Substitution and Crowding-Out Effects of Active Labour Market Policy

Transcription

1 FRIEDRICH-ALEXANDER-UNIVERSITÄT ERLANGEN- NÜRNBERG Lehrstuhl für VWL, nsbes. Arbetsmarkt- und Regonalpoltk Professor Dr. Claus Schnabel Dskussonspapere Dscusson Papers NO. Substtuton and Crowdng-Out Effects of Actve Labour Market Polcy ELKE J. JAHN AND THOMAS WAGNER JUNE 000 ISSN: Edtor: Prof. Dr. Claus Schnabel, Fredrch-Alexander-Unverstät Erlangen-Nürnberg Elke J. Jahn and Thomas Wagner

2

3 Substtuton and Crowdng-Out Effects of Actve Labour Market Polcy * Elke J. Jahn a and Thomas Wagner b ABSTRACT: The search model contans two matchng technologes, the publc employment servce (PES) wth ts type-specfc regsters for workers and vacances, and the search market where frms advertse vacances and unemployed who have not been placed by the PES search for jobs. The placement actvty of the PES ncreases the barganed wages, reduces actve job search, decreases the number of advertsed vacances, but - compared wth the lassez-fare regme - ncreases employment and per capta consumpton. Of all the nstruments of ALMP, the probabltes of a match, the porton of unsklled not nterested n a job, and the hrng subsdes generate crowdng-out effects. The productvty of the unsklled, (re-employment)bonuses, penaltes for volatons of the search rule, and the strngency of the search rule cause crowdngn effects. Assstance for "problem groups" s less effectve than promotng actve job search. ZUSAMMENFASSUNG: Das Suchmodel umfasst zwe Matchng-Technologen, de des PES mt typspezfschen Regstern für Arbetslose und Vakanzen und de des Suchmarkts, wo Frmen Vakanzen annonceren und ncht Vermttelte nach Stellen suchen. De Vermttlungstätgket des PES erhöht de Lohnkosten, reduzert de aktve Suche und de annoncerten Vakanzen, senkt m Verglech zum Lassez-fare Regme de Arbetslosenquote und stegert den Pro-Kopf-Konsum. Von den Instrumentvarablen der ALMP verdrängen de Matchwahrschenlchketen, der Antel der Arbetsunwllgen und de Lohnkostenzuschüsse Arbetsplätze. De Produktvtät der Gerngqualfzerten, (Wederbeschäftgungs-) Prämen, Sanktonen für Suchregel-Verstöße sowe de Intenstät der Regeldurchsetzung senken de Arbetslosgket. De Förderung von Problemgruppen st wenger wrkungsvoll als de Förderung der aktven Jobsuche. KEY-WORDS: Matchng model, actve labor market polcy, PES, search market, heterogeneous unemployment pool, rankng JEL-CODE: J 4, 63, 64, 68 * Ths draft benefts from comments receved at the conference of the EALE 999, Regensburg, the conference of the Veren für Socalpoltk 999, Manz and at a semnar of the Insttute for Employment Research (IAB-Colloquum) 000, Nürnberg. a Fredrch-Alexander-Unverstät Erlangen-Nürnberg, Lehrstuhl für Arbetsmarkt- und Regonalpoltk, Lange Gasse 0, Nürnberg, Germany, Elke.Jahn@wso.un-erlangen.de. b Unversty of Appled Scences, Nürnberg, Hastverstr. 3, Nürnberg, Germany, Thomas.Wagner@fh-nuernberg.de.

4 . INTRODUCTION Actve labour market polcy (ALMP) s desgned to enhance the productvty and qualfcatons of unemployed job seekers, to mprove ther search and self-presentaton technques, and gven a lmted demand for labour to allocate the avalable jobs of a country more evenly. In Europe and Scandnava, ALMP consumes a szeable amount of resources - wthn the OECD on average 0.9% of GDP n 996 (Martn 998). Effcency gans of ALMP are hard to measure. Many authors estmate that they are modest or even negatve f job-creaton and saved nsurance and transfer payments are balanced aganst the hgh expenses of the publc employment servce (PES) for tranng, wage and ncome subsdes, and especally for job-creaton programmes. Nevertheless, n the reform of socal securty systems n Europe, wage and ncome subsdes seem to have a brght future (Snower 997; Phelps 997; OECD 998). ALMP s consdered to be relatvely successful f the nstruments are talored to well-defned (problem) groups, or f they are desgned to ntensfy job search (OECD 993a, 996a). On the other hand, job-creaton schemes and subsdes are often crtcsed as havng consderable deadweght losses and crowdng-out effects. Deadweght losses exst f job seekers or supplers of vacances are subsdsed even though they would have ended up wth job contracts wthout help. Substtuton effects are present f subsdsed persons or vacances crowd out those that are not supported by the subsdes. In ths paper, we dscuss the substtuton effects and the crowdng-out effects of ALMP usng a smulaton model wth a fully ntegrated PES and a prvate search market. Asde from mcroeconometrc studes of ALMP (LaLonde 995) there are two approaches that analyse the nterdependence between ALMP and the aggregate unemployment rate. The frst conssts of macroeconometrc regressons (Layard et al. 99; OECD 993a; Calmfors 994; Calmfors and Skednger 995; Burda and Lubyova 995; Scarpetta 996; Bellmann and Jackman 996; Nckell 997; Büttner and Prey 998; Nckell and Layard 999; Blanchard and Wolfers 999). In those papers, the estmatng equaton s usually based on a labour market theory that goes back to Layard and Nckell (986) and Layard et al. (99) and has become a standard framework for labour market analyss. Calmfors (994) has modfed the standard model and, followng OECD (993a), has provded a classfcaton of ALMP effects that has been wdely used snce. Calmfors dentfes three effects that nfluence aggregate employment through the labour demand functon, the productvty effect,

