Urban Studies, Vol. 38, No. 10, 1733 –1751, 2001

Local Underemployment and the Discouraged Worker Effect

Maarten van Ham, Clara H.Mulder and Pieter Hooimeijer

[Paper Žrst received, April 2000; in Žnal form, December 2000]

Summary. The effectof poor local labour marketopportunities on occupational achievement is an important aspect of the spatial mismatchhypothesis. Much of the researchhas concentrated on the directlink betweengeographical accessto jobs and employment outcomes. In contrast, little attention has been given to the discouraging effectof poor chances on job searchactivities. The discouraged workereffect is deŽned as the decision to refrain fromjob searchas aresultof poor chances on the labour market. Discouragement effectscan arisefrom a lack of individual qualiŽcations, fromdiscrimination in the labour marketor froma high local level of underem- ployment. The empiricalŽ ndings of this paper, based on the Netherlands Labour ForceSurveys 1994– 97, show that discouragement can enterthe job searchprocess both at the stage of deciding to enterthe labour forceand at the stage of deciding to engage actively in ajob search.Gender differentials in discouragement arerevealed in the processof self-selectioninto the labour force. Poor labour marketchances lead to lessactivity in both off-the-job and on-the-job search, indicating arole of discouragement in the spatial mismatch.Individual qualiŽcations and ascribed characteristicsturn out to be moredecisive than the local level of underemployment.

1. Introduction Even though Kain’s ‘spatial mismatchhy- bias in studies aimedat estimating the com- pothesis’(Kain, 1968) was muting tolerance of the unemployed (Cooke and Ross, 1998). The second is the widening originally coined to describe abroad set of of the issue to encompass not only racebut geographical barriersto employment for also gender (Preston and McLafferty,1999). African-Americaninner city residents The third is the detailed measurementof (Preston and McLafferty,1999, p. 388) geographical access to appropriate jobs using it has also stimulated moregeneral research GIS, linking this access to the level of occu- on the effectof poor job access on occu- pational achievement (Hanson et al., 1997; pational achievement. This researchhelps to Ong and Blumenberg, 1998; van Ham et al., understand the variety of mechanisms that 2001). underlie the original hypothesis. Research in Several empiricalstudies have focused on the 1990s has shown majoradvancement in the inuence of spatial restrictions on em- three areas.The Žrst is uncovering selection ployment rates(for example, Ong and Blu-

Maarten van Ham, Clara H.Mulder and Pieter Hooimeijer are in the Urban Research Centre (URU),Faculty of Geographical Sciences, Utrecht University, POBox 80.115, 3508 TCUtrecht, The Netherlands. Fax: 31 30 253 2037. E-mail: [email protected]; [email protected]; and [email protected]. Maarten van Ham’s research was supported by the Netherlands Organization for ScientiŽc Research (grant no. 42513002). Clara Mulder’s research was made possible by afellowship from the Royal Netherlands Academy of Arts and Sciences.

0042-0980 Print/1360-063X On-line/01/101733-19 Ó 2001 The Editors of Urban Studies DOI:10.1080/ 00420980120084831 1734 MAARTENVAN HAM ET AL. menberg, 1998; Immergluck, 1998) and gen- ing engaged in job search. Individual charac- der differences in labour participation (for teristics (eitherreal or ascribed) and the local example, Hanson and Pratt, 1988, 1990, level of underemployment areboth con- 1991). However, no direct empiricalevi- sidered potential sources of discouragement. dence has been found of arelationship be- We show that discouragement can enter the tween spatial restrictions and job search. Yet job search process attwo different stages. this relation is crucial, as job search is a The Žrst stage concerns the decision to par- prerequisite forlabour marketparticipation ticipate in the labour market.At this stage, and careeradvancement. The jobless search people select themselves into or out of the to escape unemployment and those already in active labour force.This selection clearly has ajob search to Žnd abetter one (Mortensen, an effecton their chances of employment, as 1986). The relationship between poor the potentially less successful will refrain chances in the labour marketand the inten- fromparticipation moreoften. The second sity of job search has been expressed in the stage is the decision to engage actively in job discouraged worker hypothesis (Fisher and search, either on or offthe job, once one is in Nijkamp, 1987). The hypothesis states that the active labour force.In this second stage, people with poor labour marketexpectations selection effectsare expected, asthe discour- become discouraged in their job search and aged worker hypothesis stipulates that low leave or failto enter the labour force,be- chances of being unemployed will have a cause the probability of Žnding asuitable job negative effecton the search intensity. aftera certain period of timeis low. The remainderof this paper is organised in Poor labour marketchances can result four