econsto Make You Publication Visible A Sevice of Witschaft Cente zbweibniz-infomationszentum Economics Blien, Uwe; udewig, Olive Confeence Pape Technological pogess and egional dispaities in (unemployment 54th Congess of the Euopean Regional Science Association: "Regional development & globalisation: Best pactices", 26-29 August 2014, St. Petesbug, Russia Povided in Coopeation with: Euopean Regional Science Association (ERSA Suggested Citation: Blien, Uwe; udewig, Olive (2014 : Technological pogess and egional dispaities in (unemployment, 54th Congess of the Euopean Regional Science Association: "Regional development & globalisation: Best pactices", 26-29 August 2014, St. Petesbug, Russia This Vesion is available at: http://hdl.handle.net/10419/124259 Standad-Nutzungsbedingungen: Die Dokumente auf EconSto düfen zu eigenen wissenschaftlichen Zwecken und zum Pivatgebauch gespeichet und kopiet weden. Sie düfen die Dokumente nicht fü öffentliche ode kommezielle Zwecke vevielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, veteiben ode andeweitig nutzen. Sofen die Vefasse die Dokumente unte Open-Content-izenzen (insbesondee CC-izenzen zu Vefügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in de dot genannten izenz gewähten Nutzungsechte. Tems of use: Documents in EconSto may be saved and copied fo you pesonal and scholaly puposes. You ae not to copy documents fo public o commecial puposes, to exhibit the documents publicly, to make them publicly available on the intenet, o to distibute o othewise use the documents in public. If the documents have been made available unde an Open Content icence (especially Ceative Commons icences, you may execise futhe usage ights as specified in the indicated licence. www.econsto.eu
Uwe Blien, Olive udewig: Technological pogess and egional dispaities in (unemployment Uwe Blien T.: 0911/1793035 uwe.blien@iab.de Olive udewig T.: 0911/1793205 olive.ludewig@iab.de Institute fo Employment Reseach - Institut fü Abeitsmakt- und Beufsfoschung (IAB Regensbuge Staße 104 D- 90478 Nünbeg (Gemany Daft of a pape, Nuenbeg, 30 Jan 2014 Abstact One of the key issues in economics is the explanation of unemployment and its vaiation acoss diffeent economies. Doing so, moden mainsteam macoeconomics efes to the effects of financial cises and to institutional stuctues and thei vaiation acoss counties. Howeve, unemployment within the Euopean states vaies nealy as much as between these counties. In the inteio of a county, howeve, thee ae only mino diffeences in institutions. Theefoe, the lage vaiation in egional unemployment and in the development of employment is puzzling. Ou explanation of this egional vaiation of unemployment builds on the egional industy composition and technological pogess. It is shown fomally that unde vey geneal and standad peconditions the elasticity of demand on poduct makets is decisive: Technological pogess leads to an expansion of employment if poduct demand is elastic. It is accompanied, howeve, by shinkage of employment if poduct demand is inelastic. A tansition fom the elastic into the inelastic ange of the demand function fo the most impotant poduct(s can aleady suffice to plunge a egion into cisis. In ou empiical analysis we use industy level time seies data on output, pices, employment and national income fo Gemany povided by the Fedeal Statistical Office and the OECD. We estimate Mashallian type demand functions using an instumental vaiables estimato to deive the pice elasticities fo diffeent industies and link this infomation to the egional labou maket pefomance of the espective industies and egions. Keywods: Stuctual change; Poductivity gowth; abou maket dynamics Q33, R11, J23
2 Contents 1. Intoduction... 3 2. Backgound... 7 3. The labou maket model of stuctual change... 11 4. Empiical Analysis... 15 4.1. Empiical stategy... 15 4.1.1. Identifying elasticities... 15 4.1.2. Elasticities and employment... 16 4.2. Data... 16 4.3. Estimating elasticities... 18 4.4. Elasticities and labou maket pefomance... 21 5. Conclusions... 25 Refeences:... 26
3 1. Intoduction One of the key poblems in economics is explaining the level of unemployment. In this pape we contibute a new explanation by looking at the inteaction of technological pogess and of demand conditions on poduct makets. Unde the condition of technological pogess less labou is needed to poduce the same amount of poducts. This points to the possibility of technological unemployment (ayad, Nickell 1985. Howeve, since technological pogess may also lead to a decease in pice, thee is a counteacting foce which depends on the demand conditions of poduct makets. In the following we exploe the stengths of the two foces theoetically and on the basis of vey eliable data also empiically. The explanation poposed can be contasted with those of standad appoaches of economics nomally used to explain unemployment. Some of these concentate on the effects of the last cisis in financial makets and show the tansmission pocesses to the labou maket (see Blanchad 2008, Kugman 2012 fo a popula vesion. Anothe explanation povided by moden mainsteam macoeconomics efes to the natue of institutional stuctues and thei vaiation acoss counties. Fom theoetical models it is deived that counties with moe flexible labou makets have elatively low unemployment ates. One pominent mainsteam explanation of unemployment is the so-called Euopean abou Maket Model of ayad, Nickell & Jackman (2005, cf. Calin & Soskice 2006 fo an integation with taditional macoeconomics. Thee, unemployment esults fom the competing claims of goups of economic subects. The claims of wokes and fim ownes on the social poduct ae kept in balance by unemployment. In ode to incease employment, economic policy theefoe has to ceate institutions which estain these demands. In a eview witten on the occasion of the new edition of the book by ayad et al., Blanchad (2007 emphasises that the theoy contained in the book has been empiically confimed (also ayad, Nickell & Jackman 2005, intoduction. Nonetheless, since the end of the 1990s thee has been inceasing citicism on the mainsteam appoaches. Fo example, some authos state that the empiical basis has been poven to be ambivalent (e.g. Howell et al. 2007. One mao poblem of the Euopean abou Maket Model and othe macoeconomic appoaches (see the abou Maket Matching Model by Motensen and Pissaides 2011 is thei inability to explain the vaiation of unemployment and of the development of employment within counties. Afte all, unemployment within a nation shows about the same level of vaiation as it does between counties (Südekum 2005. In Euope only vey few counties which wee hit especially had by the Financial Cisis stick out in ecent yeas (Geece and Spain and ae an exception of this ule.
