Criscuolo Presentation14 November by 26WIJj8

VIEWS: 4 PAGES: 28

									Innovation and Productivity
           Chiara Criscuolo
  Centre for Economic Performance
    London School of Economics

     Paris 14th November 2007
                        Motivation
• Technological knowledge is a driver of productivity
• Not the only one:
   – Organisational knowledge/ Non technological innovation (Topic 4)
• “Productivity isn’t everything, but in the long run it is
  almost everything” (Paul Krugman)
      • Productivity growth drives growth of real wages
      • Productivity growth can be used to finance government
        expenditure
• Therefore understanding differences in innovation
  drivers (see also on IPRs and innovation) and outcomes
  and differences in the relationship between innovation
  and productivity will help us understand differences in
  performance across countries.
                     Background
• Use of comparable cross-country data on innovation
  inputs and outputs
• We need firm level data: it is firms that innovate not
  countries or industries
• Large evidence of substantial and persistent intra-
  industry heterogeneity in performance and
  characteristics across firms. (within countries and
  within industries!)
• Hurdle: firm level datasets compiled by national
  statistical offices must comply with confidentiality
  protection.
   – NOT possible to POOL the data together.
   – NEED a TEAM/Network willing to conduct the same
     analysis in each separate country on similar data with
     “same” variables.
   The project would have impossible...
• …without the invaluable and extensive efforts and VOLUNTARY contributions of
  country researchers and their institutions
        1.    Australia: David (ABS)
        2.    Austria: Martin Berger
        3.    Belgium: Jeoffrey Malek
        4.    Brazil: Bruno Araújo and João De Negri
        5.    Canada: Petr Hanel and Pierre Therrien
        6.    Denmark: Carter Bloch and Ebbe Graversen
        7.    Finland: Mariagrazia Squicciarini Olavi Lehtoranta Mervi Niemi
        8.    France: Stephane Robin and Jacques Mairesse
        9.    Germany: Bettina Peters
        10.   Italy: Francesco Crespi Mario Denni Rinaldo Evangelista and Mario Pianta
        11.   Japan: Tomohiro Ijichi (could not participate for data problems)
        12.   Korea: Seok-Hyeon Kim
        13.   Luxembourg: Anna-Leena Asikainen
        14.   Netherlands: George van Leeuwen, Pierre Mohnen, Michael Polder, Wladimir Raymond
        15.   New Zealand: Richard Fabling
        16.   Norway: Svein Olav Nås and Mark Knell
        17.   Sweden: Hans Loof
        18.   Switzerland: Spyros Arvanitis
        19.   United Kingdom: Chiara Criscuolo

        A big THANK YOU to all!
       And without Innovation Surveys
• What do we like about them?
• rich source of information on innovation activities of firms:
  innovation inputs/outputs external knowledge sources, factors
  hampering/fostering innovation and methods of protection of
  innovation.
• unique tool to study the innovation process within and across
  countries.
• harmonized across EU countries and similar across other
  countries
• in most countries can be matched to other data sources:
   – check the quality of the data and the representativeness of the sample
   – produce richer set of variables with which to investigate more questions
• it covers both the manufacturing and the services sectors
   – particularly suited to study innovation in the services sector
  What we don’t like about innovation surveys
• Although harmonized some cross-country differences
  arise along various dimensions:
   –   Sampling frames
   –   Sector covered
   –   Mandatory/voluntary
   –   Questions asked
   –   Phrasing and ordering of questions
   –   Presence of filter questions
• Some of these issues can be accounted for in the
  analysis; but some of them cannot
• Limit the breadth of the analysis: “minimum
  common denominator” variables
 Issues when doing comparative analysis

