econstor Make Your Publications Visible.

Transcription

1 econstor Make Your Publcatons Vsble. A Servce of Wrtschaft Centre zbwlebnz-informatonszentrum Economcs Malranta, Mka; Rouvnen, Petr Workng Paper Informatonal moblty and productvty: Fnnsh evdence ETLA Dscusson Papers, The Research Insttute of the Fnnsh Economy (ETLA), No. 919 Provded n Cooperaton wth: Research Insttute of the Fnnsh Economy (ETLA), Helsnk Suggested Ctaton: Malranta, Mka; Rouvnen, Petr (2004) : Informatonal moblty and productvty: Fnnsh evdence, ETLA Dscusson Papers, The Research Insttute of the Fnnsh Economy (ETLA), No. 919 Ths Verson s avalable at: Standard-Nutzungsbedngungen: De Dokumente auf EconStor dürfen zu egenen wssenschaftlchen Zwecken und zum Prvatgebrauch gespechert und kopert werden. Se dürfen de Dokumente ncht für öffentlche oder kommerzelle Zwecke vervelfältgen, öffentlch ausstellen, öffentlch zugänglch machen, vertreben oder anderwetg nutzen. Sofern de Verfasser de Dokumente unter Open-Content-Lzenzen (nsbesondere CC-Lzenzen) zur Verfügung gestellt haben sollten, gelten abwechend von desen Nutzungsbedngungen de n der dort genannten Lzenz gewährten Nutzungsrechte. Terms of use: Documents n EconStor may be saved and coped for your personal and scholarly purposes. You are not to copy documents for publc or commercal purposes, to exhbt the documents publcly, to make them publcly avalable on the nternet, or to dstrbute or otherwse use the documents n publc. If the documents have been made avalable under an Open Content Lcence (especally Creatve Commons Lcences), you may exercse further usage rghts as specfed n the ndcated lcence.

2 ETLA ELINKEINOELÄMÄN TUTKIMUSLAITOS THE RESEARCH INSTITUTE OF THE FINNISH ECONOMY Lönnrotnkatu 4 B Helsnk Fnland Tel Telefax World Wde Web: Keskusteluaheta Dscusson papers No. 919 Mka Malranta Petr Rouvnen INFORMATIONAL MOBILITY AND PRODUCTIVITY FINNISH EVIDENCE* * Ths research s a part of the Wreless Communcaton Research Program (bre-etla.org) of BRIE, the Berkeley Roundtable on the Internatonal Economy at the Unversty of Calforna at Berkeley, and ETLA, The Research Insttute of the Fnnsh Economy. Fnancal support of the Fnnsh Natonal Technology Agency (Tekes) and Noka Corporaton s gratefully acknowledged. ISSN

3 MALIRANTA, Mka ROUVINEN, Petr, INFORMATIONAL MOBILITY AND PRODUCTIVITY FINNISH EVIDENCE. Helsnk: ETLA, Elnkenoelämän Tutkmuslatos, The Research Insttute of the Fnnsh Economy, 2004, 14 p. (Keskusteluaheta, Dscusson Papers, ISSN ; No. 919). ABSTRACT: Labor productvty effects of portablty as well as wrelne and wreless connectvty of nformaton and communcaton technology () are studed wth Fnnsh frm-level data. It s found that a computer wth only processng and storage capabltes boosts labor productvty by 9% (correspondng to 5% output elastcty), portablty by 32%, wrelne connectvty by 14%, and wreless connectvty by 6%. Fndngs are n lne wth prevous lterature and comparsons to costs suggest that frms equate margnal costs and returns. Whle ncreasng penetraton can no longer be a major source of productvty growth n developed economes, the studed new characterstcs can. Keywords: JEL codes: Productvty, Computer,, Informaton and communcaton technology, LAN, Local area network, Moblty, Portablty, Wreless D24, J24, L23, O33 MALIRANTA, Mka ROUVINEN, Petr, TIEDOLLISEN LIIKKUVUUDEN TUOTTA- VUUSVAIKUTUKSET SUOMESSA. Helsnk: ETLA, Elnkenoelämän Tutkmuslatos, The Research Insttute of the Fnnsh Economy, 2004, 14 s. (Keskusteluaheta, Dscusson Papers, ISSN ; No. 919). TIIVISTELMÄ: Tässä paperssa tutktaan teto- ja vestntäteknologalatteden kannettavuuden sekä langallsten ja langattomen tetolkenneomnasuuksen tuottavuusvakutuksa suomalasella yrtysanestolla. Tulokset osottavat, että dgtaalsen tedon prosenssont- ja tallennusomnasuudet nostavat työn tuottavuutta 9 % (vastaten 5 % tuotosjoustoa), kannettavuus 32 %, langallnen tetolkenne 14 % ja langaton 6 %. Tulokset ovat yhdenmukasa aemman krjallsuuden kanssa ja nden mukaan yrtykset nvestovat tetoteknkkaan tasolla, jolla rajakustannukset ja -hyödyt muodostuvat yhtäsuurks. Tetoteknkan käytön levämnen snänsä e enää vo olla tuottavuuskasvun lähde kehttynessä talouksssa, mutta kästellyt uudet omnasuudet vovat. Avansanat: Tuottavuus, tetokone,, teto- ja vestntäteknologa, LAN, lähverkko, lkkuvuus, kannettavuus, langaton.