5 3 the substtuton effect, and the deadweght loss whch s connected wth the actve polces. The competton effect and the ncentve effect work through the wage-settng functon. Fnally, effects that drectly nfluence the matchng process can express themselves ether through the demand or through the wage-settng functon. The second approach focuses on the smulaton of ALMP and uses a labour market model whch s based on the more recent theory of two-sded-search (Mllard and Mortensen 997; Mortensen 994). Ths branch of search theory employs a matchng functon to model the nformaton mperfectons and the heterogenetes of the labour market (Damond 98; Pssardes 986, 990; Hosos 990; Mortensen and Pssardes 994, 999). The results of the macroeconometrc regressons suffer from dentfcaton and endogenety problems (OECD 993a; Calmfors and Skednger 995; Jackman and Bellmann 996). Nevertheless, Layard et al. (99), OECD (993a), Nckell (997), and Nckell and Layard (999) fnd a sgnfcant negatve correlaton between the ALMP budget and the aggregate rate of unemployment. These studes focus on the competton effect as the man cause of the dentfed correlaton. Tranng and counsellng ncrease the compettveness of the partcpants of ALMP programmes and reduce the dscouraged-worker effect. Thus the labour supply grows, and the employed are confronted wth more and better qualfed compettors and wth a reduced transton rate nto employment when becomng unemployed. Consequently, they wll moderate ther wage demands, the wage-settng schedule wll shft downwards, and regular employment wll ncrease. By contrast, Calmfors and Skednger (995), usng regonal labour market data from Sweden, fnd a negatve correlaton between job-creaton programmes and regonal employment. The programmes crowd out regular jobs and, n the perod , had no sgnfcant postve nfluence on employment anywhere n Sweden. The results of Swedsh tranng programmes are mxed. It s, however, clear that tranng programmes show postve employment effects sgnfcantly more often than job-creaton programmes. Unlke Calmfors and Skednger, Büttner and Prey (998) n ther analyss of West German regonal data for the perod fnd no sgnfcant effects of tranng programmes whle job-creaton programmes are postvely correlated wth regonal employment. Bellmann and Jackman (996) estmate the employment effects of specfc ALMPs usng cross-sectonal data from 7 OECD countres for the years The study ncludes the budget of the PES, expenses for tranng and job-creaton programmes, and wage and ncome subsdes as explanatory varables. The

6 4 authors fnd no sgnfcant correlaton between these ALMPs and aggregate employment. However, all four polcy nstruments have a sgnfcant nfluence on the proporton of long-term unemployed: Tranng and the budget of the PES reduce the ncdence of long-term unemployment whle job-creaton programmes and subsdes ncrease t. Mllard and Mortensen (997) use the matchng model wth endogenous separaton rate ntroduced by Mortensen and Pssardes (994, 999) and supplement t wth a fnte duraton of unemployment nsurance (UI) benefts, wth payroll taxes, severance payments, and a hrng subsdy for newly created jobs. The authors show that the hrng subsdy s postvely correlated wth the steady state rate of unemployment. They calbrate the model to the US labour market and fnd that a hrng subsdy of 0% of the wage bll would cause the unemployment rate to ncrease by.7 percentage ponts. Followng Pssardes (979), our smulaton model allows for the fact that n Europe government organsed placement servces (PES) and prvate search technologes coexst. Frms choose between the two search methods, and the unemployed who have not been placed by the PES decde whether to search for a job on ther own. Sklled (type ) and unsklled (type ) unemployed are regstered separately wth the employment servce. Moreover, the PES offers frms lookng for (un-)sklled workers the opportunty to post ther vacances. After the regstraton has termnated, the PES combnes the regsters and arranges a certan number of matches. Most matchng models assume that the unemployed have suffcent ncentves to actvely search for a job. Yet, actve job search s costly, and therefore the endogenous rate of job seekers s only a fracton of the total number of unemployed. Certanly, search rules compel the unemployed to engage n actve job search, but, n spte of the mpendng penaltes, the PES s not at all able to get all unemployed who have not been placed to actvely search for a posted vacancy. Whle the PES ex ante puts workers and notfed vacances n separate regsters accordng to ther type, n the search market ths separaton happens ex post: Frms rank ther applcants accordng to the expected market value of the flled job. A second-rate worker s employed only f no frst-rate worker apples. The rankng order n our model s endogenous and nfluenced by the ALMP. The structure of the paper s as follows. Secton ntroduces the PES wth ts regsters for unemployed workers and vacances. Secton 3 descrbes the search market. Secton 4 contans a smulaton of the substtuton and crowd-

7 5 ng out effects of ALMP, and experments wth the followng parameters: () the effcacy of the placement actvtes of the PES, () the percentage of unemployed who partcpate n ALMP programmes, (3) the percentage of unemployed who are unwllng to work, (4) the productvty of the unsklled, (5) the share of sklled workers among the labour force, (6) the dfferent hrng subsdes, (7) the assstance for problem groups, and (8) re-employment bonuses and penaltes desgned to ncrease the number of actve job seekers among the unemployed. The results are summarsed n Secton 5.. THE PUBLIC EMPLOYMENT SERVICE MATCHING TECHNOLOGY. As n Germany, Italy, and Japan (OECD 996a), the PES s a fully ntegrated system coverng the three functons: placement servce, UI, and ALMP. The placement actvtes of the PES are represented by a matchng functon e M( µ U, R ) whch ndcates the number of matches for type- workers f U unemployed and R vacances are on fle n the regsters for type. A percentage µ of the unemployed s partcpatng n ALMP programmes so that of all U workers only µ U are avalable for the placement actvtes. Both frms and workers have unobservable characterstcs and ex ante are not entrely dstngushable from ther compettors. But once contact s made, both sdes know wth certanty whether the partner meets ther expectatons, an event that wth respect to the gven placement effcacy of the PES wll happen wth the probablty e. The parameter e s an effort varable expressng the screenng capacty of the PES and the effort of ts agents. The matchng functon M s assumed to be dfferentable, concave, and homogenous of degree one, and to have postve dervatves n both arguments. For type- workers the transton rate nto employment s P Θ, µ, e ) ( e M (, Θ / µ ) = e M( µ U, R )/ µ U and the arrval rate at the regstered vacances s Q ( Θ, µ, e ) em( µ / Θ,) = em( µ U, R )/ R, where Θ measures the tghtness R / U between the number of regstered vacances and the number of regstered unemployed. As the tghtness decreases, the arrval rate tends to zero and the transton rate tends to nfnty, an ncreasng tghtness drves P to nfnty, whle Q approaches zero. If the PES augments the share of unemployed who partcpate n ALMP programmes ( µ decreases), the transton rate ncreases whle the arrval rate decreases. Wth growng probablty of a match, e, both the transton rate and the arrval rate ncrease.

