity and hours worked. However, the confidence intervals for this case do not exclude the possibility that hours worked may initially decrease immediately following the shock. On the other hand, the results in Table 4 do confirm that new computer technologies play a strong role in productivity movements and a moderate one in variations in hours worked in the medium run. Canadian Productivity and the U.S.-Canada Productivity Gap In Chart 5, we present Canadian and Ameri- can TFP indices f rom Madsen (2007). Two trends are apparent. First, his estimates suggest that Canadian TFP in 2005 was approximately the same as it was in the mid-1970s. Second, starting in the late 1970s, Canada’s TFP growth failed to keep pace with that of the United States, giving rise to a well-known productivity gap.15 On the face of it, the first trend would seem to be inconsistent with the analysis in the previous subsection. In addition, it appears to be at odds with the upsurge in computer titles held in Can- ada and, accordingly, with the apparent advances in computer technology in this country. As it happens, the problem lies not with the data or with our argument but with the misunderstand- ing that TFP is a proxy for technological inno- vation. As we all know but often forget, TFP is a residual that contains all those factors other than labour and capital that affect GDP growth. These include, among other things, changes in scale economies, organizational capital, utiliza- tion rates, measurement errors and so on, some of which could easily affect the size and rate of change of the residual. The bottom line, for our purposes, is that TFP does not measure pure technical change – which is exactly what our book based indicator is capturing. Moreover, although computer-based technical change did 15 This gap is also seen in labour productivity measures. ion Due to Computer Technologies VARs s Yt = α+γt+ρYt-1 +εt , where Yt = [ln(TFPt), ln(Hourst), is the value of the indicator at time t. Horizon (Years) ln(TFP) ln(Hours) omputer books 3 3.368 0.256 6 13.311 2.715 9 23.273 8.010 12 30.742 14.239 omputer books 3 7.136 0.421 6 25.226 4.658 9 39.543 13.441 12 48.307 22.863 omputer books 3 6.571 1.119 6 15.683 6.110 9 20.348 11.656 12 22.568 15.515 oductivity in Canada and the United States, 950=100) asures for the total economy are from Madison (2007). 2 6 N U M B E R 2 3 , S P R I N G 2 0 1 2 19 92 19 98 20 04 United States 19 93 19 99 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 N um be r of N ew T it le s Canadian Held Computer Books 19 50 19 56 19 62 19 68 19 74 19 80 19 86 US Held Computer Books play an important role in driving productivity advances in Canada, there were other, counter- balancing forces at work as well. Although unpacking the contents of the residual exceeds the scope of this article, it is a worthwhile project for future research. As for the second trend, the productivity gap has, naturally enough, attracted the attention of Canadian academics and policy makers.16 The central questions are the obvious ones – what caused the gap and why has it grown? The answers matter for at least two reasons. First, we cannot begin to address the problem until we identify its source and, second, our ability to compete wi th our neighbor to the south is closely linked to the relative productivity in the two countries. Results reported in papers such as Sharpe (2010), Rao (2011), Rao et al. (2004), Rao and Tang (2001) and Van Ark et al. (2003) suggest that differences in the use and the rate of adoption of information technologies, especially computers, in the two countries are likely a major contributor to gap. While there are always issues with cross- country comparisons, both our metrics, and our overall findings tend to support this view. First, as noted earlier, technological advances in the field of computers have had a significant impact on Canadian productivity. Moreover, as reported in Alexopoulos (2011) and Alexopoulos and Cohen (2011), a similar relationship can be observed in the United States. It follows, then, that if there were a gap in the adoption of new