• Knowing the differences across different
  countries
• Which variables/information are the
  “minimum common denominator”?
• Define variables in the same way across
  countries
• Estimate the same model/specification across
  countries
            …“Selection” issue
• In most countries innovation surveys “concentrates”
  on innovators (“yes to initial questions: product,
  process, ongoing and abandoned)
• those who have responded “no” to these questions
  (non-innovators) and do not have to respond to most
  of the rest of the survey.
• contains little information on non-innovators
• “Selection” issue that needs to be dealt with in the
  econometric analysis.
          …poor productivity measure
• For almost all countries innovation surveys contain
   information only on one year (cross-sectional) information on
   sales (turnover) and employment.
• Sales/turnover is different from gross output
• Only cross-sectional information, we cannot say anything on
   growth
• Sales per employee is a very rough measure of productivity.
   Ideally we would want value added per employee (Labour
   Productivity measure) or Multi Factor Productivity measures
To improve on productivity measurement innovation survey
   must be combined with other production panel datasets
   (follow-up of current project?)
  But what did we manage to do?
• We estimate:
  – the same model across 18 countries
  – extensions and variants of the model for smaller
    subsets of countries
• Within countries we distinguished where
  possible
  – Small vs large firms
  – Manufacturing vs services
  – Control also for human and physical capital
  And HOW did we manage to do?
• We agreed on an acceptable model which could be run
  in all participating countries:
   – Minimum common denominator variables
   – Corrected for selection
   – Simple productivity measure
• We had several meetings during the past 12 months to
  agree on the model
• Wrote programs/routines that could be easily
  estimated across countries (STATA do and ado files:
  Innov4Prod.do and *.ado available) and readme files
  that helped countries implement the routines
   A brief outline of the model followed…
…To estimate the effects of innovation on
  productivity controlling for selection and
  endogeneity
Following the Crepon-Duguet Mairesse “tradition”
  we estimate a recursive 3 stage/4 equation
  model:
• 1st innovation equation
• 2nd innovation input equation
• 3rd innovation output equation
• 4th productivity equation
                                The model
• 3 stage with 4 equations
• 1st stage explains firms’ decision whether to engage in innovation
  activities or not and the decision on the amount of innovation
  expenditure
    Prob(innovation=1)=f(size; Group; Foreign Market, Obstacles to innovation due
    to knowledge; costs and market; industry dummies)
    Ln(innovation expenditure per employee)=f(Group; Foreign Market;
    Cooperation; Financial Support; industry dummies)
• In the 2nd stage we estimate the knowledge production function where
  innovative sales depends on investment in innovation. Ln(innovative sales
    per employee)=f(Innovation expenditure; Size; Group; process innovation;
    Cooperation with clients; suppliers; other private and public agents; industry
    dummies; inverse Mills ratio [to correct for selection])
•   The 3rd stage we estimate the innovation output productivity link using
    an augmented Cobb-Douglas production function using IV.
    Ln(sales per employee)=f(Size; Group; Process Innovation; log innovative sales
    per employee; inverse Mills ratio; industry dummies [Human Capital and
    Physical Capital])
                      Obstacles to innovation
8.2 During the three years 2002 to 2004, how important were the following
    factors for hampering your innovation activities or projects or influencing a
    decision not to innovate?
                                                                                Degree of importance
                                                                                                Factor not
                                                                        High   Medium   Low    experienced
            Lack of funds within your enterprise or group                                           
 Cost
 factors    Lack of finance from sources outside your enterprise                                    
            Innovation costs too high                                                               

            Lack of qualified personnel                                                             
Knowledge
factors     Lack of information on technology                                                       
            Lack of information on markets                                                          
            Difficulty in finding cooperation partners for innovation                               

 Market     Market dominated by established enterprises                                             
 factors    Uncertain demand for innovative goods or services                                       
                               Cooperation partners
Please indicate the type of co-operation partner and location        (Tick all that apply)

                                                                          [Your          Other    United   All other
     CUSTOMERS
                                                                         country]       Europe*   States   countries
     C. Clients or customers                                                                               


                                                                          [Your          Other    United   All other
     SUPPLIERS
                                                                         country]       Europe*   States   countries
     B. Suppliers of equipment, materials, components, or software                                         


                                                                          [Your          Other    United   All other
     PUBLIC
                                                                         country]       Europe*   States   countries
     F. Universities or other higher education institutions                                                
     G. Government or public research institutes                                                           