4 Map 1: Development of employment in Gemany 2001-2012 Kiel Schleswig-Holstein Hambug Rostock Mecklenbug-Vopommen Rheinland-Pfalz Bemen Niedesachsen Bandenbug Sachsen-Anhalt Nodhein-Westfalen Hessen Thüingen Sachsen Fankfut Hannove Belin Halle Düsseldof Desden Saaland Saabücken Stuttgat Baden-Wüttembeg Nünbeg Bayen München von -37,09 bis -5,00 (64 von -5,00 bis 0,00 (61 von 0,00 bis 5,00 (91 von 5,00 bis 10,00 (78 von 10,00 bis 15,00 (62 von 15,00 bis 55,05 (46
5 Map 2: Unemployment ate Mach 2013 Kiel Schleswig-Holstein Hambug Mecklenbug-Vopommen Bemen Niedesachsen Hannove Belin Bandenbug Nodhein-Westfalen Düsseldof Rheinland-Pfalz Saaland Saabücken Hessen Fankfut Sachsen-Anhalt Sachsen Thüingen Bambeg Nünbeg Bayen Halle Desden Baden-Wüttembeg München von 1,76 bis 4,48 (79 von 4,49 bis 6,05 (79 von 6,05 bis 7,67 (78 von 7,68 bis 10,4 (80 von 10,5 bis 18,2 (86
6 If, howeve, within counties only mino diffeences in the institutions and the elevant maco-economic factos can be found, the huge vaiation in egional unemployment constitutes a poblem fo economic mainsteam. This is fo example tue concening the diffeence between East and West Gemany with unemployment ates of 7.0 and 12.6 % (Mach 2013. It is also tue fo egional dispaities within these two pats of the county. In Westen Gemany thee ae aeas with huge diffeences in employment development (Map1: As a consequence, thee is (nealy full employment like in the egion aound Munich. As a contast thee ae aeas with pesistent labou maket cises like the Ruh aea (see Map 2. These egional dispaities can mainly be attibuted to diffeent development paths of employment. They ae neglected by most theoies of (unemployment. Ou own explanation of employment and unemployment development builds on the egional industy composition and technical pogess. We do not ely on any technological deteminism, but look at the economic conditions shaping the effects of technological change. Fom this view the most impotant condition of the effects of technical pogess concens the elasticity of demand on poduct makets. The pice elasticity of demand tansmits the effects of technical pogess (o poductivity inceases we use the tems as synonyms and in a athe boad sense on employment. To see this we distinguish between two effects of poductivity inceases. As the same poduct can be poduced using less labou, technical pogess fist leads to a dop in the demand fo labou fo a given quantity. This is the displacement effect of technical pogess. In addition, howeve, the eduction in costs as a esult of technical pogess also leads to a dop in pice. This in tun inceases demand fo the paticula poduct and theefoe also demand fo wokes who ae employed in poduction. Theefoe, a compensation effect occus. This effect is the stonge the moe pice elastic demand is, as can be seen fom a simple fomal model we pesent in this pape. If demand is elastic the compensation effect dominates, if is inelastic the displacement effect pevails. This elation between technological pogess, demand conditions and employment we call the basic theoem of technological change. Recently, thee has been a boom of eseach on the effects of technological pogess on employment often influenced by an impotant pape of Auto, evy and Munane (2003. Howeve, this eseach teats the employment level as given and concentates on its skill composition (see Auto 2013 fo an oveview. In the following we fist pesent the theoetical backgound on the effects of technological change. Then, in the next pat of the pape, we develop a simple theoetical model. Futhemoe, we pesent the design of the empiical eseach and the esults obtained. Finally we conclude.
7 2. Backgound With technological pogess many feas and hopes ae associated. The idea of a technological unemployment comes up because of feas that wokes ae substituted by machines. Bynolfsson and McAfee (2011 quote Keynes in this espect and descibe a Race Against The Machine (booktitle. They see an acceleation of innovation pocesses and an inceasing pobability that computes will win the ace befoe wokes. ike Auto et al. (2003 they efe to diffeent skill goups of wokes but look at the global effect on the labou foce too. They descibe the cuently high levels of unemployment as geneated by technological pogess: Wokes ae substituted by machines, which ae being innovated at acceleating speed. An obection against this view is that new technologies ae intoduced due to the calculations of the fims involved. Then, the effect of employment also depends on these calculations. If poduct demand inceases enough no unemployment occus. The absolute speed of innovations is not decisive. It can be obseved that fields with shot cycles of innovation, e. g. micoelectonics, ae not affected by employment poblems. To have an intuition how the balance between labou saving and compensation effect woks it is instuctive to look at a small but ingenuously constucted maco-model developed by Appelbaum and Schettkat (e. g. 1999, see also Mölle 2001. They show that the limiting value fo the labou maket effect is the diect demand elasticity of one. Employment inceases with poductivity gains if poduct demand is elastic and it falls if poduct demand is inelastic. The model can be used to discuss seveal impotant issues, though it has not the levels of mico-foundation and geneality that we intend to each late. It begins with a definition equation fo the poductivity of labou π in a fim in which the poduction quantity Q is elated to the level of employment N. Q N z W P (2 π Q f(p, y, with dq /dp 0, dq /dy 0 (3 The second equation is a pice-setting function based on mak-up calculation. The pice is P, z is the mak-up facto, which also includes expenditues fo capital and W is the wage ate. Finally, the thid equation is a demand function, which falls with the pice and ises with the national income y. These equations on levels can be tansfomed to expessions of gowth ates: Nˆ Qˆ ˆ (1 Pˆ ẑ Ŵ ˆ (2 (1