4 INTRODUCTION Informaton and communcaton technology () devces have become ncreasngly portable and (nter)connectable at least snce the md-1990s, whch n many crcumstances have made t possble to create and access dgtally coded nformaton ndependent of tme and place. Does ths ncreasng nformatonal moblty have consequences on labor productvty? In order to answer the queston, ths paper operatonalzes the concept of nformatonal moblty and quantfes ts effects on labor productvty n Fnnsh busness. In what follows, portablty refers to the use of laptop personal computers (PCs), tablet PCs, or smlar portable data processng and storage devces (due to data lmtatons, however, personal dgtal assstants, moble phones, or smlar handhelds are excluded). Connectvty refers to wrelne (e.g., va a local area network, LAN) or wreless (e.g., va a publc access moble telephony network or a wreless local area network, WLAN) access to remotely stored data. Olaru (2003) notes that both portablty and wreless connectvty have grown explosvely n recent years n terms of both avalable technologes and servces offered. Perhaps the most mportant drvers have been rapdly droppng prces and user costs. 1 Leadng trade publcatons, see e.g. Gartenberg (2003), have repeatedly suggested albet wthout emprcal evdence that portablty and wreless connectvty offer sgnfcant cost savngs and boost productvty. Anecdotal support can be found n busness press; see e.g. Green (2003) and Nasaw (2003). For example Altmer et al. (2002) have document the benefts of portablty and wreless connectvty n health care and Zurta and Nussbaum (2003) n educaton. Whle studes n other or across sectors are nearly non-exstent, portablty and wreless connectvty mght boost productvty n vrtually any actvty where mmedate nformaton storage, processng, retreval, and exchange are benefcal. As ths s the case n many management, sales, and logstcs actvtes, most busnesses could be affected. 1 Whle relable statstcs are unavalable, for nstance Cahners InStat/MDR (as cted n Rao & Parkh, 2003) states that the U.S. WLAN equpment sales ncreased by 65% n Chwelos (2003) fndngs suggest that the qualty-adjusted prce of a laptop dropped roughly twce as fast as that of a desktop computer n the 1990s. 1

5 MODEL The three basc prncples of Lancaster s (1991) consumer theory may be gven a dual renterpretaton n a producton context as follows: (1.) goods do not boost productvty they possess characterstcs that do, (2.) goods may possess many characterstcs and many of them are shared by more than one good, and (3.) goods used n conjuncton wth others may possess characterstcs dfferent from the same goods used separately. Any good may be seen as a bundle of ts characterstcs, each havng a separate productvty effect. Even though as noted by Gordon (2000) dmnshng returns on mght be dscovered rather soon as ts usage expands, t s possble that new characterstcs added to old goods and new goods workng n conjuncton wth older ones may contnuously shft the productvty fronter. The goal here s to capture the productvty effects of varous characterstcs on an ndvdual worker. The problem s, however, that these characterstcs or more precsely, the goods that bundle them are only observed at the frm level,.e., data s grouped. Assumng that workers are after controllng for observable ndvdual qualtes perfect substtutes, a frm-level model revealng these effects can nevertheless be devsed and estmated. An extended Cobb-Douglas producton functon of frm can be wrtten as β β Y = A K L Z β (1) K L Z where Y s net output, A s dsemboded technology, K s captal, L s labor, and Z s a vector of other relevant frm and ndvdual qualtes. Workers may have dfferent margnal productvtes dependng on the set of characterstcs they use. Let L be a vector ndcatng the number of workers usng each of the characterstcs of nterest. Addng ths to (1) yelds β L K, Z Y AK L 1 β L = + θ L Z β, (2) where θ s a parameter vector capturng the possble addtonal productvty effects assocated wth the characterstcs. Slght manpulaton yelds a labor productvty specfcaton Y K L = A θ + ( + ) L + β Z, (3), ln ln βkln βlln 1 βk βl 1 ln ln L L L Z 2

6 where ( β β ) + 1lnL accounts for devatons from constant returns to scale. Approxmatng K L L, L, ln 1+ θ wth θ L yelds L Y K L + + θ + ( + ) + β Z, (4), ln ln A βkln βl βk βl 1 ln Lt Zln L L L whch s estmable once the varables and a stochastc error structure have been specfed. 2 DATA After elmnatng a few unusable observatons (see below), the sample for year 2001 conssts of 2,358 manufacturng and servce frms that are left after lnkng the survey data wth Fnancal Statements and Employment Statstcs as well as Busness Regster data. Whle Statstcs Fnland has conducted ts surveys snce 1998, only the year 2001 survey ncludes suffcent nformaton for the solaton of the desred characterstcs. Table 1 lsts the bundles and the characterstcs they possess. Note that wreless s seen as an extenson of wrelne connectvty, whch becomes mportant upon nterpretng the results the reference for WLAN s a smlar computer wth LAN. As Table 1 suggests, portablty and connectvty cannot exst wthout processng and storage capabltes. As the characterstcs are always emboded n one or more of the bundles, ther effects cannot be estmated drectly. By comparng two bundles that are dentcal n other respects except the characterstc to be solated, ts effects can nevertheless be assessed. The possble comparsons for solatng the characterstcs can be found at the bottom of Table 1: portablty can be solated va three dfferent comparsons; both wrelne and wreless connectvty va two. 2 Greenan and Maresse (2000) consder a smlar model n and, e.g., Dearden et al. (2000) and Ilmakunnas and Malranta (2003) n other contexts. 3