parts. Section 2describes atheoretical fromindividual characteristics, fromdis- frameworkwithin which (gender-related) crimination in the labour market,and froma discouragement effectsin the various stages high level of local underemployment, which of the search process can be understood. indicates amismatchbetween demand and Section 3introduces the data and methodol- supply on the local labour market(Simpson, ogy. The method consists of aseries of three 1992). Evidence fromstudies using USdata logistic regression models which areused to (Parsons, 1991; Keith and McWilliams, estimate:the chances of being in the active 1999) and British data (van Ophem, 1991) labour force;the chances of being unem- indicates that women areless likely than men ployed, given the factthat people arepartici- to be engaged in job search. Women are pating on the labour market;the chances of morespatially restricted than men (Hanson being in search of ajob dependent on and Pratt, 1988), so gender differences in whether one is employed. Section 4reports search behaviour may be explained in part by the results of the empiricalvalidation of the gender differences in the discouraged worker models. Selection effectsare measured, using effect.In general, discouragement can be the two-step Heckman procedure, and also regarded as anextra mechanism that hampers given asubstantive interpretation in termsof the occupational achievement of groups with discouragement. The Žnal section comprises poor chances on the labour market,like asummaryand adiscussion of the implica- Kain’s inner-city African-Americanresi- tions. dents, and researchon discouragement might thereforecontribute to amoregeneral under- 2. Job Search: Theory and Background standing of spatial mismatches. The aimof this paper is to Žnd empirical To explain job search and the inuence of evidence forthe discouraged worker hypoth- local underemployment on job search, we esis by looking atdirect evidence of job use insights fromvarious theoretical points search activity. The main issue is the extent of view. We commencewith job search the- to which poor labour marketchances have a ory and human capital theory. Individual and discouraging effecton the probability of be- household restrictions areconsidered, paying LOCAL UNDEREMPLOYMENT 1735 special attention to racialand gender differ- search; people search in order to avoid un- ences. Finally, job search is placed in aspa- deremployment. For unemployed people, tial context and the discouraged worker there areno returns on previous investments effectis worked out in moredetail. in human capital. The higher the level to which an unemployed person has been edu- cated, the greateris the loss of income, so the 2.1 Job Search Theory moreintensive is the job search. For the Since the seminal papers of Stigler (1961, employed, the effectof human capital on job 1962), job search theory has conclusively search cannot be seen independently from become one of the main theoretical and em- the level of their present job. The human pirical tools forunderstanding the working of capital of an employed person is best utilised the labour market.In the past four decades, when that person’s job level and education labour economists have produced an exten- level arein keeping. Workers therefore sive body of researchrelated to job search search moreintensively when the educational theory (Lippman and McCall, 1976; Kiefer requirements of their job arelower than their and Neumann, 1989; Devine and Kiefer, level of education (seeSimpson, 1992). On 1991). In the basic sequential job search the basis of the foregoing, it can be hypothe- model, individuals choose areservation sised that the probability of being engaged in wage; this isthe lowest wage level atwhich an off-the-job search increases with educa- they would be willing to accept ajob. Ajob tional level. Itis further expected that, fora offerwould only be accepted ifthe wage given level of education, the probability of offerwere at least as high as the reservation being engaged in on-the-job search decreases wage. The arrivalrate of job offersdepends with the level of the job. on an individual’s search intensity; this in In addition to job level, other job charac- turn depends on the potential gains of the teristics can also indicate that aworker’s search (seeMortensen, 1986). By varying present job is sub-optimal, given past invest- search intensity, individuals can inuence ments in human capital. According to Blau the search outcome. Ifan offeris accepted, a (1991), the number of hours worked per worker maycontinue to search on-the-job week is an important determinant of on-the- until abetter job is found. Job search theory job search, because the returns on invest- is based on the idea that individuals maxi- ments inhuman capital aremaximised when miselifetime utility by moving through dif- aworker is employed full-time.The returns ferentstates; the theory isexplicitly dynamic. on previous investments in human capital are Over their lifetime,people adjust their reser- best assured in asecure job, so job security vation wage. They increase their job search also