8 Qˆ ŷ Pˆ (3 If ẑ 0 fom (1 to (3 the following expession fo a fim s employment development is deived: Nˆ η ŷ (ε 1πˆ ε Ŵ (4 In (4 an expession is obtained that includes two elasticities, the income elasticity of demand and the pice elasticity. It is easy to see that the stated popeties concening employment, technological pogess and demand elasticity ae implied with (4. Howeve, the focus on single fims in the equation is fo many puposes not vey instuctive, since it is not clea, whethe the output and ob gains of innovating fims ae achieved at the expense of competitos, o whethe thee is a net effect on aggegate industy employment (Pianta 2006. The elasticity of demand is diffeent if measued at the level of single fims o at the level of an industy. Fo the individual fim that is neithe a monopolist no an oligopolist, the behaviou of othe fims appeas to be given. If the fim lowes its pice, demand fo its poducts may incease vey stongly because othe fims, which maintain thei pices, ae displaced. If all the fims lowe thei pice, howeve, the quantity sold may change only slightly. Theefoe, elasticities at the fim and at the industy level diffe and an inteesting conceptual multi-level poblem has to be taken into account. To focus on whole makets it is necessay to aggegate all fims of an industy in the basic equations. We assume at the moment that the elevant units ae egions, though the model constuction is the same fo national economies. By aggegating all fims of a paticula industy i in a egion it is assumed that these fims ae identical: Nˆ i η ŷ (ε 1πˆ ε Ŵ (5 i i i i i i The model descibes poductivity gains as Hicks-neutal technical pogess, which is defined in such a way that the input atio of the poduction factos emains constant. This assumption ensues that shifts in labou demand ae not stemming diectly fom the technologicalpogess itself in a tivial way but that they ae the consequence of the maket mechanism. Additionally, the assumption simplifies the model stuctue. The tem technological pogess is used hee and in the following in a wide sense, which includes any outwad movement of the poduction function. Fo example changes in the oganisational pactices of a fim, which incease poductivity, ae included in this definition of technological pogess. As a consequence of technological pogess, wokes ae displaced when poduct demand is inelastic (i. e. i < 1. When demand is elastic ( i > 1 on the othe hand, employment inceases. This can be seen diectly fom (5. Theefoe the basic theoem of the employment effects of inceases in poductivity can be deived fom this simple model. The wate line is the elasticity of unity ( i = 1. Thee is no eason fo optimism like it is nouished by many basic appoaches of economics. In moden aticles using the Dixit/Stiglitz (1977 model of monopolistic competition, fims always choose a pice level located in the elastic pat of the poduct demand cuve. Theefoe, no poblem with employment aises, even if the above mentioned theoem is tue. Howeve, the Dixit/Stiglitz model equies athe specialised conditions of the econ-
9 omy. If they ae not met, the model does not hold. Taking a boade view on the topic, it is impotant to note that in cases in which fims do not have a monopolistic position in the maket, elastic demand is not secued. We will show late empiically that this is in fact elevant in a subset of makets. In (5 income elasticity is also impotant. When it is high, the demand fo a poduct can incease even unde conditions of pices ising seculaly. Thus, within the model positive employment effects on industy level can stem fom both, high pice elasticity and high income elasticity. Futhemoe, the model can be used to examine the effect of wage inceases. Accoding to (5, in the ealistic ange of values fo the demand elasticity (i. e. fo i > 0, wage ises lead to deceasing employment. The effect is stonge the moe elastic demand is, as we also know fom the Mashall-Hicks-ules. These ules of labou demand also establish a link between employment and the pice elasticity of demand. Howeve, this link is elated to the wage effects on labou demand and not to technical pogess. Neisse (1942: 53 appaently was the fist aguing that the elasticity of aggegate demand plays an impotant ole fo the balance between displacement and compensating effects. He pesented no model, only an agument which is based on an analogy between poduct and labou makets: If the demand elasticity on the poduct maket is below one then the tunove of the poduct shinks and it is assumed that in paallel less wok is equied in this specific industy. Neisse s ideas about technological unemployment wee hadly noticed within economics. Appelbaum and Schettkat wee the fist who came up with a model based fomulation. It was, howeve, a simple maco-model without complete mico-foundation. Next, in thei seminal pape about The dynamics of local employment in Fance, Combes, Magnac and Robin (2004 developed a small model stating fom the behaviou of individual agents. In this model the compensating effect dominates if the demand elasticity is geate than unity. This model was the foundation of thei empiical analyses which became influential in eseach on local labou makets. At about the same time in an empiical pape by Cingano and Schivadi (2004 a vesion of the basic theoem was included. By quoting the esult of Combes et al. they deived the theoem en passant in a simple model stuctue with only one poduction facto. The authos wee inteested in the analysis of agglomeation effects on poductivity and employment. They agued that agglomeation foces might push these taget vaiables in opposite diections. In the case of inelastic poduct demand agglomeation effects might incease poductivity but decease employment. Of couse thee is the possibility that both effects coincide in the case of an industy with elastic demand. In an economy with a mixed industy stuctue the net employment effect depends on its composition with espect to industies. With egional economies, the net effect depends on the specialization of egions. Cingano and Schivadi pesented empiical evidence suppoting diffeing agglomeation effects on poductivity and employment. On the elation between technological pogess, demand elasticity and employment a genealized theoetical model is developed by Blien and Sanne (2014. They stat fom individual behaviou, use homothetic poduction functions and intoduce a lage numbe of