7 Table 1. The bundles, ther characterstcs and ways of dstngushng them. Bundle d. bundle: Processng and storage capab. Portablty characterstc: Wrelne connectvty Wreless connectvty 0 Not usng computer at work a Desktop Yes b Laptop Yes Yes c Desktop wth LAN Yes Yes d Laptop wth LAN Yes Yes Yes e Desktop wth WLAN Yes Yes Yes f Laptop wth WLAN Yes Yes Yes Yes Characterstc dstngushable a 0 b a c a e c by comparng bundle(s): d c d b f d f e Note: The emprcal defntons of the bundles dscussed n the data secton. In some estmatons below varous ways of calculatng bundle coeffcents are constranted to be equal. The emprcal defnton of the bundles s complcated by the fact that only the workers usng desktops and laptops are avalable as fractons; the use of LAN and WLAN are only observed as frm-level dummes. Thus, the bundles are derved under the assumpton that mplementng a LAN or WLAN nvolve solely a fxed cost, n whch case every computer s connected once a network s ntroduced at the frm level. Ths s not entrely unreasonable and the relatvely large data set used n the analyss allevates problems that ths mght cause. The possble practcal consequences of ths assumpton are studed n the Appendx. Table 2 lsts the varables along wth weghted descrptve statstcs. One tenth of employment n Fnnsh busness uses a non-connected desktop and a lttle over one per cent a nonconnected laptop at work. Roughly one thrd uses a LAN-connected desktop and one tenth a LAN-connected laptop at work. WLAN usage s qute rare. An extensve set of control varables n employed n order to avod dscoverng spurous relatonshps. Earler work by Malranta and Rouvnen (2004) has shown that controllng for ndvduals educatonal backgrounds s partcularly mportant. Besdes the varables n Table 2, also a constant term as well as 40 NACE rev. 1 two-dgt ndustry and 20 NUTS level-three regonal dummes are ncluded. 4

8 Table 2. Varable defntons and descrptve statstcs. Varable Descrpton Type 1 Source 2 Mean St. dev. Mn. Max. CD: ln(value added/labor) Log of real value added per labor nput R FSS : Desktop Sh. of workers w. non-connected desktop F : Laptop Sh. of workers w. non-connected laptop F : Desktop+LAN Workers w. LAN-connected desktop F : Laptop+LAN Workers w. LAN-connected laptop F : Desktop+WLAN Workers w. WLAN-connected desktop F : Laptop+WLAN Workers w. WLAN-connected laptop F CD: ln(captal/labor) Real physcal captal stock per labor nput R FSS CD: ln(labor) Real labor nput L FSS Frm: young Frm's establshments' avg. age 5 yrs or less D BR Frm: old Frm's establshments' avg. age 15 yrs or more D BR Frm: mult-est. Mult-establshment frm D BR Edu: lo tech Workers w. lower (bachelor eq.) techncal ed. F ES Edu: m tech Workers w. mddle (master eq.) techncal ed. F ES Edu: h tech Workers w. hgher (doctor eq.) techncal ed. F ES Edu: lo non-tech Workers w. lower (bachelor eq.) non-tech ed. F ES Edu: m non-tech Workers w. mddle (master eq.) non-tech ed. F ES Edu: h non-tech Workers w. hgher (doctor eq.) non-tech ed. F ES Labor: young Workers that are under 35 years old F ES Labor: old Workers that are over 45 years old F ES Labor: female Female workers F ES Note: Although not lsted above, all estmatons also nclude a constant term as well as NACE rev. 1 two-dgt ndustry and NUTS level-three regonal dummes. 1 Varable types: D = dummy (1 f true, 0 otherwse), F = fracton (share of all workers at the frm), L = natural logarthm, and R = natural logarthm of a rato. 2 Sources: BR = Busness Regster, ES = Employment Statstcs, FSS = Fnancal Statements Statstcs, and = survey. ESTIMATION Snce labor weghts are employed, the domnatng effect of very large frms on the results s an ssue of concern. Thus, 23 frms wth over two thousand employees are elmnated. Outlers are elmnated by usng the standardzed or Pearson resduals: a prelmnary regresson s performed and 23 observatons wth the standardzed resduals over four standard devatons away from the mean are dropped, whch wth normally dstrbuted errors s roughly equvalent of elmnatng 3 out of 100,000 observatons. Whte (1980) heteroskedastcty consstent standard devatons are reported. Consequences of these choces are dscussed n the Appendx. Three sets of regresson results are reported n Table 3: unconstranted ordnary least squares (OLS) n Column (1), OLS wth the constrant that LAN should have a symmetrc effect n Column (2) (.e., the followng constrant s mposed on the bundle coeffcents: 5