                                                                          [Your          Other    United   All other
     OTHER PRIVATE
                                                                         country]       Europe*   States   countries
     D. Competitors or other enterprises in your sector                                                    
     E. Consultants, commercial labs, or private R&D institutes                                            
  Extensions/variations of the model ...
• Some countries could add Human Capital (H); Physical Capital
  (K) and Materials (M) in the productivity equation:
   – Austria (H,K); Belgium (H,K); Brazil (H,K,M); Canada (H,K); Finland
     (H,K,M); Germany (H,K,M); New Zealand (K,M); Switzerland (H,K,M);
     UK (H)
• Sales per employee is a very rough measure of
  productivity. Ideally we would want value added per
  employee (Labour Productivity measure) or Multi Factor
  Productivity measures
• To do this innovation surveys must be combined with
  other production datasets. (follow-up of current
  project?)
…Extensions/variations of the model ...
• Most countries estimate separately for small
  vs large firms and manufacturing vs services
  firms;
• Korea and Canada only manuf;
• Luxembourg serv.
• Standard size threshold 250 employees
• Different estimation strategies to deal with
  endogeneity of innovation expenditure
• Alternative definition of innovative firm
    …Extensions/variations of the model…
• Germany/Netherlands suggested a modification of
  the model to deal with endogeneity
• Canada: could only estimate on manufacturing and
  weighted regressions and no information on
  obstacles to innovation
• Austria: estimate it on CIS3 rather than CIS4
• Australia: no information on foreign market; inputed
  group information and 2005 rather than 2004
• New Zealand: again differences in variable definitions
• Switzerland no group variable
                    RESULTS
• “structural” model:
  – Heckman
  – Innovative sales eq.
  – Productivity eq.
• Some extensions:
  – Manufacturing vs services
  – Small (vs large)
  – Controlling for Human capital; Physical capital and
    Materials
         Controlling for Selection: innovation
          equation (Heckman selection eq.)