10 poducts which could be complements o substitutes. Only the quoted souces ae known about the basic theoem on technological pogess. In the suvey of Pianta (2006: 579 the impotance of the pice elasticity of demand is mentioned, albeit without knowing the wateshed between labou-saving and compensating effects, the elasticity of unity. Howeve, it may be that due to the influential pape by Combes et al. the ole of the demand elasticity is egisteed moe and moe in the economics pofession. It is mentioned by Patidge et al. (2013 with a hint to this pape. Howeve, the geneal discussion about the effects of technological pogess on employment is much olde. The labou displacement effect is obvious and was discussed in Ricado s (1817 opus magnum and of couse also in Max s witing. Seveal mechanisms concening a compensating effect have been discussed extensively. The most impotant one concens geneal equilibium pocesses, which ae teated unde the heading of Say s law o in othe teminology: The disequilibium in the labou maket geneated by labou displacement is counteacted by a wage eaction stong enough to estoe a balanced labou maket. Othe pice eactions point in the same diection. Neisse (1942 discusses seveal vaiants of a compensation effect which ae taken up late in the liteatue. One basic idea is that a eduction of demand measued in tems of money due to a pice decease will geneate demand fo othe poducts. This is a vaiant of Say s law. It is not necessay hee to discuss at length whethe this is sufficient to secue full employment. At least it can be stated that if thee is a shift of labou demand between industies and, if thee is an uneven distibution of industies, also between egions. If migation between sectos and egions takes time and effots, egional unemployment is geneated. We will addess this point again late. Howeve, the specialization of egions is an inteesting condition fo the explanation of unemployment. Fo Kugman (1991: 5 the most stiking chaacteistic of the geogaphy of economic activity is its concentation. ocalisation effects lead to the specialisation of egions in a few pefeed industies. Although this concentation of paticula industies in specific egions has declined somewhat in time (cf. fo the USA: Kugman 1991: 75ff.; fo westen Gemany: Mölle, Tassinopoulos 2000, its extent emains astonishingly. Regions typically specialise in a subset of industies and poducts. Thus, industy compositions of egions ae vaying substantially. Consequently some egions have a highe shae of high elasticity industies, wheeas in othe egions industies with low elasticities ae dominating. Depending on these diffeences in the industy composition egional employment will develop diffeently as the consequence of technical pogess. The highe the shae of industies with elastic demand, the bette is the egional labou maket pefomance. In the eseach on specialization often thee diffeent effects ae mentioned, which wee chistened the thee Mashallian foces, following the mao wok of Alfed Mashall (1920. These ae labou maket pooling, fowad-backwad linkages and knowledge spilloves. Thee is a vaiety of theoetical models and a numbe of empiical papes (Ke, Komines 2010, Dauth 2010 showing the effectiveness of these foces. Kugman (1991: 123ff. deives a theoetical model on labou maket pooling as a diving foce of egional concentation and specialisation. His fundamental assumption is that the business cycles fo diffeent fims do not develop entiely synchonously. Fims ae hit by idiosyn-
11 catic shocks. It is theefoe advantageous fo fims and wokes to fom a oint pool of labou. Fims will settle in places whee thee ae aleady fims fom the same industy in ode to be able to hie wokes when thei own demand is high and that of the othe fims is low. Such behaviou educes unemployment o, in the case of flexible wages, ensues a steady wage development. Kugman s model shows that the advantage associated with this can cay moe weight that the deteioation of the competitive position that subsequently esults fo a fim. Specialisations ae impotant at national level, too. The developed economies poduce clealy diffeent national poduct mixtue as put fowad in the vaieties of capitalism appoach (Hall, Soskice 2001; Paunescu, Schneide 2004. If, fo example, the poduction of the Geman economy is compaed with that of othe developed counties, a dispopotionately lage specialisation can be seen in the aea of manufactuing. In addition many highquality goods ae manufactued in elatively small seies. In ode to explain this specialisation of nations, usually the theoy of compaative advantages o the New Tade Theoy is used. Howeve, such a specialisation can also be explained by the Geman institutional stuctue geneating a specific poduction system. Steeck (1991, 1997 and Soge and Steeck (1988 descibed this system and named it divesified quality poduction. This poduction system equies among othe things paticulaly highly skilled wokes. In Gemany the institutional pe-equisite fo this is the so-called dual system of vocational taining. This system (with fims and schools as leaning places is geaed mainly towads occupations in the manufactuing industy. In geneal the Vaieties of Capitalism liteatue suggests that institutional settings acoss a boad set of subsystems like the educational system, the financial system and the labou maket ae geaing economies towads specific industy compositions. Fo example coodinated maket economies like Gemany ae expected to have thei stengths in sophisticated but not to innovative manufactuing industies. ibeal maket economies like the Anglo-Saxon counties ae expected to be stong in new sevices and innovative high tech manufactuing (Hall, Soskice 2001. 3. The labou maket model of stuctual change In the last section we gave an outline of the Appelbaum-Schettkat model of technological pogess. In ode to obtain statements about unemployment a iche model is now developed which explicitly contains the labou maket. The change in employment is modelled in the usual way as the development of labou demand. This is a main diffeence to the models developed by Combes et al. and by Blien and Sanne. It has the additional advantage that a wage eaction like the one descibed in the Euopean abou Maket Model o in the wage cuve eseach can also be included. Since technological pogess leads to gowth o shinkage of employment in the espective industy, the model is called the abou maket model of stuctual change.