8 6 Only τ of the regstered unemployed of type are nterested n fndng a job. For smplcty we do not model the preferences of ths type explctly. How many of the unemployed are actually wllng to work s a queston that s rarely studed. Von Rosenbladt (99) reports that accordng to estmates of the German PES agents ntervewed, approxmately % of all regstered unemployed are not really lookng for a job. Some PES agents even guess that 60% of ther clents are not nterested n a new job. The nformaton about those preferences s asymmetrcal. Only the unemployed themselves know whether they prefer to lve on UI benefts. Snce avalablty for the placement servce s a precondton for UI payments, each unemployed must hde those preferences from nvestgatons by the PES. Ex ante, we assume, frms only know that the porton τ of the unemployed are not nterested n takng up work; ex post, all uncertanty about the job match s resolved, but nether the frm nor the PES are able to tell whether a rejected job match s due to a lack of nterest on the sde of the applcant or due to a msmatch, an event whch happens wth the probablty e. Thus, f the probabltes of the three events - contact, nterest, and apttude - are ndependent from each other, τ Q s the arrval rate of nterested applcants who meet the job profle. ASSET EQUATIONS. In the steady state, the asset equatons for the regstered vacances, the unemployed who are wllng to work, and the employed of type have the form: rv = k + τ Q ( Π V + G ) () R R rπ = y w + s( V Π ) () rn = B + P ( W N ) (3) rw = w + s( N W ) (4) TP = ( W N ) + ( Π + G ) (5) W R N = βtp, (6) where r denotes the nterest rate, k R the costs of regstraton wth the PES, y the output of a job occuped by a worker of type, w the wage measured n unts of output, B the reservaton utlty, and s the exogenous separaton rate. V, Π, N and W are the market values of a regstered vacancy, a flled job, a regstered worker, and an employed worker respectvely. Fnally TP s the present value of the match rent, the dstrbuton of whch s barganed blaterally between the frm and the job seeker. The result of the bargan depends on the parameter β, representng the barganng power of the job applcant. The nstruments of ALMP nclude hrng subsdes G R that n-

9 7 crease the quas-rent of a match. Hrng subsdes are pad to the frms that fll ther regstered vacances wth unemployed workers sent by the PES. REGISTERED VACANCIES. Supplers of vacances frst decde whether to advertse n the search market or to regster ther job wth the PES. If they prefer regstraton, they must choose between the two regsters for the dfferent types of workers, unsklled and sklled. Both decsons depend on the market value of the vacances. Frms prefer the search method and the regster that maxmses the market value of ther vacancy. Access to the search market and to the regsters s unlmted; thus, n equlbrum, vacances have the same value on all three submarkets. In addton, the perfectly elastc nflow of new vacances guarantees that n the steady state V = 0. Therefore t follows from () that the value of a flled job equals the expected costs of regstraton less the hrng subsdy: kr y w Π = GR =. (7) τ Q r + s FILLED JOBS. Π denotes the value of a job flled wth a worker of type. In the steady state, an nvestor who tes up captal n a flled job receves the permanent ncome rπ, and, n the case of job destructon, suffers a captal loss V Π, an event whch occurs wth the exogenous probablty s. Snce n equlbrum V = 0 t follows from () that the value of a flled job as n (7) equals the present value of the expected cash flow. From (7) we can derve the wage cost at whch market entry of vacances for type- workers no longer pays: w = y kr ( r + s) GR, (8) τ Q WORKERS. The permanent ncome rw of an employed worker s determned n equaton (4), where N W s the captal loss suffered by the worker when hs job s destroyed. In equaton (3), the Bellman equaton of the unemployed human captal, B s the endogenous reservaton ncome for whch B = b ( P ) γg ; here b represents UI benefts and ( P ) γg the expected penalty for volatons of the search rule. Unemployed workers who do not receve a job offer through the PES must actvely search for a job. The probablty that an unemployed worker wll not be assgned a regstered vacancy by the PES s P. If he does not search actvely but prefers to wat for future job offers through the PES he volates the search rule. The PES wll detect the

10 8 shrker wth probablty γ and reduce hs UI benefts by a penalty equal to such as n Swtzerland (OECD 996b). g For the captal gan W N that an unemployed worker realses upon transton nto employment t follows from (3) and (4) that W w B w rn N = =, (9) r + s + P r + s where w denotes the wage of a worker who accepts a regstered vacancy offered through the PES. WAGE BARGAINING. The quas-rent of a match s gven by equaton (5) consderng that V = 0. The frm and the job applcant negotate the wage. As n the generalsed Nash soluton, they dstrbute the rent so that equaton (6) holds n equlbrum. β measures the barganng power of the worker and s a constant between zero and one determned by the prevalng labour law and socal legslaton. Wth (3) and (5) through (9) the result of wage barganng s w = B + β ) β ( y B ) + βθ kr / µ τ + ( r + s G. (0) R The wage ncome conssts of two components: the endogenous reservaton ncome B and the nsder ncome. The nsder ncome s the sum of the nsder s share of the statc quas-rent β ( y B ), hs share of the transacton costs β Θ k R / µ τ, and hs share of the subsdy β ( r + s) G R. The share of the search costs depends on the tghtness Θ / µ τ measured n effcency unts that prevals n the PES-admnstered market segment for type- workers, where Θ represents the tghtness between the regsters for vacances and for unemployed workers, µ the porton of unemployed workers who are avalable for the placement actvtes by the PES, and τ the porton of unemployed who are nterested n fndng a job. The greater the number of workers who ether partcpate n ALMP programmes or are unwllng to work ( µ and τ decrease), the hgher the wage (0). Snce the probablty of a match, e, has a symmetrcal effect on the transton rate and the arrval rate of type, t has no drect nfluence on the wage level. PES EQUILIBRIUM. The PES-admnstered labour market segment for workers ( of type s fully charactersed by the wage and the tghtness w, Θ ). In equlbrum, the wage (0) reaches the level of wage costs (8), and the nflow of new vacances nto the regster for workers of type stops. The wage-settng