9 c a= d b; also mples a constrant on portablty), and OLS wth the constrants that both LAN and WLAN should have symmetrc effects n Column (3) (.e., e c= f d also mposed). As can be seen, mposng constrants leads to gans n effcency but qualtatvely the results reman the same. The null hypothess that constrants are vald cannot be rejected. Table 3. Estmaton results of the labor productvty equaton (the bundle coeffcents not adjusted for the output elastcty of labor). ln(value added/labor) (1) (2) (3) regressed on Unconstranted LAN constr. W/LAN constr. : Desktop * ** (0.057) (0.040) (0.040) : Laptop *** *** *** (0.145) (0.067) (0.065) : Desktop+LAN *** *** *** (0.054) (0.034) (0.034) : Laptop+LAN *** *** *** (0.110) (0.060) (0.059) : Desktop+WLAN ** *** *** (0.092) (0.053) (0.043) : Laptop+WLAN ** *** (0.256) (0.121) (0.065) CD: ln(captal/labor) *** *** *** (0.015) (0.007) (0.007) CD: ln(labor) (0.012) (0.008) (0.008) Frm: young (0.060) (0.032) (0.032) Frm: old (0.029) (0.019) (0.019) Frm: mult-est *** *** *** (0.032) (0.020) (0.020) Edu: lo tech (0.122) (0.103) (0.103) Edu: m tech * ** ** (0.143) (0.099) (0.099) Edu: h tech ** *** *** (0.251) (0.157) (0.155) Edu: lo non-tech (0.142) (0.116) (0.116) Edu: m non-tech ** *** *** (0.142) (0.098) (0.098) Edu: h non-tech *** *** *** (0.224) (0.155) (0.155) Labor: young ** *** *** (0.159) (0.084) (0.084) Labor: old * * (0.117) (0.087) (0.087) Labor: female *** *** *** (0.079) (0.054) (0.054) Adjusted R-squared 0.49 H0: Constrant(s) are vald (t-test) (0.172) (0.309) Note: Also ncludng (not reported due to space lmtatons) a constant term as well as NACE rev. 1 two-dgt ndustry and NUTS level-three regonal dummes. Weghted OLS wth Whte (1980) heteroskedastcty consstent standard errors n the parentheses. A subsample of 2,358 year 2001 observatons wth very large frms and outlers elmnated as dscussed n the text. ***, **, *, and + respectvely ndcate sgnfcance at 1, 5, 10, and 15 per cent level. 6

10 The results n Table 3 do not take nto account the output elastcty of labor,.e., βlθ rather than θ s reported. In the upper secton of Table 4 the bundle coeffcents have been adjusted for the output elastcty of labor by dvdng through wth β L to get θ. Table 4. The bundle coeffcents (adjusted for the output elastcty of labor) and the estmated effects of the characterstcs. Labor productvty effects of the bundles (1) (2) (3) Unconstranted LAN constr. W/LAN constr. (a) : Desktop 8% 9% * 9% ** (b) : Laptop 52% *** 44% *** 42% *** (c) : Desktop+LAN 23% *** 23% *** 24% *** (d) : Laptop+LAN 58% *** 58% *** 56% *** (e) : Desktop+WLAN 20% ** 20% *** 15% *** (f) : Laptop+WLAN 30% 30% 48% *** Labor productvty effects of the characterstcs (also see Table 1) (1) (2) (3) Unconstranted LAN constr. W/LAN constr. Processor, storage a 0 8% 9% * 9% ** b a 44% ** Portablty d c 34% *** 35% *** 32% *** f e 10% Wrelne connectvty c a 15% *** d b 6% 14% *** 14% *** Wreless connectvty e c -27% -28% ** f d -3% -3% -8% ** Note: Calculatons wth the delta method. ***, **, and *, respectvely ndcate sgnfcance at 1, 5, and 10 per cent level. The lower secton of Table 4 reports the effects for the characterstcs. The results wth respect to wreless connectvty are somewhat unstable, although t s perhaps not surprsng that tryng to nfer ts productvty effect by comparng two otherwse smlar desktops does not generate ntutve results, as wth an nherently mmoble computer the benefts of wreless over wrelne connectvty are perhaps prmarly lmted to arguably faster network deployment as well as to ncreased flexblty n offce floor plan desgn and ts alternatons. As also the test statstcs ndcate that the constrants mposed are vald (see above), the rghtmost Columns (3) 7