12 3.1 Fixed wage We begin with a case in which we teat the wage as fixed. As aleady mentioned a vey boad view of technological pogess is addessed, since all positive influences on a geneal technology paamete A in a poduction function Q ae descibed as technological pogess. Q A 1 K poduction function, with 0 < β < 1, K fixed (6 Q Q(P poduct demand (7 ike Combes et al. we use a Cobb-Douglas type poduction function in (6. In addition we stat out fom the assumption of pice-setting with pefect competition. With the function fo poduct demand (7 we abstact at the moment fom national income, but we will include it late. The equations ae fomulated fo individual fims, but the subscipt is dopped hee. The cost function c (e.g. accoding to Vaian 1992: 54f. shows the minimalcost facto combinations at given facto pices. Fo this it is necessay to detemine in each case the quantity of a poduction facto that is necessay fo a cetain poduction level (: labou, K: capital, A: technology facto, c: costs, W: wages; : inteest. 1 c (, W, Q min( K W s. t.: Q AK (8 min K WA 1 1 1 1 1 K Q c K WA 1 1 1 Q 1 1 K 1 1 0 The demand function fo capital with a given poduction quantity and given facto pices (conditional demand function is then: W K(, W, Q (1 1 A 1 Q The coesponding demand function fo labou takes the following fom: (9 (1 (, W, Q W A 1 Q It then follows fo the cost function with (maximum-pofit demand quantities inseted: (10 c(, W, Q K(, W, Q W(, W, Q 1 1 1 c (1 W A Q (11 1 The pice is equal to the maginal costs (with (1 :
13 c( W,, Q ( P Q (1 W Q A Q ( W A Q 1 1 1 1 1 Q P W A 1 1 (12 We deive via (12 the change in labou demand esulting fom technological pogess: 1 W A Q( P( A labou demand (1 d K da A 1 W (1 1 P Q dq dp Equation (14 yields diectly the fundamental theoem on the employment effects of technological pogess. The employment esponse to poductivity inceases is positive if the elasticity of demand is geate than 1. Howeve, this is always fulfilled fo individual fims unde pefect competition (η >> 1. If the fims of an industy ae aggegated, howeve, the employment in an industy can be elated to the oveall demand fo this aggegate. Then equation (14 applies fo the entie industy. The aggegation is possible since the poduction function shows constant economies of scale. 3.2 Reaction of wages to unemployment In the following we stat out fom the (exteme simplification that the economy only poduces one single good. This assumption allows establishing a connection with the labou maket, because now the function fo labou demand depicts the oveall demand on a labou maket. The aim of the following analysis is to constuct a model that is simila to a cetain degee to that of ayad et al. Since the fomalization is standad, only some basic equations ae given. Hee we do not bothe with the micofoundations of the model. Fo easons of simplification, in the following employment is measued as a shae of the active population, which is in tun standadised to 1 (N = 1. Unemployment esults accodingly with U = 1 -. In the spiit of the wok by ayad, Nickell & Jackman (2005 and Calin, Soskice (2006 fo the national level and by Blanchflowe, Oswald (1994, 2005, see Baltagi, Blien, Wolf 2009 fo Gemany fo the egional level, it is assumed that the wage esponds invesely to egional o national unemployment (wage-setting cuve o wage cuve. In ode to make the calculations easie it is assumed that the wage cuve is not semi-logaithmic but linea. The following expession esults: (14 W ' U (15 1 ' 1 ' W (16 The ationale behind this fomalisation is quite analogous to that of ayad et al. The wage (setting cuve can be deived concening efficiency wage appoaches and wage negotia-
14 tion models. The fact that a linea and not a log-linea fomulation is adopted hee does not constitute a limitation. Empiical studies on the egional wage cuve do not clealy favou eithe of the two fomulations ove the othe (Blien 2001. In the following the wage is endogenised, using (10 and (16: Q A (1 ( 1 (17 Q A (1 ( 1 Implicit function: 0 (1 ( 1 Q A G (18 ( (1 (1 ( ( 1 (1 2 1 1 A P Q QdP PdQ A K G A G da d (19 Diffeence between (14 and (19: S A P Q ( (1 (1 ( ( ( 2 1 (20 with 0 < S < 1 if P Q < 0 Thus, the effect of inceases in poductivity is weake in the case of endogenous wages. Howeve, the tuning point of the development, i. e. the elasticity of one, emains the same. Thus the pevious finding, that employment on industy level depends on the pice elasticity of demand and that consequently the egional development of employment is depending on the industy composition is still holding. In the model of Combes et al. (2004 the labou maket is also included, but in a diffeent way. In thei case the supply elasticity of labou is egaded. If this elasticity is infinite, the effect of poductivity changes is like the one in the model without labou maket. If the supply elasticitiy of labou is smalle the poductivity effect is dampened as in ou model. Finally, we could also include the income level of the elevant maket aeas fo which the poducts ae addessed. This income level influences total demand of the espective poduct. Theefoe (17 could be witten with espect to the social poduct Y: ( (1 ( 1 Q Y A (21 The consequence of this extension is that the social poduct has the effect of an additional shift paamete in the equation fo labou demand. The social poduct influences poduct demand and theeby also labou demand.
15 4. Empiical Analysis Ou model links the pice elasticity and the income elasticity diectly to labou maket outcomes. In ode to establish this link empiically we have to deive in a fist step these elasticities fom industy level data. In a second step we use the elasticities to explain the pefomance of diffeent labou makets like egional o industial labou makets. 4.1. Empiical stategy 4.1.1. Identifying elasticities Despite the theoetical simplicity of the pice elasticities of demand its empiical identification faces some challenges (Mölle 2001. Fo example, estimating a classical Maschallian demand function fo a specific good would equie the inclusion of a vecto of the pices of all othe goods o at least of all othe industies. This is, howeve, hadly feasible because of the limited numbes of obsevation available. Following Mölle (2001 we assume that poducts of each industy ae substituted against a composite good, which is epesenting the poduct mixtue of all othe goods. Additionally we assume that the espective industies ae small compaed to the total economy yielding the following Mashallian type demand function: q pit pt i yt uit it oi 1 i 2 (22 whee q it is the industy eal output, y t is the eal national income, p it is the industy pice level and p t the national pice level. All vaiables ae in logaithms, thus p it -p t is giving the pice of industy i elative to the geneal pice level p t. Estimates fo β 1i povide the pice elasticities on industy level and those fo β 2i give the income elasticities (η. This specification implies also that domestic and foeign consumes ae identical and that the income elasticity concept is also applying to intemediate goods. As in ou theotetical model we define the pice elasticity ε positively and multiply β 1i with -1. Thus ε should be negative with inelastic demand between 0 and 1. Demand is pice elastic if 1 ε holds. Industies with η > 1 face income elastic demand. They ae poducing supeio goods; those with 0 η 1 ae selling elative infeio poducts and those with η < 0 offe absolute infeio ones. In ou fist step we estimate equation 22 and get estimates fo the pice and income elasiticities. These ae then enteed into ou second step which establishes the link between elasticities and labou maket development.