11 9 functon (0) gves us a strctly monotoncally ncreasng relatonshp between wage and tghtness n the w, Θ ) plane; the entry condton (8) gves us a ( strctly monotoncally decreasng schedule. The ntersecton of the two schedules determnes the equlbrum of the PES-admnstrated labour market segment for workers of type. COMPARATIVE STATIC ANALYSIS. Passve labour market polcy (PLMP) through hgher UI payments, b, leads to an upward shft of the wage-settng schedule (0) n the ( w, Θ ) plane; consequently, the wage rate ncreases, the tghtness of the regster Θ decreases, and both the value of a flled job (7) and the ncentve to accept the job (9) decrease. Due to the dstrbuton rule (6), the ncentve (9) and the value of the flled job (7) always change n the same drecton. ALMP n the form of a hgher probablty γ that shrkng s detected or a hgher penalty g decreases the reservaton ncome and shfts the wagesettng schedule downwards so that the wage decreases, the tghtness ncreases, and both the market value and the ncentve to accept the job ncrease. In contrast, the hrng subsdy G has an ncreasng effect on the R wage and the tghtness of the regsters. The probablty of a match e works only through the entry condton (8) causng t to shft upwards. Thus, the wage and the tghtness ncrease, and the ncentve to accept a job offered through the PES decreases. Both the porton of unemployed workers who are avalable for the placement actvtes of the PES, µ, and the porton of those wllng to work, τ, move the entry condton (8) upwards and the wage-settng schedule (0) downwards; thus, the tghtness between the regsters for type ncreases when those portons grow. At frst sght any reacton of the equlbrum wage to changes n µ and τ seems possble. Whether the wage ncreases, decreases, or remans unchanged depends on the reacton of the arrval rate τ Q as equaton (8) shows. Implctly dfferentatng (8) and (0) one can show that the arrval rate τ Q does not react to changes n µ but ncreases wth τ even though a hgher tghtness reduces the frequency of contacts Q. Thus, usng (8), t follows that the equlbrum wage ncreases wth τ whle t does not react to µ. Wth growng productvty y both the wage-settng curve (0) and the entry condton (8) are shfted upwards so that the wage ncreases. However, the tghtness between the regsters also ncreases because the shft of the wagesettng curve s smaller than the shft of the entry condton. Wth growng productvty the current proft, y w, also ncreases. Thus, n the natural equlbrum (wthout ALMP), the market value of a job occuped by a type- worker

12 0 s hgher than the market value of a job that s flled wth a type- worker, snce y > y and, moreover, snce all unemployed get the same UI benefts b and have the same barganng power β. 3. THE SEARCH MARKET Unemployed workers who have not been placed through the PES can and because of the search rule must make an effort to screen the search market for vacances. But actve job search s tme-consumng and produces monetary and psychologcal costs. Moreover, wth a growng number of actve job seekers, the congeston externaltes from search ncrease, the transton rates nto employment decrease, and the expected gan from search dsappears, such that the search strategy has no advantage n comparson to watng passvely for the PES to fnd an approprate job. Therefore, n spte of the mpendng penaltes, n equlbrum only a certan fracton of the unemployed workers who have not yet been placed by the PES are actvely engagng n job search. Frms post vacances on the search market as long as the market value of the advertsed vacances s at least as hgh as that of the regstered ones. Snce there are no barrers to entry the perfectly elastc nflow of new vacances contnues untl the value of the advertsed jobs drops to zero. Frms want to fll ther vacances as soon as possble. Consequently, job advertsements are not type-specfc so that job seekers of both types apply. Frms employ those applcants who maxmse the expected market value of the flled job. For that purpose they rank the applcants, acceptng a second-rate worker only f no frst-rate worker shows up. MATCHING TECHNOLOGY. The arrval rate q for the S actve job seekers among the U unemployed type- workers at the A advertsed vacances s generated by the transacton technology of the search market. The number of matches s determned by the matchng functon f m( a S, A). f s a shft factor whch can be nterpreted as the probablty of a match. The matchng functon of the search market has the same propertes as that of the PES. The arrval rate of job seekers of type s gven by q q θ, a ) f m( a / θ,) = ( f m( a S, A) / A. Due to search externaltes, the arrval rate q s a monotoncally decreasng functon of the tghtness of submarket, θ = A / S, and a monotoncally ncreasng functon of the search ntensty a.

13 RANKING ORDER. Frms rank applcants accordng to the expected market value of the flled jobs. Let + represent the group of frst-rate workers and the group of second-rate workers, then the market values of the flled jobs are J + J. The rankng order protects frst-rate applcants aganst the congeston externaltes caused by the group of second-rate workers. In contrast, the transton rate of second-rate workers not only depends on the tghtness n the submarket of the lower rankng type, θ A / S, but also on the tghtness + = + = θ A / S of the hgher rankng segment of the search market, p p + p( θ p( ) = θ q( θ ) () θ, θ ) [ ( )] ( ) + = q θ θ + q θ. () + Due to postve externaltes from search, both transton rates ncrease strctly monotoncally wth the tghtness of the respectve submarket. In addton, the transton rate of the second-rate group grows strctly monotoncally wth θ. + ASSET EQUATIONS. The Bellman equatons of the advertsed vacances and those employed who have found a job through random search are rv re TS + { J J } qva A = k( A) + q ( q) J + q( q) J + qq max, (3) = ω + s( N E ) (4) = ( E N + G ) J A, (5) where k (A) s the advertsement costs of a vacancy as a functon of the aggregate number of advertsed vacances, q s the arrval rate of job seekers at an advertsed job, q = q + q qq, V A s the market value of an advertsed vacancy, J s the market value of a flled job, E s the value of an employed worker of type who has found a job through the search market, ω s the wage rate negotated by the frm and the applcant, TS s the quas-rent of the match, and G s a (re-employment) bonus whch the PES as n Japan A (OECD 993c) pays to actve job seekers. The space for advertsements s scarce. Thus, the frms search costs ncrease strctly monotoncally wth the aggregate number of ads, k > 0. FIRMS. The nflow of advertsed vacances nto the search market stops as soon as V = 0. Therefore, n the steady state, the value of an occuped job s A J = ( y ω ) /( r s), (6) +