plays an important part in job search intensity when they areunderemployed – that (van Ophem, 1991). Jobs with apermanent is, when their present income falls under employment contract and regular working their reservation wage. hours offerthis security. Itis thereforeto be expected that the probability of being en- gaged in job search increases when aperson 2.2 Human Capital and Underemployment is employed part-time,works irregularhours According to the human capital theory or does not have an employment contract. (Becker, 1962), people invest in productiv- Most job mobility occurs in the Žrst dec- ity-enhancing skills and strive to maximise ade of work experience (Topel and Ward, the utility of this accumulated capital. Hu- 1992). Job shopping enables individuals to mancapital accumulates over alifetimein try out several jobs to determine their com- the formof (formal)education and working parative advantage (Johnson, 1978); Žnd experience. When, given past investments in higher-quality job matches (Jovanovic, human capital, the labour marketposition of 1979); and achieve better pay (Parsons, an individual is sub-optimal this leads to job 1973; Burdett, 1978). People accumulate hu- 1736 MAARTENVAN HAM ET AL. man capital with age through their work sibilities. Many women areplaced outside experience; their human capital becomes the labour marketas aresult, because of their morespeciŽ c. The costs of ajob change are domestic responsibilities and restricted ac- considerable when aworker with accumu- cess to childcare facilities(Bowlby, 1990). lated speciŽc human capital moves to ajob Restrictions also cause women to preferpart- wherethese speciŽc skills cannot be utilised. timejobs, because these enable them to com- Furthermore, the pay-off period forsearch bine domestic work with paid employment. and job change costs becomes shorter as age We deduce fromthe above that, even increases. The probability of being engaged when women decide to participate in the in job search is thereforeexpected to de- labour market,the domestic workload in creasewith age. combination with the presence of young chil- dren mayrestrict the opportunities of search- ing fora suitable job. We expect women to 2.3 Household Situation and Gender participate less in the labour marketthan The labour forceparticipation of women is men, and forwomen who do participate to be much lower than that of men. Women are less frequently engaged in job search than also less often engaged in job search than men. We further expect the probability of men(Keith and McWilliams, 1999). Men women being engaged in job search to de- traditionally have afull-timejob and only a creaseif young children arepresent in the smallproportion of the malelabour force household and the effecton job search of would voluntarily step out of the labour mar- working part-timeto be less strong for ket. In contrast, many women seemto have women than formen. other priorities than paid work. For awoman to stay outside the active labour forceand 2.4 Spatial Restrictions and Discouragement become afull-timehousewife is an accept- able alternative, especially when there are Labour economists traditionally look atspa- young children in the household. Making tial restrictions in termsof the monetary such achoice is inconsistent with the as- costs of migration and commuting. Commut- sumption that all individuals maximisethe ing costs lead, forexample, to adjustment of utility of their accumulated human capital. the reservation wage – the minimumwage a The new home economics theory (Becker, worker is willing to accept fora job ata 1976, 1991) offersa theoretical framework certain location, given his or her location of that resolves this inconsistency. According to residence. Thereforejob search intensity this theory, the labour participation decision rises signiŽcantly with rising commuting of amother is purely Žnancial and depends time(van Ommeren,1996). on her earning capacity. Ifa mother’s earn- Spatial restrictions are,however, more ing capacity is low, she will decide to be- than just the costs of covering distance. For comea full-timehousewife. Mothers who the majority of the workforce, the set of job have ahigh earning capacity maydecide to opportunities that is actually available or se- participate inthe labour marketand contract- riously considered is highly constrained spa- out part of the domestic workload. tially (Hanson and Pratt, 1992). Spatial According to Hanson and Pratt(1990; restrictions inuence the arrivalrate of suit- Prattand Hanson, 1991) neo-classical theory able job opportunities. The quantity and pays insufŽcient attention to the part played quality of jobs within one’s job search area by constraints in the explanation of female depend on both its location and its size (see labour participation. Although femalelabour also Simpson, 1992). For most people, the participation has risen spectacularly in the location of their job search areais Žxed in past fewdecades, many households arestill the space around their current residence. traditional in the sense that women undertake During their lifetime,people build up loca- most of the household and childcare respon- tion-speciŽc capital attheir current residence LOCAL UNDEREMPLOYMENT 1737