16 4.1.2. Elasticities and employment Ou model states that the employment esponse to poductivity inceases is positive (negative if demand is pice elastic (inelastic. acking appopiate poductivity measues we assume that thee ae poductivity inceases in each industy without quantifying them. Thus we expect that industies with 1 ε have bette labou maket pefomance than those with 1 > ε. The bigge the shae of industies with elastic demand in any administative unit (e.g. county, national state the bette will be the labou maket pefomance of the espective unit. Thus we have two diffeent units of analysis: industies and some geogaphical defined administative unit. We use Geman NUTS-III Regions (Keise as administative unit. This gives us thee diffeent levels of analysis, industies, industies within diffeent egions and egions. We define labou maket pefomance as the change in employment within a specific peiod of time. We begin with a desciptive analysis. That is, we compae the mean of elative employment change in pice elastic industies and in inelastic industies. Then we poceed with egession analysis. We egess the development of employment on the two elasticities, which wee deived in the fist step: i 0 1 i 2 2 i 3 X v (23a i i In ou second appoach we estimate the same egession function fo industy in each egion. (23b i, 0 1 i 2 i 4X i, 5X vi i, Finally we aggegate the elevant infomation on egional level. We calculate the aveage elasticities fo each egion by weighting the industy specific elasticities with the employment shae of the especitve industies in the specific egion. The sum of these weighted elasticities in a egion gives the employment weighted aveage egional pice and income elasticity. ˆ X (23c ˆ 0 1 2 5 Fo all appoaches we expect fom ou model that the moe pice elastic demand is, the bette the employment development will be. Thus we expect positive sings fo α 1 and α 2. 4.2. Data One main souce of data is the national account of Gemany fom the Geman Fedeal Statistical Office. 1 The national accounts povide infomation fo goss value added on industy level (two digit and the national GDP. The industy value added is given in nominal and eal tems which allow calculating industy specific pice indices. The fedeal statistical office is also poviding the national consume pice index, which we take as an appoximation of the national pice level. All these vaiables ae indexed with the base yea 2000 1 To be moe pecise it is the Fachseie 18 Reihe 1.4.
17 (index value = 100. The national accounts also include the wage bill fo each industy and the numbe of employed. We use this data to calculate the deflated wage pe capita. The data of the Fedeal Statistical Office is in pinciple available fo a athe long time peiod. Howeve, almost all economic data on Gemany is suffeing fom the stuctual beak caused by the unification. Fo this eason we skip data befoe 1994 esulting in a obsevation peiod anging fom 1994 to 2007. Anothe data set is taken fom the employment statistics of the Geman Fedeal Employment Agency. 2 It coves all employees who ae subect to the social insuance system and it povides a ich set of infomation on these employees. Fulltime equivalents ae calculated by weighting pat-time employed by 0.5. In contast to the infomation of the Statistical Office the data of the employment agency is available on egional level but has no infomation on industy pices and industy poduction. Thus the data of the Statistical Office is used to estimate ou fist step and equation 23a. The data fom the Employment agency is used to estimate equations 23b and 23c. The data fom the employment agency is only available fom 1996 onwads, because of some poblems with the industy classification. Thus, we ae estimating the elasticities fo a peiod a bit longe than we can calculate the employment change. Synchonizing both obsevation peiods by cutting of the fist yeas off the national account data would educe the degees of feedom of the elasticity estimation. This seems to us a pice too high to pay. Additionally using diffeent data souces and diffeent obsevation peiods can povide a fist obustness check. While ou industy cycle agument is based on changing elasticities this is holding fo long peiods and is less elevant fo the shot peiods we use fo estimation. Thus we ae not estimating time vaying elasticities but constant ones. We ae not captuing the industy life cycle by investigating the changing elasticities fo each industy acoss time but by analyzing elasticities acoss 50 industies at vaious stages of the life cycle. We calculate the pecentage change in employment fo the peiod fom 1994 to 2006 fo the industy level data fom the Statistical Office. This seves as dependent vaiable in estimating equation 23a. The same is done fo the peiod 1996 to 2007 fo the industy employment in each egion and the aggegated egional employment using the data of the Employment Agency. Ou model is based on maket mechanisms. Thus we exclude state diven industies (Agicultue; Fishing; Mining and quaying; Public administation and defence; Compulsoy social secuity; Activities of households as employes of domestic staff fom the analysis. 2 Beschäftigungsstatistik de Bundesagentu fü Abeit, Febuay 2009
18 4.3. Estimating elasticities We estimate the elasticities using equation 22. q it is appoximated by the eal goss added value on industy level, the industy pice level (p it is deived by dividing the eal goss added value by the nominal goss added value. p t is appoximated by the consume pice index and y t by the eal GDP. Remembe, all values ae indexed with the base yea 2000 (100 and logaithms ae taken. We estimate fou diffeent specifications. The fist vaiation is, that we substitute p t fo p it - p t. Thus we ae not solely looking at the elative pices but also at the absolute pice levels in each industy. The two esulting specifications ae then estimated using OS and an instumental vaiable estimato. We suspect that the pices might suffe fom endogeneity. To account fo this poblem we instument p t and p it -p t with the lagged values of q it, p it and p t. While we pefe the instumental vaiable estimato of the oiginal equation, we give also the esults of the othe thee specifications in Table 1 in ode to check the stability of esults. In ou view the table indicates a high stability of esults (signs and magnitude if the small numbe of obsevations is taken into account. Additionally, we define a dummy that is one if the industy poduction level is pice elastic. That is, if β 1 is not significantly diffeent fom 1 o significantly geate than 1. These dummies and the estimation esults ae then used in ou second step to test whethe pice elastic industies have a bette labou maket pefomance o not.