computer tech- nologies between the two countries, this may have been a non-trivial contributor to the pro- ductivity gap. The question is: did such a gap exist? Cross border data on ICT investment in Sharpe (2010) points to a gap, as does our new book-based indicators. Specifically, Chart 6 shows the num- ber of new computer titles held by libraries in the United States as recorded by OCLC along- side the number of new computer titles held by Canadian libraries. It shows that a gap begins to emerge in the early 1970s.17 A similar pattern can be seen in Chart 7, based on indicators cre- ated from the holdings of the largest library in the United States, the Library of Congress, and the largest in Canada, the University of Toronto 16 See, for example, Rao and Tang (2001), Rao et al. (2004, 2008), Baldwin, Gu and Yan (2008), and Rao (2011). Chart 6 New Computer Books Held in Canada and the by Copyright Date Chart 7 New Computer Titles by Copyright Date 0 200 400 600 800 1000 1200 1400 1600 1800 N u m b er o f N ew T it le s University of Toronto Libraries 19 57 19 63 19 69 19 75 19 81 19 87 Library of Congress 2000 I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 27 Libraries.18 Since our VAR results suggest that a lag exists between the commercialization of new computer technologies (as measured by our new indicators) and their impact on productivity, the appearance of a productivity gap in the 1980s is perfectly consistent with the emergence of an adoption gap a few years earlier.19 In short, the computer-related technology gap measured by our book-based indicators does appear to have contributed to the emergence of the productiv- ity gap beginning in the 1980s.20 Conclusion In this article, we draw on the holdings of Canadian libraries to develop new book-based indicators of technical change in the field of computers for the years 1950-2005 and use them to determine the impact in Canada of techno- logical advances in this area on output, produc- tivity, and employment. As we have argued elsewhere (Alexopoulos (2011, 2008) and Alex- opoulos and Cohen (2009, 2011), these new indicators resolve many of the problems that plague traditional measures of innovative activ- ity such as patent citations and research and development expenditures. They also have the additional advantage that because they are con- sistent across countries as well as over time, they facilitate international time series comparisons. We are able to show, for example, that most of the computer innovations identified in publica- tions held in Canada actually originate in the United States. More, we can demonstrate, using VARs, that similar to our results for the United States, positive computer-related technology shocks in Canada lead to increases in GDP, TFP and hours worked in the short and medium run. Finally, we can make use of our new approach to show that starting in the 1970s, the number of new computer titles in Canada began to lag sig- nificantly the number in the United States (the appearance, our indicators suggest, of a technol- ogy gap), contributing a decade or so later to the emergence of a productivity gap between the two countries. This finding still leaves open the question of why the technology gap emerged in the first place, but with the identification of this problem, we hope policy makers can take steps to address it. 17 One might be concerned that funding differences could affect the comparability of the indicators across the two countries. However, statistics available from http://www.oclc.org/reports/escan/economic/educationli- braryspending.htm suggest that Canada spends slightly more per capita (4.6 per cent) on its library collections than does the United States, and more as a fraction of GDP (0.20 per cent versus 0.12 per cent). Given that all major U.S. and Canadian libraries are represented in our sample, and that the budgets are sufficient to allow Canadian libraries to accumulate in aggregate the same titles as their American counterparts, we believe the indicators do provide important information about the knowledge gap. 18 The University of Toronto Libraries has one of the largest collections in North America. According to sta- tistics based on the number of titles and volumes held, its collection is approximately 53 per cent of the size of the Library of Congress despite the fact that the Library of Congress serves a much larger popula- tion than the University of Toronto Libraries. 