11 n Tables 2, 3, and 4 are the preferred set of results. The above results suggest that processng and storage capabltes themselves ncrease a workers productvty by 9%. Portablty boosts productvty by nearly one thrd. Wrelne connectvty boosts productvty 14% and wreless by 6%. 3 DISCUSSION The results derved n the above secton are not drectly comparable to the prevous lterature, as the output elastcty of s the most commonly reported measure n these contexts. As nvestment flows and captal stocks are unobserved n the data, alternatve ways of dervng the results cannot be compared drectly. In any case the above approach s arguably better, as they drectly measure usage as well as avod the mmense dffcultes n constructng real nvestment flows and captal stocks. The relatonshp of the above results to those found n prevous lterature can nevertheless be studed by consderng a smplfed verson of the above model Takng a logarthm yelds the dervate of whch s β ( θ ) L Y = L+ L. (5) lny β ln ( L θ L ) = +, (6) L dy θdl = βl Y L+ θ L, (7) whch can be used to derve output elastcty dy L θ χ = = β = β L L dl Y ( L + θ L ) L L L + θ θ. (8) The descrptve statstcs n Table 2 suggest that 65.2% of employment n Fnnsh busness use at least the basc characterstc processng and storage capabltes at work. The pre- 3 Note that the comparson pont of wreless s wrelne connectvty, and thus the negatve sgn on wreless ndcates that t boosts productvty less than wrelne. 8

12 ferred estmate of βlθ for a plan desktop s 8.1% n Table 3 and the correspondng estmate for θ s 9.1% n Table 4. Smply pluggng these numbers nto (8) yelds an output elastcty estmate of 5.0%. Ths s the mean output elastcty estmate across nternatonally avalable comparable studes consdered by Stroh (2002). Obvously ths only refers to the key characterstc not to usage at large unlke most studes consdered by Stroh, although ths s arguably what the estmates of these studes capture not least because the average sample year of In order to get an overall measure of the output elastcty n the current context, the above model s re-estmated wth a sngle measure of usage derved as a sum of the sx bundles above. The respectve estmates of βlθ and θ become 21.7% and 24.2%. Pluggng these n nto (8) yelds an overall output elastcty estmate of 12.2%, whch s stll well wthn Stroh s 90% confdence nterval. Assumng that frms are ratonal, they wll contnue to nvest n untl assocated margnal costs and returns are equal, at whch pont also the average costs and revenues should be smlar. The year 1999 survey by Statstcs Fnland ncludes nformaton on -related costs. Ther estmated mean rato to labor costs ranges from 11.2% to 14.2% accordng to Malranta and Rouvnen (2003). If the bundle coeffcents n Column (3) of Table 3 are weghted by ther means n Table 2, the estmated return on the overall usage relatve to labor nput becomes 16.8%; f one were to take nto account that especally portablty but also connectvty were less common n 1999, the estmated return would be towards the mddle of the above range. Whle the two data sets refer to dfferent ponts n tme and are not entrely comparable, also ths comparson suggests that the derved results are qute reasonable. As above results show, portablty n essence the fact that one can carry processng power and storage capacty and possbly other characterstcs around has a consderable effect on productvty. Whle employed measure of portablty s rather specfc, t undoubtedly proxes for the overall organzatonal aglty, whch s unobserved. Due to portablty many tasks are not ted to hours spent physcally at the offce, whch may possbly lead to an ncreasng amount of unmeasured and unpad work. On the other hand, t may well have the opposte effect, as n 9

13 the absence of punchng n tmecards remote work s harder to montor. In branwork the dstncton between work and lesure has always been blurry and dong at least some unmeasured and unpad work durng after hours s often unavodable regardless of avalable technology a laptop can nevertheless rase the productvty of those same hours by an order of magntude. Great deas often comes n unexpected bursts and they are easly lost unless a sutable recordng technology anythng from a paper napkn to a laptop computer s avalable. A moble worker can more easly make hs/her efforts to concde wth these erratc eruptons of ngenuty. As compared to portablty, returns on wreless connectvty seem less stellar, although nevertheless postve n the preferred specfcaton. It should be kept n mnd, however, that, as of year 2001, wreless connectvty was n ts nfancy; the mmedate effect of ntroducng a new technology s almost surely negatve, as resources are ted to purchasng, nstallng, and learnng the technology not to menton co-nventon, e.g., n the form of organzatonal changes and some current producton s forgone. It should also be noted that less than ten per cent of employment uses wreless connectvty at all and less than two per cent use t bundled wth portablty. The modest usage suggests that the technology was stll at an experment stage as of Furthermore, as noted n the ntroducton, mportant and more prevalent forms of wreless connectvty are not captured by the measures employed. As dscussed by e.g. Wooldrdge (2002), results derved as above are consstent n large samples wth a relatvely weak set of assumptons. It s nevertheless true that smultanety mght bas the results, although causalty runnng from productvty to nvestment seems to suggest that the frms were not proft maxmzers. In lack of panel data and external nstruments the ssue of smultanety cannot be studed further here. As ably ponted out by DNardo and Pschke (1997), selectvty mght also bas the results n these contexts: more sklled and productve workers mght also be the ones usng at work. Above ths problem s hammered down 4 There are obvous problems n measurng the productvty gans assocated wth a gven pece of technology at an early stage. Lkewse there are problems n measurng the productvty gans assocated wth a technology that s nearly completely dffused. 10