Table 1: Estimated pice elasticities of the fou specifications p it p it -p t p it p it -p t Elasticity p-value Elasticity p-value Elasticity p-value Elasticity p-value Manufactue of food poducts and beveages 0.711*** 0.000 0.685*** 0.001 0.456** 0.005 0.313 0.236 Manufactue of tobacco poducts 1.403** 0.003 1.284* 0.011 1.307** 0.002 1.110* 0.022 Manufactue of textiles 1.053 0.085 0.304 0.541 0.070 0.922-0.543 0.253 Manufactue of weaing appael; dessing and dyeing of fu 0.692 0.273 0.082 0.850-0.729 0.479-0.573 0.254 Tanning and dessing of leathe; manufactue of luggage, handbags, saddley, haness and footwea 0.964 0.167 0.229 0.792 1.874* 0.023 0.624 0.629 Manufactue of wood and of poducts of wood and cok, except funitue; manufactue of aticles of staw and plaiting mateials -0.530 0.342-0.440 0.225-1.001 0.190-0.527 0.146 Manufactue of pulp, pape and pape poducts 0.807*** 0.000 0.711*** 0.000 0.645*** 0.000 0.541*** 0.000 Publishing, pinting and epoduction of ecoded media 0.293 0.661-0.345 0.543 0.757 0.284-0.334 0.588 Manufactue of coke, efined petoleum poducts and nuclea fuel 1.014*** 0.000 1.008*** 0.000-3.076 0.869-4.454 0.870 Manufactue of chemicals and chemical poducts 1.440* 0.049 1.378** 0.001 3.503* 0.010 1.674*** 0.000 Manufactue of ubbe and plastic poducts 1.119** 0.002 1.002*** 0.000 1.932** 0.001 1.210*** 0.000 Manufactue of othe non-metallic mineal poducts -0.101 0.817-0.175 0.568-0.542 0.283-0.339 0.250 Manufactue of basic metals 0.363*** 0.000 0.397*** 0.000 0.341*** 0.000 0.382*** 0.000 Manufactue of fabicated metal poducts, except machiney and equipment 1.769** 0.004 1.012 0.063 2.601* 0.025 0.529 0.497 Manufactue of machiney and equipments n.e.c. 0.718 0.142 1.903*** 0.000 0.697 0.111 2.529*** 0.000 Manufactue of office machiney and computes 0.932*** 0.000 0.906*** 0.000 0.955*** 0.000 0.920*** 0.000 Manufactue of electical machiney and appaatus n.e.c. 4.067** 0.004 1.799 0.080 1.666 0.235 0.449 0.600 Manufactue of adio, television and communication equipment and appaatus 1.100*** 0.000 1.060*** 0.000 0.986*** 0.000 0.946*** 0.000 Manufactue of medical, pecision and optical instuments, watches and clocks 0.972 0.343 3.010** 0.005-0.168 0.837 1.942* 0.028 Manufactue of moto vehicles, tailes and semi-tailes 0.432 0.567 1.135 0.071 0.675 0.428 1.669* 0.012 Manufactue of othe tanspot equipment 3.755* 0.011 3.038* 0.039 0.395 0.914 0.889 0.686 Manufactue of funitue; manufactuing n.e.c. 2.533*** 0.000 2.495** 0.007 2.653*** 0.000 2.787*** 0.001 Recycling 0.949*** 0.000 0.988*** 0.000 0.994*** 0.000 1.048*** 0.000 Electicity, gas, steam and hot wate supply -0.138 0.564-0.091 0.732-0.383 0.113-0.377 0.176 Collection, puification and distibution of wate 0.002 0.995-0.092 0.825-0.759* 0.049-0.966** 0.008 Constuction 0.546 0.290-0.321 0.634 1.163 0.073-0.252 0.786 Sale, maintenance and epai of moto vehicles and motocycles; etail sale of automotive fuel 1.018* 0.022 1.069** 0.001 1.394** 0.003 1.319*** 0.000 Wholesale tade and commission tade, except of moto vehicles and motocycles 0.002 0.993-0.020 0.916 0.116 0.608 0.011 0.951 OS IV
20 Retail tade, except of moto vehicles and motocycles; epai of pesonal and household goods -0.051 0.861-0.373 0.143-0.096 0.865-1.067* 0.010 Hotels and estauants 0.706** 0.007 0.524 0.162 0.699*** 0.001 0.573 0.152 and tanspot; tanspot via pipelines 0.775*** 0.000 0.726** 0.002 0.753** 0.005 0.405 0.245 Wate tanspot 0.692* 0.031 0.690* 0.020 1.241** 0.003 1.075*** 0.001 Ai tanspot 0.104 0.848-0.003 0.995-0.792 0.299-0.690 0.259 Suppoting and auxiliay tanspot activities of tavel agencies -0.049 0.925 0.716 0.190-0.883* 0.045 0.384 0.557 Post and telecommunications 0.213 0.677 0.700 0.213-0.042 0.956 1.431 0.136 Financial intemediation, except insuance and pension funding 0.190 0.126 0.202 0.152 0.188 0.120 0.224 0.124 Insuance and pension funding, except compulsoy social secuity 0.595*** 0.000 0.615*** 0.000 0.528*** 0.000 0.549*** 0.000 Activities auxiliay to financial intemediation 0.575*** 0.000 0.612*** 0.000 0.506*** 0.000 0.561*** 0.000 Real estate activities 0.867 0.106 1.261** 0.002 0.429 0.418 1.117*** 0.000 Renting of machiney and equipment without opeato and of pesonal and household goods 2.380* 0.013 1.423 0.089 2.041 0.103 0.783 0.368 Compute and elated activities 0.786 0.499-0.110 0.881 1.269 0.347-0.210 0.771 Reseach and development 1.768 0.057 2.689 0.225 1.931* 0.013 6.437* 0.036 Othe business activities 0.470* 0.011 0.800*** 0.001 0.347* 0.021 0.731*** 0.000 Public administation and defence; complusoy social secuity 0.220 0.235-0.239 0.290 0.349** 0.009-0.312 0.270 Education -0.086 0.721-0.319 0.306-0.174 0.395-0.869* 0.047 Health and social wok 0.093 0.906 1.782* 0.025-1.037 0.410 3.347*** 0.001 Sewage and efuse disposal, sanitation and simila activities 0.843* 0.019 1.060** 0.002 0.809* 0.014 1.139*** 0.000 Activities of membeship oganizations n.e.c. 0.513 0.355-1.007 0.310 0.591 0.202-1.781 0.111 Receational, cultual and spoting activities 0.795 0.151 0.005 0.994 1.313 0.231-2.564 0.090 Othe sevice activities -0.235 0.378 0.463 0.273-0.376 0.104 0.027 0.953