 Selection Eq.      GP      FOR_MKT       LEMP      HAKNOW HAMARKET HACOST          rho* Observations P-value LR test
    Austria       0.213*     0.454***    0.253***    -0.0765    -0.182  -0.00122 0.223      1001           0.226
    Belgium      0.198***    0.617***    0.267***     0.0427   -0.0500   0.455***   0.41    2695          0.0012
     Brazil      0.424***   -0.264***    0.123***    0.152***  0.131***   0.0320 2.019***   9384           0.000
    Canada       -0.105*     0.290***    0.140***                                 1.005***  5355           0.000
   Denmark       0.186**     0.637***    0.253***    0.243**    0.0288   0.391*** 0.324**   1729          0.0202
    Finland       0.0649     0.532***    0.254***    0.190**   0.259***  -0.0266 0.477***   2155         0.00178
    France       0.227***    0.778***    0.204***    0.201*** 0.0678*** 0.227*** 0.643***   18056          0.000
   Germany       0.144***    0.529***   0.0884***     0.0144    -0.107   0.173*** 0.256**   3242          0.0656
      Italy      0.203***    0.478***    0.185***    0.110*** -0.0680** 0.0908*** 0.753***  15915          0.000
 Luxembourg       0.267*     0.314**     0.248***     0.191     -0.101    0.359*   0.192     545           0.701
 Netherlands     0.164***    0.546***    0.213***    0.175***  -0.111**   0.0123 0.727***   6858           0.000
 New Zealand     0.113**     0.349***   0.0785***    0.0892*    0.0270   0.138*** 1.337***  3426           0.000
    Norway       -0.0724     0.643***    0.320***    0.301***   0.0478   0.301*** 0.739***  1852           0.000
United Kingdom   0.174***    0.464***   0.0468***    0.287***  0.0883** 0.0883** 0.189      11162          0.261
           Heckman outcome equation:
            innovation expenditure eq.
 Outcome eq.          GP    FOR_MKT COOP       FINSUP Observations
    Austria          0.161   0.737*** 0.408*** 0.746***     1001
    Belgium         0.233*   0.524*** -0.0205  0.714***     2695
     Brazil        0.875*** -0.204* 0.384*** 0.332***       9384
    Canada          0.145*   0.448*** 0.173**   0.183*      5355
   Denmark         0.477*** 0.762***   0.182   0.735***     1729
    Finland        0.260**    0.361*  0.495*** 0.460***     2155
    France         0.231*** 1.158*** 0.427*** 0.683***     18056
   Germany          0.0538 0.610*** 0.402*** 0.469***       3242
      Italy        0.268*** 0.511*** 0.310*** 0.412***     15915
 Luxembourg          0.212    0.434    0.102    0.352       545
 Netherlands       0.247*** 0.675*** 0.389*** 0.569***      6858
 New Zealand       0.664*** 0.740*** 0.225*** Confidential  3426
    Norway         -0.0436 0.706*** 0.354*** 0.657***       1852
United Kingdom      0.0508 0.513*** 0.377*** 0.537***      11162
  Careful: group and foreign market are not marginal effects
                   GP       LEMP
                                   Innovation Sales eq.
                                       PROCESS MILLSstrict COOP_client COOP_supplier COOP_private COOP_public   Obs
    Austria        0.32     0.0996        0.32     0.629      -0.142      0.0573        0.0828       -0.533      359
    Belgium        0.244    0.0432       0.0374  -0.0760      0.230*      -0.150       0.307**     -0.273**      411
     Brazil      0.748*** 0.136***       0.0580  0.747**     -0.0878       0.164        0.0340     0.313***     1954
    Canada        0.0717    0.0214     0.263***    0.244     -0.0676     -0.0250       0.281**       -0.145     2273
   Denmark       0.631*** -0.101*        0.0522  -0.0849       0.184      -0.111       -0.0260       0.0547      584
    Finland       0.264*    -0.172*     0.297**   -0.632      -0.189     0.00830         0.250      -0.0291      698
    France       0.552*** -0.0117       0.155**   0.0785      0.116*      0.0437       0.147**      -0.0437     2511
   Germany        0.0723    0.0367       0.0593 -0.469**     0.00700       0.127       -0.0507      -0.0601     1390
      Italy       0.0851   -0.0378     0.528***   -0.443       0.041       0.109        0.197*       0.0659      747
 Luxembourg        0.183   -0.284**    0.509*** -1.130*        0.144      -0.127        0.0246        0.224      207
 Netherlands     0.320*** -0.0301       -0.0125   -0.186      0.0481    -0.000654      -0.0957       0.0889     1374
 New Zealand     0.300*** -0.0885***   0.000126 0.548*      -0.278***     0.190*        -0.103      -0.0765      993
    Norway       0.239** 0.169**         0.147     0.219     -0.336*       0.304        0.0609     -0.00662      672
United Kingdom   0.225*** -0.0349*      0.125**   -0.163      0.0138      0.0660       0.00828     -0.00909     2989