14 wth y denotng the output of the job and ω the wage rate measured n unts of output. Furthermore, usng V A = 0, t follows from (3) that n the steady state H A ( θ, θ, A) q J + ( q ) q J k( A) = 0. (7) WORKERS. Equaton (3) determnes the permanent ncome of an unemployed worker, rn, and equaton (4) the permanent ncome of an employed worker who has found a job through actve search, re. The captal gan = E N that an actve job seeker realses upon transton nto employment follows from (3) and (4): E N ω = B P ( W r + s N ), (8) where the ncentve to accept a job placement by the PES, W N, s determned by equaton (9). The greater the gan from a PES placement, and the hgher the probablty of such a match, P, the smaller the ncentve to engage n actve search. WAGE BARGAINING. The (re-employment) bonus G A s part of the quas-rent (5) that s dstrbuted between the frm and the applcant through the wage rate ω so that n the barganng equlbrum E N + GA = βts. (9) From (5), (9), (3), and (4) t follows that the wage for type- workers who have found a job through random search s ω = βy + ( β ) rn ( β )(r + s). Insertng (3) agan and takng account of (5), (6), and (7), we G A get the result of the barganng between the partes n the search market: ω = B + β( y B ) + βθ kr / µ τ ( β )( r + s) GA. (0) The regstraton costs k R and the tghtness of the PES-admnstered labour market measured n effcency unts, Θ / µ τ, also play a crucal role n the wage negotatons n the search market. Ths s because an actve job seeker can base hs wage demands on the permanent ncome of an unemployed worker (3), whch s the sum of the reservaton ncome B and the expected captal gan P ( W N ) whch he can count on n case the negotatons fal. However, t follows from (5), (6) and (7) that the captal gan W N upon transton nto a job offered through the PES equals the expected regstraton

15 3 costs weghted wth the relatve barganng power β /( β ). From ths we can nfer that the wage on the search market s a functon of Θ / µ τ. k R Comparng (0) wth (0) we can see that wthout nterventon of the PES n the submarkets for type- workers, the law of one wage holds. Hrng subsdes and (re-employment) bonuses drve a wedge between the wage barganed n the search market and that barganed n the PES-admnstered labour market such that the wage dfferental for type- worker s non-negatve, w R A ω = ( r + s)[ βg + ( β ) G ] 0. Search costs mply that we could observe a strctly postve wage dfferental and at the same tme a strctly postve number of actve job seekers, whch s strctly smaller than the number of unemployed who have not been placed by the PES. ACTIVE JOB SEARCH. The captal gan (8) and the bonus G are the ncentves to search actvely for an advertsed vacancy. However, actve job search A s costly. The number of unemployed workers who decde to actvely engage n job search ncreases untl ether all unemployed workers of one type who have not been offered a job by the PES search actvely, or H and H +, the gans from search for frst-rate and second-rate applcants, dsappear, and the unemployed are ndfferent between actve search and passve watng for a job offer through the PES H H ( θ ) p ( G + ) c( a ) = A + + () + ( θ, ) ( + ) ( ) = 0 θ + p G A c a, () where p s the transton rate generated by the matchng technology of the search market, and c ( a ) are the prvate search costs for type- job seekers that are a functon of the ntensty of the search. If we assume that the partcpants n ALMP programmes also have an opportunty to search actvely for a job, the upper lmt for the number of actve job seekers of type, S, s determned as follows. Out of U unemployed P µ U are well matched and could sgn a contract but only τ are nterested n acceptng the offered job. Therefore we have S τ ( µ P ) U. In the followng, we analyse equlbra wth nteror solutons so that for both types of workers: 0 < S < τ ( µ P ) U. SEARCH MARKET EQUILIBRIUM. A search market equlbrum wth the varables ( A, θ, θ ) denotng the number of advertsed vacances, the tghtness n both + submarkets, and the endogenous rankng order s charactersed by the equatons (7), (), and (). Snce the ncentve to search and the rankng order

16 4 are exogenous wth respect to the search market, the equaton system s recursve. Frst we determne θ usng (), then θ usng (), and fnally the + number of advertsed vacances usng (7). Each perod, out of L workers of type, s( L U ) lose ther job. At the same tme, P τ µ U + p S unemployed fnd a new job. In the steady state, the nflow nto the pool of unemployed equals the outflow so that the type-specfc rates of unemployment are gven by u = U / L and the aggregate rate of unemployment u u = s, =, s + τ µ P + σ p, u = λ u + λu, (3) where σ = S / U s the rate of actve job seekers among unemployed workers of type, and λ s the fracton of type- workers among the labour force L = L. 4. SIMULATION In order to measure the crowdng-out and employment effects of the ALMP, we frst smulate a basc scenaro wthout actve measures (Table A n the Appendx). Then we experment wth the nstruments of ALMP and show ther effects through comparson wth the basc scenaro (3) of the mxed economy (Tables A and A3). The vector of polcy nstruments s e, µ, τ, y, ( GR, GA, γ, g ) : e measures the effcacy of the placement servce wth regard to the unemployed workers of type, µ and τ are the portons of the unemployed who are avalable for job placement by the PES and who are wllng to work, y denotes the productvty of type- workers, G s a hrng subsdy for R frms that have flled ther regstered vacancy wth an unemployed worker of type sent by the PES, G s a (re-employment) bonus for workers who have A found a job through the search market, γ represents the probablty that an unemployed worker who does not engage n search s detected by the PES, and g s a penalty for the unemployed who are not searchng actvely. PARAMETERS AND MATCHING FUNCTIONS. To the extent that estmates for Germany exst, we base our choce of parameters on those values. Where we have degrees of freedom, the choce s guded by the ntenton of generatng steady states wth nteror solutons and type-specfc as well as aggregate rates of unemployment smlar to the German rates for 998. Table A n the Appendx shows the parameters of the basc scenaro. Takng nto account the