14 by ncludng detaled controls on varous aspects of labor qualty. Sample selectvty should not be an ssue of concern, although the data does not nclude self-employed or one-person frms. Especally these one-person-shows mght gan enormously from portablty and wreless connectvty. Upon dervng the above results several choces had to be made on the emprcal setup, the consequences of whch are studed n the Appendx. Consderng the complexty of the ssues at hand and the early development stage of especally wreless connectvty, the fndngs seem qute robust. CONCLUSION Informatonal moblty s truly transformng whte-collar work, whch has s essence remaned the same snce the dawn of the ndustral revoluton. Whle t s lkely that the stll relatvely lttle used characterstcs portablty and wreless connectvty have been employed n uses where ther net returns are the hghest, t s also clear that the current ways of organzng whte-collar work do not explot these characterstcs to the fullest. As vrtually all frms are already usng computers and the Internet for some of ther busness processes, smply exposng more frms to cannot be a source of productvty growth n developed economes. Portablty and wreless connectvty reman relatvely rare at work, yet the above fndngs suggest they have potentally large productvty effects n the shorter run and undoubtedly more so n the longer run, once work practces have fully adjusted to ther presence. Ths paper has studed the general magntude of productvty effects of varous characterstcs. Data permttng, t would also be nterestng to study how frms dffer n ther abltes to beneft from varous characterstcs. Prevous work by Malranta and Rouvnen (2004) has shown, that younger frms seem to be better able to make most of ther nvestments. Prelmnary experments wth ths data suggest, that especally mult-establshment frms beneft from portablty. 11

15 REFERENCES Altmer, L., Besuner, P., Hasselfeld, K. and Johannemann, T. (2002), Go wreless to Spark Productvty, Nursng Management 33, Chwelos, P. (2003), Approaches to Performance Measurement n Hedonc Analyss: Prce Indexes for Laptop Computers n the 1990 s, Economcs of Innovaton and New Technology 12, Dearden, L., Reed, H. and Van Reenen, J. (2000), Who Gans When Workers Tran? Tranng and Corporate Productvty n a Panel of Brtsh Industres, IFS Workng Paper 00/04. DNardo, J. and Pschke, J.-S. (1997), The Returns to Computer Use Revsted: Have Pencls Changed the Wage Structure Too?, Quarterly Journal of Economcs 112, Gartenberg, M. (5 Dec. 2003), W-F: Why and Why Not, Computerworld 31, 21. Green, H. (24 Nov. 2003), TaylorMade, Busness Week 3859, Greenan, N. and Maresse J. (2000), Computers and Productvty n France: Some Evdence, Economcs of Innovaton and New Technology 9, Ilmakunnas, P. and Malranta M. (2003), Technology, Labor Characterstcs and Wage- Productvty Gaps. ETLA Dscusson Paper 860. Lancaster, K. (1991), Modern Consumer Theory, Edward Elgar, Aldershot. Malranta, M. and Rouvnen P. (2003), Teto- ja vestntäteknologan tuottavuusvakutukset Suomen lke-elämässä. Kansantaloudellnen akakauskrja 99, Malranta, M. and Rouvnen P. (2004), and Busness Productvty: Fnnsh Mcro-Level Evdence, n OECD (ed.), and Economc Growth, Organsaton for Economc Co- Operaton and Development, Pars. Nasaw, D. (23 Oct. 2003), W-F Is Becomng Popular at the Offce, Wall Street Journal, B.4. Olaru, S. (2003), An Issue Dedcated to Wreless Networks and Moble Computng, Internatonal Journal of Foundatons of Computer Scence 14, Rao, B. and Parkh, M. (2003), Wreless Broadband Drvers and Ther Socal Implcatons, Technology n Socety 25, Stroh, K. (2002). Reassessng the Impact of IT n the Producton Functon: A Meta-Analyss, Federal Reserve Bank of New York. Whte, H. (1980), A Heteroscedastcty-Consstent Covarance Matrx Estmator and a Drect Test for Heteroscedastcty, Econometrca 48, Wooldrdge, J. (2002), Econometrc Analyss of Cross Secton and Panel Data, MIT Press, Cambrdge, MA. Zurta, G. and Nussbaum M. (2004), Computer Supported Collaboratve Learnng Usng Wrelessly Interconnected Handheld Computers, Computers & Educaton 42,