                                              LRTOTPE                                            LRTOTPE
                       Austria                   0.0501     Germany                              0.340***
                      Belgium                   0.274***       Italy                             0.210***
                        Brazil                  0.354***  Luxembourg                             0.211***
                      Canada                    0.367***  Netherlands                            0.260***
                      Denmark                   0.141***  New Zealand                            0.516***
                       Finland                  0.218***     Norway                              0.471***
                       France                   0.214*** United Kingdom                          0.263***
                 Productivity equation
                  GP        LEMP      PROCESS MILLSstrict LISPE Observations
    Austria     0.137      -0.0236      -0.0103   -1.077**  0.401   359
    Belgium    0.289***    -0.0235       -0.125*    -0.240 0.467*** 411
     Brazil    0.183**    0.140***     -0.211***    -0.315 0.647*** 1954
    Canada     0.250***   0.0772**      -0.122** -0.00113 0.436***  2273
   Denmark     0.186**    0.0732***     -0.0405      0.146 0.345*** 584
    Finland    0.244***   0.0859**      -0.0677      0.135 0.314*** 698
    France     0.232***   0.0536***    -0.129***     0.089 0.474*** 2511
   Germany     0.0838**   0.0625***    -0.116***  -0.259** 0.500*** 1390
      Italy     0.093      0.00391      -0.192**   -0.312* 0.485*** 747
 Luxembourg 0.434***        0.0349        -0.142     0.170  0.226*  207
 Netherlands    0.0219    0.0902***     -0.0440    -0.0765 0.409*** 1374
 New Zealand 0.128**      0.0662***    -0.135***    -0.200 0.682*** 993
    Norway     0.256***     0.0407      -0.0716     -0.168 0.344*** 672
United Kingdom 0.150***   0.0580***    -0.121*** -0.272*** 0.550*** 2989
              …Manuf vs Serv; SME; HC
                          Manuf      Serv     small     Large      without     with HC
Australia     LISPE     0.399***    0.0155
              Process    -0.0443    0.515
Austria       LISPE       -0.35    0.394**   0.308**                0.401        0.334*
              Process     0.229    0.322**     0.157               -0.0103       -0.036
Belgium       LISPE     0.446***   0.525*** 0.607***   0.353***   0.467***     -0.00867
              Process   -0.170**   -0.0913    -0.109    -0.125     -0.125*       -0.121
Brazil        LISPE                         0.758***   0.589***   0.647***     0.117***
              Process                      -0.316***    0.0218    -0.211***     -0.0373
Canada        LISPE                         0.507***   0.368***   0.436***     0.380***
              Process                       -0.219**   -0.0703    -0.122**     -0.126**
Denmark       LISPE      0.439***   0.229   0.308***              0.345***         no
              Process     0.0536   -0.168* -0.0884                 -0.0405         no
Finland       LISPE      0.376***   0.213   0.289***              0.314***      -0.0929
              Process    -0.119*    0.105    -0.0941               -0.0677     -0.00189
France        LISPE      0.495*** 0.443*** 0.361***    0.605***   0.474***         no
              Process   -0.154*** -0.107* -0.115***    -0.0999*   -0.129***        no
Germany       LISPE      0.405*** 0.613*** 0.421***               0.500***     0.329***
              Process   -0.107*** -0.137 -0.0979**                -0.116***   -0.0877***
Luxembourg    LISPE               0.450***                          0.226*
              Process             -0.317**                          -0.142
Netherlands   LISPE      0.459*** 0.390*** 0.386***    0.429***   0.409***
              Process   -0.0803* 0.00225 -0.0507       -0.0116     -0.0440
New Zealand   LISPE      0.589*** 0.707*** 0.685***    0.639***   0.682***     0.245***
              Process    -0.0629 -0.186*** -0.133***   -0.192*    -0.135***    -0.0334
             Summary of results
• When significant, coefficients are surprisingly similar
• Serving a foreign market; being large and being part
  of a group are generally associated with higher
  probability of being innovative and financial support
  and cooperation activity with higher investment in
  innovation
• Using a selection model is appropriate for most
  countries (exc. Austria; Luxembourg and UK)
• In the innovation sales eq. the elasticity of innovative
  sales to innovation expenditure is mostly between
  20%-35%
• In productivity eq.: elasticity of productivity to
  product innovation is 30%-60%
   Some counterintuitive results
• Obstacles to innovation have mostly positive
  coefficients. More innovative firms try harder
  and therefore find more obstacles?
• Process innovation is mostly negative in
  productivity equation. Measurement issues?
  Adjustment costs? Firms in crisis more likely to
  do process innovation? (Possibly future
  work?)
   Lessons learned and the work
              ahead
• Surprising similar results across different
  countries
• Innovation matters for productivity
• High coordination costs!
• More work on the modelling and
  understanding the differences in results
• Extend the analysis to better measures of
  productivity: match CIS with production data
      THE END
     THANK YOU!


C.CRISCUOLO@LSE.AC.UK

								
To top