17 5 barganng power of workers n Germany and other OECD countres n Europe, we assume that β = As n Germany (OECD 997), a match has an expected lfetme of / s = 0 perods; an unsklled worker has 70% of the productvty, and UI benefts amount to 37.5% of the productvty of a sklled worker. In the basc scenaro of the mxed economy (s. Table A) wth b = 37.5 the replacement rato for the sklled s 39% and that for the unsklled s 56%; usng the fractons of those groups n the labour force as weghts we can calculate a weghted replacement rato of 44%. Accordng to Franz (999, p. 66), replacement ratos n Germany are 4,9% for men and 38,8% for women. Unt costs for the advertsement of a vacancy are a lnear functon of the aggregate number of vacances, k ( A) = k A, where k = The search costs for an actve job seeker are a c, wth c = 30. The ntensty of the search depends on the poston of the job seeker n the rankng order of the search market. Frst-rate job seekers make up to a =. 3 and second-rate job + seekers up to a = 0. 6 applcatons per perod. The matchng functons of the PES and the search market are of the Cobb-Douglas type Φ Q ( Θ, µ, e ) = e ( µ / Θ ), (4) φ q ( θ, a ) = f ( a / θ ), =,, (5) wth Θ = R / U, and θ = S / A. For the scale factor f of the matchng functon representng the search market t s assumed that f = Wthn the PES technology, the unemployed domnate the number of contacts whch the PES arranges wth an elastcty of Φ = 6 / 7. In the search market, on the contrary, the advertsed vacances wth an elastcty of φ = 7 / 0 determne the number of successful job fllngs. By comparson: Burda and Wyplosz (994) estmate elastctes of the number of unemployed between 0.70 and 0.80 for matchng technologes wth constant returns to scale, and Burda (994) reports an estmate for western Germany of BASIC SCENARIO. The basc scenaros n Table A nclude the steady states of three regmes. () s a lassez-fare economy wth a partally ntegrated PES responsble only for UI. Unemployed workers can fnd a job only through actve search. In () there s no prvate search market. Only the PES regsters jobs and vacances and arranges matches between them. In ths economy there s no rankng of the unemployed. The separaton of types takes place ex ante when jobs and workers are assgned to dfferent regsters. (3) represents the mxed economy wth a fully ntegrated PES and a search market where workers who have not been placed by the PES randomly search for a job.

18 6 Under the regmes () and (3), job seekers base ther wage demands not only on the UI benefts but also on the placement actvty of the PES. Compared to the lassez-fare economy, the placement servce obvously causes a redstrbuton from proft to wage ncome. In the mxed economy, the wage for sklled workers ( w = 96. ) s 8.3% hgher and the wage for the unsklled ( w ) s.6% hgher than n the lassez-fare economy. The search = rates n the lassez-fare economy are σ = 90 % and σ = 60%, where σ = S / U, snce n ths economy all unemployed who are wllng to work prefer to search actvely, such that σ = τ. A completely ntegrated PES wth job placement actvtes curbs the ncentve to engage n actve job search. In the mxed economy the search rate of type- workers s only 48% and that of type- workers s 3.%. Snce the number of job seekers has a postve nfluence on the number of advertsed vacances va search externaltes, a fully ntegrated PES also reduces the number of advertsements from A = 85 under the lassez-fare regme to A =. 4 n the mxed economy. If we deduct prvate and publc search and placement costs from the gross product of the economy we can take the resultng per capta consumpton C as a welfare measure C = λ [( u ) y σ uac krr / L ] k A / L kp u. (6) = Wth placement costs of the PES equal to k = 40 per unemployed worker we have a per capta consumpton of 77. n the mxed economy and 75.5 n the PES-admnstered economy. Under the lassez-fare regme, on the other hand, per capta consumpton s only The placement costs of the PES P would have to more than double ( k = 90 ) n order for the per capta consumpton n the mxed economy to drop to the level observed n the lassezfare regme. P The last columns of Tables A and A3 contan ndces for the actve and passve measures of labour market polcy. Here, PLMP s defned as UI benefts per unemployed worker dvded by the gross product per capta of the workng populaton. The gross product s BP = L λ ( u ) y. (7) = Fnally, the expenses per unemployed worker for ALMP are measured as a percentage of the gross product per capta of the labour force and the sum of