16 APPENDIX Varous specfcatons of the derved model are estmated n Table 5. Column (1) s derved usng an unweghted and homoskedastc OLS estmator wth data ncludng both very large frms Table 5. Estmaton results of the labor productvty equaton wth alternatve specfcatons. ln(value added/labor) regressed (1) (2) (3) (4) (5) (6) wth the followng optons on Smplest Reference Unweghted Homosked. W. large W. outlers Labor weghts: Yes Yes Yes Yes Heteroskedastcty consstent: Yes Yes Yes Yes Large frms dropped: Yes Yes Yes Yes Outlers dropped: Yes Yes Yes Yes : Desktop *** *** (0.043) (0.057) (0.036) (0.044) (0.058) (0.060) : Laptop *** *** *** *** ** *** (0.121) (0.145) (0.103) (0.152) (0.160) (0.183) : Desktop+LAN *** *** *** *** *** *** (0.045) (0.054) (0.039) (0.035) (0.053) (0.062) : Laptop+LAN *** *** *** *** *** (0.088) (0.110) (0.103) (0.062) (0.128) (0.194) : Desktop+WLAN ** ** *** (0.069) (0.092) (0.061) (0.053) (0.089) (0.097) : Laptop+WLAN * *** ** (0.158) (0.256) (0.170) (0.121) (0.264) (0.265) CD: ln(captal/labor) *** *** *** *** *** *** (0.007) (0.015) (0.009) (0.007) (0.016) (0.016) CD: ln(labor) * (0.009) (0.012) (0.008) (0.008) (0.012) (0.014) Frm: young *** *** (0.032) (0.060) (0.032) (0.032) (0.072) (0.089) Frm: old (0.024) (0.029) (0.019) (0.019) (0.033) (0.034) Frm: mult-est ** *** *** *** ** (0.026) (0.032) (0.021) (0.020) (0.034) (0.046) Edu: lo tech * * (0.086) (0.122) (0.070) (0.103) (0.143) (0.190) Edu: m tech *** * *** ** ** ** (0.096) (0.143) (0.085) (0.099) (0.159) (0.201) Edu: h tech ** ** * *** *** * (0.145) (0.251) (0.154) (0.157) (0.281) (0.336) Edu: lo non-tech (0.091) (0.142) (0.075) (0.116) (0.164) (0.171) Edu: m non-tech *** ** *** *** *** ** (0.086) (0.142) (0.078) (0.098) (0.159) (0.167) Edu: h non-tech *** *** *** *** *** *** (0.133) (0.224) (0.149) (0.156) (0.231) (0.282) Labor: young *** ** *** *** ** *** (0.067) (0.159) (0.063) (0.084) (0.153) (0.172) Labor: old ** (0.070) (0.117) (0.063) (0.087) (0.131) (0.146) Labor: female *** *** *** *** *** * (0.054) (0.079) (0.044) (0.054) (0.081) (0.102) Observatons 2,404 2,358 2,358 2,358 2,381 2,381 Adjusted R-squared H0: Homoskedastc *** (degr. of freedom) (2) Note: Also ncludng (not reported above n the nterest of space) a constant term as well as NACE rev. 1 two-dgt ndustry and NUTS level-three regonal dummes. Estmated wth OLS; standard errors n the parentheses. ***, **, *, and + respectvely ndcate sgnfcance at 1, 5, 10, and 15 per cent level. 1 2 A χ dstrbuted Lagrange multpler test for heteroskedastcty as proposed by Whte (1980) but due to large number of regressors calculated usng predcted values from the regresson as dscussed n Wooldrdge (2002, p. 127). 13

17 and outlers (the smplest specfcaton). Column (2) s a reproducton of Column (1) n Table 3,.e., weghted heteroskedastcty consstent results wth large frms and outlers elmnated (the reference specfcaton). In Columns (3) (6) each of the four optons s relaxed one at the tme. As can be seen, qualtatvely results reman the same regardless of the specfcaton. As mentoned above, only the workers usng desktops and laptops are avalable as fractons; LAN and WLAN are only observed as frm-level dummes. In the text t s assumed that once a network s ntroduced at the frm, every computer s connected. The consequences of ths assumpton are studed n Table 6 wth nformaton from earler surveys recordng LAN usage as a fracton. The number of observatons drops severely due to the lackng overlap of the samples. Column (1) n Table 6 s a re-estmaton of Column (1) n Table 3 wth the subsample for whch the lagged LAN fracton s avalable. In Column (2) the lagged LAN fracton s used to recode the LAN bundles (nformaton on WLAN gnored). In Column (3) the lagged LAN fracton s used but the WLAN bundles are derved as done n the text (WLAN s not observed as a fracton n any of the avalable surveys). The results do not seem to be partcularly senstve to the choces made upon codng the bundles. Table 6. Estmaton results of labor productvty equaton wth alternatve bundle specfcatons. ln(value added/labor) (1) (2) (3) regressed on Reference Lag LAN Lag LAN, WLAN dmy : Desktop (0.100) (0.152) (0.151) : Laptop (0.247) (0.343) (0.318) : Desktop+LAN *** *** *** (0.078) (0.075) (0.077) : Laptop+LAN *** *** *** (0.155) (0.180) (0.169) : Desktop+WLAN (0.125) (0.164) : Laptop+WLAN (0.318) (0.361) Also: CD: ln(captal/labor), CD: ln(labor), Frm: young, Frm: old, Frm: mult-est., Edu: lo tech, Edu: m tech, Edu: h tech, Edu: lo non-tech, Edu: m non-tech, Edu: h non-tech, Labor: young, Labor: old, Labor: female, a constant, as well as Ind. and reg. dummes. Observatons Adjusted R-squared Note: Weghted OLS wth Whte (1980) heteroskedastcty consstent standard errors n the parentheses. Non- coeffcents not reported due to space lmtatons. ***, **, *, and + respectvely ndcate sgnfcance at 1, 5, 10, and 15 per cent level. 14