21 4.4. Elasticities and labou maket pefomance Fist evidence on the elationship between the pice elasticity and labou maket pefomance is given by the compaison of the mean employment change of elastic industies and nonelastic industies. The esults ae shown in Table 2. Pice elastic and inelastic industies ae defined as descibed above. Fo ou pefeed elasticity estimation (IV-estimato and using elative pices we find that the eduction in employment was substantially smalle (by about 4.2 pecentage points fo the elastic industies than fo inelastic industies. This finding is holding fo two othe estimations as well. Only the elasticities deived fom the OS estimato using the pice level is indicating that the inelastic industies ae bette pefoming. Fo the esults in the lowe pat of Table 2 we attempt to adust fo the quality of the point estimatos of the pice elasticity. Fo this pupose we deive the width of the 95% confidence inteval and use its invese as weights fo the industies. The esults stay basically the same. Table 2: Mean employment change fo elastic and inelastic industies Estimato Pice measue Pice elastic Pice inelastic Diffeence (inelastic- elasltic OS evel -8.73-4.57 4.16 OS Relative -4.24-6.44-2.20 IV evel -3.37-6.54-3.17 IV Relative -2.70-6.91-4.21 Adusted fo the quality of the elasticity estimate OS evel -7.72-4.26 3.46 OS Relative -4.68-4.94-0.26 IV evel -3.58-5.11-1.53 IV Relative -.79-6.42-5.63 Elaboating ou industy level analysis a bit futhe we use again the industy level infomation on employment change. But now we ae putting it in a egession famewok. We estimate equation 23a using thee diffeent estimatos. The fist one is plain OS. The second estimato is an outlie obust egession, which is weighting the diffeent industies in an iteative pocess with the invese of the esidual. As final vaiation we weight the industies in the OS estimato by the width of the 95% confidence inteval of the point estimates of the pice elasticity. In ode to contol some basic intevening factos we put in the estimation function the income elasticity and the change in wages. The esults ae mixed (Table 3 but still pointing in the expected diection. The coefficient of the pice elasticity has a positive sign, which is expected because. Howeve, the effect of the
22 pice elasticity is only significant in the outlie obust estimation with the elasticities deived with OS, which is not ou pefeed model. Adusting fo the quality of the elasticity estimates (not shown in the table is not changing the esults. The weak significance level might be due to the small numbe of obsevations. Howeve, taking this small numbe of obsevations into account, the consistently positive sign of the coefficient of the pice elasticity and one significant esult ae eassuing. The coefficient fo the income elasticity has the expected sign and it is in most cases significant. Table 3: Effects of pice elasticity on employment change on industy level Second stage: OS Elasticity estimates with OS Second stage: Outlie obust estimato Pice elasticity 2.51 4,85** (2.68 (2.28 Income elasticity 6.27*** 4.14*** Adusted R 2 / F- Tes (1.20 (1.02 0.35 F(3, 46 = 7.49 *** Second stage: OS Elasticity estimates with IV Second stage: Outlie obust estimato Pice elasticity 0.76 2.25 (2.09 (1.50 Income elasticity 0.53 4.37*** Adusted R 2 / F- Test (0.50 (1.11-0.02 F(3, 45 = 6.22*** Contol vaiable: wage change 1996-2007 Numbe of obsevations: 50 standad eos in paenthesis * p<.1; ** p<.05; *** p<.01 We can ovecome the poblems associates with the small numbe of obsevations by looking at industies in egions fo all egions. This is inceasing the numbe of obsevations by moe than the facto 300. Howeve, looking at egions is not piceless. It is giving ise to at least two poblems. Fistly, some industies in some egions ae athe small. In such a case employment change might be pimaily due to some happy o unhappy coincidence (o decision in one o two establishments. Thus we add a substantial souce of noise to ou data. To contol this we eestimate ou specification afte successively excluding industies with less than 25, 50, 100, 250, 500, 1000, 2500 and 5000 employees. Secondly, thee might be some egion specific effects on the employment development. To contol fo these we inset some egional contol vaiables. The wage vaiable becomes now the median wage of an industy in a egion. Additionally we add the intequantiles ange of industy wages in a egion. We put dummies fo the siedlungstuktuelle Keistypen, a widely used classification of Geman disticts (Goema and Imen 1991 povided by the
23 Fedeal Office fo Building and Regional Planning (BBSR, in ou estimations. Finally we contol fo the effects of egional density using the inhabitants pe squae kilomete and the distance to the next motoway (accessibility. We again estimate thee diffeent specifications: OS, an outlie obust egession and OS with industies weighted by the invese of the width of the 95% confidence inteval of the point estimates of the pice elasticity. The esults in Table 4 ae suppoting the findings of ou model. The coefficients of the pice elasticity as well as fo the income elasticity ae positive acoss all estimates. All coefficients ae significant. These esults suggest that the employment gowth ove the obsevation peiod in an industy with the ε=1 is two to ten pecentage points highe, than the employment gowth in an industy with ε=0. Howeve, the explanatoy powe of some specifications, expessed by the adusted R 2, is athe weak fo some estimates. This seems to be due to the white noise mentioned above, because the R 2 inceases with the theshold. The R 2 becomes satisfactoy at the theshold of 1000. Finally we aggegate employment change, elasticity and wages at the egional level. Accounting fo the white noise poblem fom above we have thee vesions of this aggegation, which ae again excluding industies below the 1000, 2500 and 5000 thesholds. The esults ae given in table Table 5. Fo the estimates with the aggegate above the 1000-theschold the effect of the pice elasticity is insignificant. The income elasticity has the ight sign and is significant. When inceasing the theshold the coefficients of the pice elasticity become negative and significant.