19 It should be noted that a widening productivitty gap beween Canada and the United States does not require that there be a growing gap in computer titles in relative terms. The current level of TFP depends on the lags of all of the new titles in the economy (not in relative terms). The fact that the Americans are still accumulating more new titles would imply that the gap should be there. However, as the current gap on the new books shrinks, the gap would widen less provided that the coefficients on the lagged titles are the same in the two countries. 20 Alexopoulos and Tombe (2011) identify a gap in management techniques which may also contribute to the presence of the gap. 2 8 N U M B E R 2 3 , S P R I N G 2 0 1 2 http://www.oclc.org/reports/escan/economic/educationlibraryspending.htm http://www.oclc.org/reports/escan/economic/educationlibraryspending.htm References Alexopoulos, Michelle (2008) “Extra! Extra! Some positive technology shocks are expansionary!” Economics Letters, Vol. 101, pp. 153-156. Alexopoulos, Michelle (2011) “Read All About it!! What happens following a technology shock?” American Economic Review, Vol. 101, No. 4, pp. 1144-79. Alexopoulos, Michelle and Jon Cohen (2009) “Mea- suring our Ignorance, One Book at a Time: New Indicators of Technical Change, 1909-1949,” Journal of Monetary Economics, Vol. 56, pp. 450- 70. Alexopoulos, Michelle and Jon Cohen (2011) “Vol- umes of Evidence: Examining Technical Change Last Century Through a New Lens,” Canadian Journal of Economics, Vol. 44, No. 2, pp. 413-450. Alexopoulos, Michelle and Trevor Tombe (2011) “Managerial Knowledge and Canadian Produc- tivity,” University of Toronto Working Paper. Baldwin, John R., Wulong Gu and Beiling Yan (2008) “Relative Multifactor Productivity Levels in Canada and the United States: A Sectoral Analysis,” Statistics Canada Research Paper, Cata- logue No. 15-206-X, No. 019 Basu, S., J. Fernald, N. Oulton and S. Srinivasan (2003) “The Case of the Missing Productivity Growth: or, Does Information Technology Explain Why Productivity Accelerated in the US but not the UK?” NBER Macroeconomics Annual. Gospodinov, N., A. Maynard and E. Pesavento (2011) “Sensitivity of Impulse Responses to Small Low-Frequency Comovements: Reconcil- ing the Evidence on the Effects of Technology Shocks,” Journal of Business & Economic Statistics, Vol. 29, No. 4, pp. 455-467. Khan, H. and M. Santos (2002) “Contribution of ICT use to Output and Labour Productivity Growth in Canada,” Bank of Canada Discussion Paper 2002-07. Library of Congress Classification. A-Z, Library of Congress, Cataloguing Distribution Services, Washington, D.C. Various years. Maddison, A. (2010) Statistics on World Population, GDP, Per Capita GDP 1-2008 AD. Groningen Growth and Development Center. Madsen, J. (2007) “Technology spillover through trade and TFP convergence: 135 years of evi- dence for the OECD countries,” Journal of Inter- national Economics, Vol. 72, No. 2, pp. 464-480. Oliner, S., D. Sichel and K. Stiroh (2007) “Explain- ing a Productive Decade,” Brookings Papers on Economic Activity, Vol. 38, No. 1, pp. 81-152. Rao, Someshwar (2011) Cracking Canada’s Productiv- ity Conundrum. IRPP Study 25 (Montreal: Insti- tute for Research on Public Policy). Rao, Someshwar and Jianmin Tang (2001) “The Contribution of ICTs to Productivity Growth in Canada and the United States in the 1990s,” International Productivity Monitor, Number 3, Fall, pp. 3-18. Rao, Someshwar, Jianmin Tang and Weimin Wang (2004) “Measuring the Canada-U.S. Productiv- ity Gap: Industry Dimensions,” International Productivity Monitor, No. 9, Fall, pp. 3-14. Rao, Someshwar, Jianmin Tang and Weimin Wang (2008) “What Explains the Canada-U.S. Labour Productivity Gap?” Canadian Public Policy, Vol. 34, No. 2, pp. 