18 E L I N K E I N O E L Ä M Ä N T U T K I M U S L A I T O S (ETLA) THE RESEARCH INSTITUTE OF THE FINNISH ECONOMY LÖNNROTINKATU 4 B, FIN HELSINKI Puh./Tel. (09) Telefax (09) Int Int KESKUSTELUAIHEITA - DISCUSSION PAPERS ISSN Julkasut ovat saatavssa elektronsessa muodossa nternet-osotteessa: No 886 PEKKA SULAMAA, Palvelutyönantajat ry:n suhdanne-ennusteet tlastollsta analyysä kyselyanestosta s. No 887 ARI HYYTINEN LOTTA VÄÄNÄNEN, Could Mr. and Mrs. Captal Market Imperfecton Please Step Forward? An Emprcal Analyss of Adverse Selecton and Moral Hazard n Captal Markets p. No 888 RAINE HERMANS, Fnance of Small Bo-pharmaceutcal Industry n Fnland Descrptve Analyss p. No 889 ARI HYYTINEN TUOMAS TAKALO, Investor Protecton and Busness Creaton p. No 890 ARI HYYTINEN PEKKA ILMAKUNNAS, Entrepreneural Aspratons, p. No 891 JYRKI ALI-YRKKÖ ARI HYYTINEN MIKA PAJARINEN, Does Patentng Increase the Probablty of Beng Acqured? Evdence from Cross-Border and Domestc Acqustons p. No 892 NIKU MÄÄTTÄNEN MARKKU STENBORG TARMO VALKONEN, Vapaaehtonen eläkevakuutus ja kansantalous s. No 893 ARI HYYTINEN TUOMAS TAKALO, Multhomng n the Market for Payment Meda: Evdence from Young Fnnsh Consumers p. No 894 RAINE HERMANS, Projected Growth Effects of the Botechnology Industry the fourth pllar of the fnnsh economy?. Frst verson , Revsed verson p. No 895 PEKKA TSUPARI PETRI ROUVINEN, Inter-Frm Collaboraton and Electronc Busness: Effects on Proftablty n Fnland p. No 896 TOMI HUSSI, Intellectual Captal and Mantenance of Work Ablty The Wellbeng Perspectve. (Revsed verson ). 30 p. No 897 TOMI HUSSI, The Manageral Relevance of Mantenance of Work Ablty A Fnnsh Perspectve p. No 898 HANNU PIEKKOLA, Actve Ageng Polces n Fnland p. No 899 Demographc Aspects of Ageng and Tme Use n a Set of European Countres p. No 900 ANTTI-JUSSI TAHVANAINEN, Academc Spn-offs n Fnnsh Botechnology A Portrat of Frm Characterstcs p.

19 No 901 PETRI ROUVINEN, Dffuson of Dgtal Telephony Are Developng Countres Dfferent? p. No 902 MAARIT LINDSTRÖM OLLI MARTIKAINEN HANNU HERNESNIEMI, Tetontensvsten palvelujen rool metsäklusterssa s. No 903 MARKKU STENBORG, Are There Bases n the Market Defnton Procedure? p. No 904 PEKKA YLÄ-ANTTILA JYRKI ALI-YRKKÖ MARTTI NYBERG, Foregn ownershp n Fnland boostng frm performance and changng corporate governance p. No 905 ESSI EEROLA NIKU MÄÄTTÄNEN, On the Poltcal Economy of Housng s Tax Status p. No 906 RITA ASPLUND, A Macroeconomc Perspectve on Educaton and Inequalty p. No 907 RITA ASPLUND, The Provson and Effects of Company Tranng. A bref revew of the lterature p. No 908 KEREM TOMAK, Behavoral Economcs of Moblty p. No 909 LIISA LEIJOLA, The Educaton System n Fnland Development and Equalty p. No 910 ANNI HEIKKILÄ HANNU PIEKKOLA, Explanng the Desre for Local Barganng: Evdence from a Fnnsh Survey of Employers and Employees p. No 911 HANNU PIEKKOLA, Actve Ageng and the European Labour Market: Synthess Report p. No 912 DMITRY EFREMOV DANIL SMIRNYAGIN OLGA VALERIANOVA HANNU HER- NESNIEMI, Esco Companes n Northwest Russa, Legal Issues and Organzatonal Schemes p. No 913 PETRI ROUVINEN PEKKA YLÄ-ANTTILA, Palveluyrttäjyys kasvaa nopemmn ja työllstää enten s. No 914 KARI E.O. ALHO, The Fnnsh EMU Buffers and The Labour Market under Asymmetrc Shocks p. No 915 HANNU HERNESNIEMI, Funktonaalsten elntarvkkeden kluster, Esselvtys s. No 916 CHRISTOPHER PALMBERG MIKA PAJARINEN, The Global Dsperson of Innovatve Actvtes The Case of Fnnsh Multnatonals p. No 919 MIKA MALIRANTA PETRI ROUVINEN, Informatonal Moblty and Productvty Fnnsh Evdence. (Tedollsen lkkuvuuden tuottavuusvakutukset Suomessa) p. Elnkenoelämän Tutkmuslatoksen julkasemat "Keskusteluaheet" ovat raportteja alustavsta tutkmustulokssta ja välraportteja tekellä olevsta tutkmukssta. Tässä sarjassa julkastuja monsteta on mahdollsta ostaa Talousteto Oy:stä kopont- ja tomtuskuluja vastaavaan hntaan. Papers n ths seres are reports on prelmnary research results and on studes n progress. They are sold by Talousteto Oy for a nomnal fee coverng copyng and postage costs.