(DaVanzo, 1981)—for example, contacts cause past attempts to Žnd ajob werefruit- with familyand friends upon which they rely less. This effectmight be exacerbated ifthe forsocial support. Aresidential move may person stems fromthe migrant population. engender considerable costs, because of the Discouragement maybe intensiŽed when loss of location-speciŽc capital (Hey and other menwith the samecharacteristics are McKenna, 1979; see also Sjaastad, 1962). In also seen to be unemployed. addition, in households whereboth partners Discouragement is most obvious when a areengaged in paid work, aresidential move person states that he or she wants to work, maylead to job loss and thereby to loss of but does not employ any job search activi- income forone of them (Mincer,1978). Asa ties. However, discouragement might also consequence, most people only search for occur in the decision to participate in the jobs inthe vicinity that would not necessitate labour market.When people state that they aresidential move. The size of the job search do not want to work, the underlying reason areais thereforedetermined formost people can still be discouragement. Consider, for by their commuting tolerance—the timethey example, awoman with achild who is look- arewilling to spend on commuting. ing fora part-timejob. Ifshe cannot Žnd a Apart fromthe coupling constraints de- suitable job close to her home, she may scribed above, authority constraints can also decide not to enter the labour marketand to impose restrictions on job search (seeHa ¨ger- become afull-timehousewife instead. This strand, 1970). For the migrant population, phenomenon can be understood with the so- racialdiscrimination in the labour market cial-psychological theory of cognitive disso- mayseverely hamper access to labour oppor- nance (Festinger, 1957; fora geographical tunities. As aresult, people become more application of the theory, see Adams, 1973). dependent on ethnic networks that provide The woman in our example has committed morelocalised formsof employment. We herself to being active in the labour market. thereforeexpect that migrants and their off- When facedwith information that is discor- spring have lower chances of Žnding em- dant with that commitment(she does not ployment. succeed in getting the job she wants because Spatial restrictions may lead people to be- of the high local level of underemployment), comediscouraged in their search forjobs. she can reduce the dissonance by changing According to the discouraged worker hypoth- her commitment.Becoming afull-time esis, people with asmallchance of Žnding a housewife leads to agreatercognitive con- suitable job may become discouraged in their sistency. job search and leave or failto enter the Research shows that men and women dif- labour forcebecause the probability of ferin their commuting tolerance, so their job Žnding asuitable job aftera reasonable pe- search areasdiffer in size: men will tolerate riod of timeis too low (Fisher and Nijkamp, longer commuting timesthan women (Mad- 1987). Inother words: if, given the expected den, 1981; Gordon et al.,1989; Johnston- returns of search, the costs of job search are Anumonwo, 1992). Women with children too high people maygive up searching. Poor have been shown to be particularly averse to chances in the labour marketmay result from long commuting times(Rouwendal, 1999). ahigh level of underemployment in one’s job Compared with men, women aremore likely search area,which would indicate alocal to have to cope with severe day-to-day space mismatchbetween demand and supply and timeconstraints dictated by their dom- (Simpson, 1992). Poor labour market estic workload (Hanson and Pratt, 1991). We chances may also result fromindividual thereforeexpect ahigh local level of under- characteristics, either realor ascribed. For employment to discourage women in particu- example, a52-year-old manwith alow level lar.We further hypothesise that women in of education and little work experience may regions with ahigh local level of underem- become discouraged in his job search, be- ployment, state that they do not want to work 1738 MAARTENVAN HAM ET AL. moreoften than women in morefavourable work and will thereforenot search atall. In labour markets. statistical analysis, this leads to selection The rationale of discouragement can be bias. People who decide not to work select summarised in three statements. First, dis- themselves out of the population atrisk of couragement can arisefrom two sources: a job search. lack of individual qualiŽcations or ascribed To deal with these effects,we decided to negative characteristics atthe microlevel, split the analyses into three steps (Figure 1). and alack of job offersat the local or The Žrst is ananalysis of participation in the regional level. We expect an extra effectof labour marketamong the potential labour discouragement among the migrant popu- force.In this analysis, weexamined the ex- lation due to their extra-poor chances in the tent towhich the local level of underemploy- labour marketand their residential location ment inuences participation. Should it be in areaswith ahigh level of underemploy- inuenced, wewould have an indication of ment. Secondly, discouragement can enter discouragement in the participation decision the job search process attwo different stages: (that is, in wanting ajob apart fromdeciding the stage of deciding toenter the labour force to search). The dependent variable indicates (avoid underemployment by choosing not to whether (1)or not (0)a respondent is in the work), and the stage of deciding to engage active labour force.Respondents in the active actively in job search (become resigned to labour forceeither have ajob of morethan underemployment and stop searching). 