163-92. Sharpe, Andrew (2006) “The Relationship between ICT Investment and Productivity in the Cana- dian Economy: A Review of the Evidence,” Research Report 2006-05 (Ottawa: Centre for the Study of Living Standards) Sharpe, Andrew (2010) “The Canada-U.S. ICT Investment Gap in 2008: Gains in Communica- tions Equipment and Losses in Computers,” CSLS Research Note 2010-01(Ottawa: Centre for the Study of Living Standards). Shea, John (1998) “What Do Technology Shocks Do?” NBER Macroeconomics Annual, Vol. 13, pp. 275-310. Stiroh, K. J. (2002) “Information Technology and the US Productivity Revival: What Do the Industry Data Say?” American Economic Review, Vol. 92, No. 5, pp. 1559-76. Van Ark, B., Robert Inklaar and Robert H. McGuckin (2003) "The Contribution of ICT- Producing and ICT-Using Industries to Produc- tivity Growth: A Comparison of Canada, Europe and the United States," International Productivity Monitor, Number 6, Spring, pp. 56-63. I N T E R N A T I O N A L P R O D U C T I V I T Y M O N I T O R 29 http://ideas.repec.org/a/bes/jnlbes/v29i4y2011p455-467.html http://ideas.repec.org/a/bes/jnlbes/v29i4y2011p455-467.html http://ideas.repec.org/s/bes/jnlbes.html http://ideas.repec.org/a/bin/bpeajo/v38y2007i2007-1p81-152.html http://ideas.repec.org/a/bin/bpeajo/v38y2007i2007-1p81-152.html http://ideas.repec.org/s/bin/bpeajo.html http://ideas.repec.org/s/bin/bpeajo.html http://ideas.repec.org/a/sls/ipmsls/v6y20035.html http://ideas.repec.org/a/sls/ipmsls/v6y20035.html http://ideas.repec.org/s/sls/ipmsls.html http://ideas.repec.org/s/sls/ipmsls.html Appendix A Sample Marc Record and Associated Online Display Marc Record: 00992cam 2200253 a 45000010008000000050017000080080041000250350021000 66906004500087010001700132020004600490400018001 95050002700213082001600240100002700256245008900 28326001220037230000340049450000200052865000430 0548650004300591630003800634991006600672- 4768599-19930312102159.8-860214s1986 waua 001 0 eng - 9(DLC) 86002512- a7bcbccorignewd1eocipf19gy-gen- catlg- a86002512 - a0914845705 (pbk.) :c$17.95 ($27.95 Can.)- aDLCcDLCdDLC-00aQA76.8.I2594bA541986- 00a005.265219-1 aAndrews, Nancy,d1945--10aWindows :bthe official guide to Microsoft's operatingenvironment / cNancy Andrews.- aRedmond, Wash. :bMicrosoft Press ;a[New York] :bDistributed to the book trade in the U.S. by Harper & Row,cc1986.- axii, 292 p. :bill. ;c24 cm.- aIn- cludes index.- 0aIBM Personal Computer XTxProgramming.- 0aIBM Personal Computer ATxProgramming.-00aMicrosoft Windows (Computer file)- bc-GenCollhQA76.8.I2594iA54 1986p00034791090tCopy 1wBOOKS Online display of information in Marc Record: Windows: the official guide to Microsoft's operating environment 4768599 LC control no.: 86002512 Type of material: Book (Print, Microform, Electronic, etc.) Personal name: Andrews, Nancy, 1945- Main title: Windows : the official guide to Microsoft’s operating environment / Nancy Andrews. Published/Created: Redmond, Wash. : Microsoft Press ; [New York] : Distributed to the book trade in the U.S. by Harper & Row, c1986. Description: xii, 292 p. : ill. ; 24 cm. ISBN: 0914845705 (pbk.) : $17.95 ($27.95 Can.) Notes: Includes index. Subjects: Microsoft Windows (Computer file) IBM Personal Computer XT --Programming. IBM Personal Computer AT --Programming. LC classification: QA76.8.I2594 A54 1986 Dewey class no.: 005.265 Appendix B Computer Classifications in Library of Congress Classification and the Dewey Decimal System In the Library of Congress, the books pertain- ing to Computers and Computer science are typically listed under the subclass QA Mathe- matics. For the indicators created in the paper, we used books classified under the QA 75 and QA 76 groups. Specifically, these are the books c l a s s i f i e d u n d e r Q A 7 5 - 7 6 . 