19 7 the expenses for actve and passve labour market polces are measured as a percentage of the gross product, ( P + A) / BP. The expenses for PLMP amount to 5.9% n the lassez-fare economy, 5.5% n the mxed economy, and 7.5% n the PES economy. By way of comparson, n the years Denmark s expenses for PLMP were 4% of GDP, n Germany the rate was.5% (OECD 998). Unlke the GDP, the gross product as defned n (7) does not nclude publc expendtures or nvestments. PLMP s 46.4% n the mxed economy, and 46.9% n the lassez-fare economy; n the PES economy, due to a hgher rate of unemployment among the unsklled, the percentage s 48.3%. Result : Not only the UI benefts but also the publc placement servce per se consderably strengthens the workers barganng poston and causes a redstrbuton from proft to wage ncome. Moreover, the placement servce curbs the ncentves to engage n actve job search, and reduces the search rates and the aggregate number of advertsed vacances. Nevertheless, the per capta consumpton may be hgher, and the aggregate rate of unemployment as well as the expendture rato may be lower n the mxed economy than n the lassezfare regme. PES PLACEMENT, JOB SEARCH ASSISTANCE AND EMPLOYMENT TRAINING. The placement effort of the PES s the most mportant nstrument of ALMP. Tables A3. and A3. show varatons of the placement effcacy manfestng n the match probablty e. e denotes the probablty that a contact between a regstered worker of type and a regstered vacancy wll lead to a job contract. Wth growng e the tghtness between the regsters for type ncreases and so does the PES transton rate P. For example for e = we fnd an ncrease n the tghtness between the regsters of R / R U U = 3. ( 5.) = /.0%. The PES and the search market are competng matchng devces. Wth hgher effcacy, the PES crowds out actve job seekers and advertsed vacances. Wth the probablty of a match ncreasng to e =, the number of type- job seekers falls by 9.%, S / S = 9.. The smaller number of actve job seekers reduces the ncentve to advertse vacances. Thus, compared to the basc scenaro, there are 8.4% fewer ads. Durng the adjustment, actve job seekers of type are subject to two opposng forces. On the one hand, the search market becomes more attractve for workers of type because the congeston externalty from the hgher-rankng type s reduced; on the other hand, the number of advertsed vacances drops. The negatve effect domnates, and the number of actve job seekers of type decreases by 4.6%. The number of unemployed type- workers s reduced because both the transton

20 8 rate of the PES and the transton rate of the search market ncrease, U / U = 5.. Whle the PES transton rate for type- workers remans unchanged, ther transton rate n the search market s reduced, and the number of unemployed of type grows, U U. 4. The unemployment rate for / = type decreases by 0.5 percentage ponts to 8.5%, whle the unemployment rate for type ncreases by 0.3 percentage ponts to 8.7%. Nevertheless, the aggregate rate of unemployment s slghtly reduced, snce type at 70% represents the larger part of the labour force. Both search rates decrease. That of type s reduced by 7 percentage ponts when the placement effcacy = reaches e, and that of type by.9 percentage ponts. The hgher effcacy of the PES placement enables type- applcants to demand wages that are 0.3% hgher than those n the basc scenaro. Per capta consumpton grows by %, and PLMP decreases by 0. percentage ponts to 46.%, whle the expendture rato ( P + A) / BP falls by 0. percentage ponts to 5.4%. If the PES expands ts placement capacty for unemployed type- workers, ths reorgansaton has no nfluence on the transton and arrval rates of type- job seekers. Ths s because the PES has separate regsters for the two types, and n the rankng order of the search market, type s preferred over type. Wth growng e the number of actve job seekers decreases snce the chance of beng offered a job through the PES ncreases. For e = 0. 80, the number of actve type- job seekers decreases by 6.%, and the search rate σ falls by.8 percentage ponts to 9.4%. Due to the search externaltes, the number of advertsed vacances decreases, whch prompts type- job seekers to reduce ther search actvtes. The number of unemployed workers of type ncreases by.%, and ther unemployment rate u reaches 9.%. By contrast, the unemployment rate of type- workers falls by.5 percentage ponts to 6.%, and the aggregate rate decreases by 0.4 percentage ponts to.4%. Layard et al. (99) gve two arguments for the postve employment effects of ALMP. The competton argument says that the partcpants n tranng programmes become more compettve, thus ntensfyng competton n the labour market and reducng the upward pressure on wages. Accordng to the productvty argument, actve polces augment the productvty of the partcpants, thus ncreasng the demand for labour. To test these hypotheses we conduct the followng experments. Frst we ncrease the porton of unsklled workers partcpatng n ALMP programmes ( µ decreases, s. Tab. A3.3). Second we assume that schoolng programmes of the PES, the tranng of

21 9 prosocal behavour, the mprovement of search strateges etc., enlarge the number of unsklled who are able and wllng to work and, through that channel, ncrease labour supply ( τ ncreases, s. Tab. A3.4). Thrd actve measures rase the productvty of the unsklled ( y ncreases, s. Tab. A3.5). Fnally we can compare the consequences of the productvty effect generated by PES programmes wth an educatonal strategy whch would ncrease not the productvty of the unsklled but nstead the porton of sklled workers among the labour force ( λ ncreases, s. Tab. A3.6). If the PES ncreases the number of unemployed n ALMP programmes, µ decreases and as shown n Secton so does the transton rate of unemployed type- workers who are avalable for job placement P. Therefore, the rate µ P must fall as well. Thus, we can nfer from equaton (3) that the unemployment rate of the unsklled ncreases. In our smulaton t ncreases by.5 percentage ponts to 9.9%, f µ = Wth µ = 0. 90, 0% of the unsklled unemployed partcpate n ALMP programmes, that s 0.6% of the labour force. By comparson, n Germany n 997 the nflow nto ALMP programmes amounted to 3.6% of the labour force (OECD 998). If ALMP ncreases the porton of unsklled who are wllng and able to work, then the tghtness between the regsters for type ncreases, and so do the PES transton rate P and the rate µ P. However, the growng attractveness of the PES placement at the same tme reduces the number of actve job seekers among the type- unemployed, an effect that has a negatve mpact on the number of advertsed vacances and fnally also on the number of actve type- job seekers. Nevertheless, the tghtness n the submarkets for type remans unchanged so that wth respect to equaton () the ncreased tghtness n the search market for type- workers ncreases ther transton rate p. Snce the search rate σ also ncreases, u decreases. In our smulaton u decreases by.3 percentage ponts to 6.%, f τ = Smultaneously the per capta consumpton C grows by %, whle the expendture rato decreases by 0.4 percentage ponts to 5.%. Thus we see, that contrary to the receved competton argument t s not a lower wage whch stmulates the demand for type- workers but the ncreased arrval rate of the unemployed at the regstered vacances, τ Q. Wth growng τ all three rates P, µ P and τ Q ncrease and, due to the growng tghtness, there s even a small wage ncrease for type- workers.