12 hours aweek (the employed labour force), Thirdly, the choice of strategy not to enter or state they would like to have such ajob the labour forceor to acquiesce in underem- (the unemployed labour force). ployment can be expected to be gender- In the second analysis, the probability of related. Ifthe chances of employment are being in the unemployed labour forcewas low, women choose moreoften than mennot estimated forthose in the active labour force. to enter the labour force.To some extent, this The dependent variable indicates whether (1) option is triggered by the earning capacity of or not (0)a respondent is unemployed. The the partner. Ifthis werethe only factor, one function of this analysis was to produce a might expect that people whose partners had variable predicting the probability of being high earning capacity would participate less, unemployed fromthe independent variables, irrespective of gender. However, since it is including ethnic origin, in the model. This less socially acceptable formen not to work, variable was used in the search analyses to gender differentials arebound to occur. test whether having apoor chance of Žnding ajob leads to discouragement in searching for one. 3. Data and Methodology The third analysis is the analysis of job search. The dependent variable indicates 3.1 Method whether (1)or not (0)the respondents had In amethodological sense, the second state- searched forwork in the four weeks preced- ment above—that discouragement can enter ing the interview among those in the em- the job search process attwo stages—is far- ployed and unemployed labour force—those reaching. Ifindeed some categories of people who areeither working or state that they refrainfrom entering the labour forcealto- would like to work. In this analysis, we gether as aresult of discouragement, the excluded those people who did not want to outcomes of an analysis of whether people work atall (and so by deŽnition werenot search or not will be biased. The substantive engaged in job search). Sothis analysis is of argument is that the category of people not in discouragement in searching among those employment consists of two sub-groups: who have decided to participate on the labour those who areunemployed and will therefore market.We wish to include the job charac- search hard; those who have decided not to teristics of the employed labour population, LOCAL UNDEREMPLOYMENT 1739

Figure 1. The three analyses. so the analyses foron-the-job and off-the-job (self)selection exists. Ifthe coefŽcient is search have been separated. positive, then it is clearthat people with a In all three analyses, the dependent vari- smallpredicted probability of participating able is binary. We have thereforeused logis- have ahigh chance of being unemployed. In tic regression models. other words, acategory of people might have Ifdiscouragement enters the decision to been indicated which has chosen not to par- participate, then the active labour forcebe- ticipate, because their chances of unemploy- comes aselective category. Those with alow ment arehigh: they have been discouraged. chance of employment, as aresult of per- The predicted values of the second model sonal characteristics or alack of job offers, represent the chances of unemployment on will be underrepresented. To correctfor this the basis of the personal characteristics in- selectivity, wehave used Heckman’s two- cluded in the model. These values enter the step procedure (Heckman, 1979), by includ- analyses of the search in step three in their ing acorrection factorLambda-1 in the transformed formLambda-2. Lambda-2 does analysis of unemployment. Inits transformed not just serve as acorrection factor;it also form,Lambda-1 represents the predicted val- measures an individual’s chances in the ues of participation fromthe Žrst model labour market.Lambda-2 can also range and ranges from0 to inŽnity. The higher from0 to inŽnity; the higher the predicted the predicted probability of participation, the probability of being unemployed in model 2, lower is Lambda-1. Two conclusions can be the lower lambda-2 will be. In the search inferredfrom the coefŽcient of Lambda-1 in analyses, weexpect respondents with poor the unemployment model. Ifthe coefŽcient chances in the labour marketto be discour- is signiŽcant, then it is evident that aged in job search. The coefŽcient for 1740 MAARTENVAN HAM ET AL.