9 5 C a l c u l a t i n g machines which include titles on electronic computers, computer science, and computer software. The indicators also include books classified under the Dewey Decimal System classifications 004 – 006. The items under these designations are grouped as follows: 004 Data processing & computer science 005 Computer programming, programs & data 006 Special computer methods 3 0 N U M B E R 2 3 , S P R I N G 2 0 1 2 Table 2 P-values of Granger Causality Tests Table 1 Comparison of Dates for Selected Computer Innovations The Effects of Computer Technologies on the Canadian Economy: Evidence from New Direct Measures Michelle Alexopoulos and Jon Cohen University of Toronto Abstract New indicators of technical change in the field of computers based on new titles held by Canadian libraries are presented, and a... Résumé De nouveaux indicateurs du changement technique dans le domaine de l'informatique, basés sur les nouveaux titres des bibliothèqu... The Indicators Output, Productivity, and Technical Change Canadian Productivity and the U.S.-Canada Productivity Gap Conclusion References Appendix A Sample Marc Record and Associated Online Display Appendix B Computer Classifications in Library of Congress Classification and the Dewey Decimal System Table 3 Per cent of Variation Due to Technology in Two Variable VARs Table 4 Per cent of Variation Due to Computer Technologies in the Tri-variate VARs Chart 1 Indicators of Computer-Related Titles by Copyright Date Chart 2 Fraction of English-Language Held Computer Books in Canada Published in the United States (per cent) Chart 3 Impulse Response Functions- Bi-Variate VAR Chart 4 Impulse Response Function: Tri-Variate VAR Chart 5 Total Factor Productivity in Canada and the United States, 1950-2005 (1950=100) Chart 6 New Computer Books Held in Canada and the United States by Copyright Date Chart 7 New Computer Titles by Copyright Date << /ASCII85EncodePages false /AllowTransparency false /AutoPositionEPSFiles true /AutoRotatePages /All /Binding /Left /CalGrayProfile (Dot Gain 20%) /CalRGBProfile (sRGB IEC61966-2.1) /CalCMYKProfile (U.S. Web Coated \050SWOP\051 v2) /sRGBProfile (sRGB IEC61966-2.1) /CannotEmbedFontPolicy /Warning /CompatibilityLevel 1.4 /CompressObjects /Tags /CompressPages true /ConvertImagesToIndexed true /PassThroughJPEGImages true /CreateJDFFile false /CreateJobTicket false /DefaultRenderingIntent /Default /DetectBlends true /ColorConversionStrategy /LeaveColorUnchanged /DoThumbnails false /EmbedAllFonts true /EmbedJobOptions true /DSCReportingLevel 0 /SyntheticBoldness 1.00 /EmitDSCWarnings false /EndPage -1 /ImageMemory 1048576 /LockDistillerParams false /MaxSubsetPct 100 /Optimize true /OPM 1 /ParseDSCComments true /ParseDSCCommentsForDocInfo true /PreserveCopyPage true /PreserveEPSInfo true /PreserveHalftoneInfo false /PreserveOPIComments false /PreserveOverprintSettings true /StartPage 1 /SubsetFonts true /TransferFunctionInfo /Apply /UCRandBGInfo /Preserve /UsePrologue false /ColorSettingsFile () /AlwaysEmbed [ true ] /NeverEmbed [ true ] /AntiAliasColorImages false /DownsampleColorImages true /ColorImageDownsampleType /Bicubic /ColorImageResolution 300 /ColorImageDepth -1 /ColorImageDownsampleThreshold 1.50000 /EncodeColorImages true /ColorImageFilter /DCTEncode /AutoFilterColorImages true /ColorImageAutoFilterStrategy /JPEG /ColorACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /ColorImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000ColorACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000ColorImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasGrayImages false /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /GrayImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000GrayACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000GrayImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasMonoImages false /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict << /K -1 >> /AllowPSXObjects false /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile () /PDFXOutputCondition () /PDFXRegistryName (http://www.color.org) /PDFXTrapped /Unknown /Description << /FRA /ENU (Use these settings to create PDF documents with higher image resolution for improved printing quality. The PDF documents can be opened with Acrobat and Reader 5.0 and later.) /JPN /DEU /PTB /DAN /NLD /ESP /SUO /ITA /NOR /SVE /KOR /CHS /CHT >> >> setdistillerparams << /HWResolution [2400 2400] /PageSize [612.000 792.000] >> setpagedevice