22 0 If ALMP ncreases the productvty of type- workers, the numbers of both regstered and advertsed vacances for type- ncrease. The ncrease n advertsements also prompts the frst-rankng unemployed to ntensfy ther search so that the unemployment rates for both types drop. For y = 80 n the steady state, the unemployment rates reach values of 8.7% for type and 6.7% for type, s. Tab. A3.5. At the same tme, per capta consumpton grows by 4%, whle the expendture rato for actve and passve measures falls by 0.5 percentage ponts to 5%. Result : The probablty of a match, e, and the percentage of unemployed who are able and wllng to work, τ, produce crowdng out effects. Increasng the labour supply through actve measures whch rase τ causes a hgher transton rate nto employment as well as a hgher arrval rate at the regstered vacances such that the type-specfc and the aggregate rate of unemployment declne, although contrary to the competton argument the transtons are accompaned by a hgher wage w. Table shows the elastctes of the unemployment rates and the per capta consumpton wth respect to the dfferent actve measures calculated as unweghted averages of the smulaton results. If, for example, the probablty e that a contact between an unsklled worker and a regstered vacancy leads to a match ncreases by one percent, the unemployment rate for type- workers decreases by 0.63%, the aggregate rate falls by 0.4% and per capta consumpton ncreases by 0.05%. Table : Elastctes of unemployment rates and per capta consumpton wth respect to the probablty of a match e, the populaton portons µ, τ, y, and the productvty λ. u u u C e e µ τ y λ HIRING SUBSIDIES. Hrng subsdes, G, are pad only to frms that regster R ther vacances wth the PES and employ workers sent by the PES. Hrng subsdes augment the placement success of the PES, curb the dynamcs of the search market, crowd out job seekers and advertsed vacances, ncrease unemployment, drve a wedge between the barganed wage of the PESadmnstered labour market and the wage of the search market, and reduce

23 per capta consumpton. The PES mproves ts placement success through wage subsdes, because the latter have a postve effect on the transton rate of subsdsed workers P, on the number of unemployed workers U, and thus also on the number of new job contracts. However, the PES competes wth the search market, where the dynamc s weakened as soon as the PES subsdses the wage costs of frms wth regstered vacances. Especally among those workers who are elgble for subsdes, the number of actve job seekers decreases and the search rates drop. Then, due to negatve externaltes, the supplers of advertsed vacances wthdraw, and the number of actve job seekers decreases even among those unemployed who are not elgble for subsdsaton. In addton, subsdsed workers negotate over hgher wages, thus acqurng a share of the hrng subsdes the sze of whch depends on ther barganng power. Although hrng subsdes are pad only to regstered frms, they can nfluence the rankng order n the search market (Table A3.7). Wth a subsdy G 35, unemployed type- workers who have not been offered a job through the PES but have found one through ther own search effort, acheve a wage ncrease of.3% compared to the basc scenaro. Ths ncrease of wage costs for jobs that are flled through the search market s suffcent to change the rankng order n favour of type- workers. The new rankng order prompts a sharp ncrease n unemployment for workers of type who now rank second and a correspondng decrease n unemployment for type- workers who n the new steady state, are preferred by employers as a consequence of the wage cost effects of the hrng subsdy. R = In the ALMP column, the tables show the hrng subsdy per unemployed worker as a percentage of the gross product per capta of the labour force. In the OECD countres between 989 and 994, ALMP vared between 3.0% n the US and 59.3% n Sweden; the correspondng value for western Germany was 5.7% (Nckell 997). In our smulaton, ALMP s 4.9% for a hrng subsdy G 30 but only 5.7% for G 30. Tables A3.7 and A3.8 also show R = R = the expendture rato for actve and passve labour market measures whch, for example, for G 30 s 6.5%, one percentage pont hgher than n the basc R = scenaro. Fnally, the tables ndcate that hrng subsdes are concomtant wth declnng per capta consumpton. Wth G 30, for example, C s 0.9% R = lower than n the basc scenaro. Result 3: Hrng subsdes augment the placement success of the PES, but on the aggregate level takng nto account the effects of search externaltes they

24 crowd out actve job seekers and advertsed vacances, drve a wedge between the wage of the PES-admnstered labour market and the wage of the search market, reduce per capta consumpton, and ncrease the aggregate unemployment rate and the expendture rato. (RE-EMPLOYMENT) BONUSES. A (re-employment) bonus pad by the PES to actve job seekers who have found a new job through ther own search effort s an often recommended but rarely mplemented nstrument of actve labour market polcy. Bonuses mprove the dynamcs of the search market at the expense of the central placement servce, reduce aggregate unemployment, ncrease per capta consumpton, and decrease the expendture rato for actve and passve labour market polces, thus fnancng themselves. A bonus pad only to actve job seekers has no effect on the tghtness between the PES regsters. Therefore, the transton rates P of the regstered unemployed as well as ther arrval rates τ Q at the regstered vacances reman unaltered. In the search market, the number of advertsements grows wth the bonus, and postve externaltes ncrease the number of actve job seekers. For G = A 5 the search rate of type s already 30 percentage ponts hgher than n the basc scenaro so that σ = 78 %. In spte of the greater number of advertsed vacances, there s only a slght ncrease n the number of actve job seekers of type ; ths s due to the rankng order that job seekers are confronted wth n the search market. More vacances also mprove the chances for type, but the growng number of type- job seekers produces congeston externaltes, the postve and negatve externaltes almost compensatng each other. Promotng actve job search among the unsklled through a bonus G A produces smlar adjustments; n contrast to G A, G A ntally has no negatve external effects on type- workers, snce the rankng order protects unemployed workers of type aganst competton from type- workers. However, wth an ncreasng bonus, the wage ω negotated n the search market s reduced; wth G A = 5 the bonus has almost reached the level where the rankng order s reversed, and, because of the bonus, frms prefer to employ type- workers. Result 4: In a labour market wth several states whch are nterconnected by postve and negatve externaltes from search, the nstruments of ALMP cause not only deadweght losses and substtuton effects but also postve employment effects on unemployed workers and vacances not drectly supported by the PES. The effects of the bonus payments substantate ths result. In addton, the bonus for type- workers fnances tself. For G A = 5, the number of actve job seekers among the subsdsed almost doubles, ther rate of unemploy-