Lambda-2 is thereforeexpected to be posi- (vbo, mavo); upper-level secondary edu- tive: respondents with alow predicted proba- cation (mbo, havo, vwo); higher vocational bility of being unemployed areexpected to education (hbo); and university. Age is in 4 be morelikely to search than respondents categories: younger than 25; 25 –34 years; with ahigh probability of being unemployed. 35– 44 years; and 45 – 54 years. Adummy has An important condition forthe application been used which indicates whether (1)or not of the two-stage Heckman procedure is that (0)there is achild younger than 5years old the model issufŽciently identiŽed in order to present. avoid multicolinearity and unstable par- Four variables measurethe characteristics ameterestimates. The Žrst, second and third of the partner. Adummy indicates whether analyses thereforehave slightly differentsets (1)or not (0)the respondent has apartner. of independent variables. The ethnicity vari- Another dummy indicates whether (1)or not able has been included in the second step, the (0)the partner works. For the respondents analysis of unemployment, as its effectwas without apartner, the average of the respon- most marked in this step. dents with apartner is substituted forthis dummy. Because the model contains avari- able indicating whether apartner is present, 3.2 Data and Variables this substitution of the means leads to un- The data used in this paper werederived biased coefŽcients of the ‘partner works’ fromDutch Labour Force Surveys conducted dummy forthose with aworking partner in 1994, 1995, 1996 and 1997 by Statistics (compareCohen and Cohen, 1975, ch. 7). Netherlands. The Labour Force Survey is The educational level of the partner is mea- representative of the Netherlands population sured in Žve categories. Substitution of the aged 15 and above and not living in an means is used todeal with respondents with- institution. The data-set includes detailed in- out apartner. The job level of the partner is formation concerning individual and house- allotted to one of the Žve levels of the Stan- hold characteristics such as level of dard Job ClassiŽcation (SBC – 1992) of education, number of children, job character- Statistics Netherlands: elementary; low; mid- istics, partner characteristics and detailed in- dle: high; academic.The substitution of formation on the workplace and location of means method has again been used to deal residence. Further, the data-set includes a with respondents without apartner, or with- direct question regarding job search. Respon- out aworking partner. dents wereasked, “Have you undertaken any Local underemployment was calculated as activity to Žnd a(nother) job in the last four apercentage of the local potential labour weeks?”. Merelylooking atjob advertise- force,using the 1994 – 97 Labour Force Sur- ments in the newspaper could count as search veys. Being underemployed is deŽned as activity. having no job atall, having ajob of less than The analyses arerestricted to respondents 12 hours aweek or having ajob whose level aged between 15 and 54 years excluding is too low with respect to the educational students, the armedservices, the self-em- level of the respondent. With the GISexten- ployed and the disabled. The potential labour sion FLOWMAP (de Jong and Floor, 1993; forcein the data-set amounts to 143 930 men van Ham et al.,2001), wehave calculated a and 156 196 women. The unemployed labour measureof underemployment on the local forceconsists of 16 366 menand 30 490 labour marketfor every respondent in the women, while the employed labour force data-set. The starting-point was avery low consists of 125 202 menand 79 094 women. spatial level; the almost 4000 4-digit post- In the analysis of participation, 8indepen- code areas.This is the Žnest measurementof dent variables have been included. Level of residential locations in our data-set. For ev- education is in 5categories: primaryedu- ery postcode, wecalculated the percentage of cation; lower-level secondary education underemployment in an areathat could be LOCAL UNDEREMPLOYMENT 1741 reached within 30 minutes by car.Since in old; and youngest child between 12 and 17 the Netherlands 80 per cent of the working years old. Hours worked per week arein four population travels less than 30 minutes per categories: 12 – 20 hours; 21– 35 hours; 36–40 single journey to work, this was thought to hours; and morethan 40 hours aweek. Com- be areasonable approach to the local labour muting timeis measured in Žve categories: markets. The local percentage of underem- 0– 30 minutes; 31 –45 minutes; 46 – 60 min- ployment ranges from41 to 54 per cent of utes; morethan 60 minutes; unknown. Regu- the potential labour force.Two areasstand larity of working timeshas been reduced to a out in having above-average levels of under- two-category variable, indicating whether (1) employment: the inner-city neighbourhoods or not (0)the respondents have irregular of the two largest cities (Amsterdamand working times. Job security has also been Rotterdam) and the moreperipheral rural reduced to atwo-category variable, indicat- areas.Below-average underemployment is ing whether (1)or not (0)respondents have a found in the suburban areasin between the permanent employment contract. cities. In the analysis of unemployment, seven 4. Results independent variables have been included. Level of education and age aremeasured in As expected, menhave ahigher probability the sameway as in the Žrst analysis. Type of of participating in the labour marketthan household is categorised as: single; couple women. Fromour data, weŽ nd that 98 per with unemployed partner; couple with work- cent of the malepotential labour forceeither ing partner; and others. Adummy indicates have ajob or would like ajob of at least 12 whether (1)or not (0)a respondent is a hours aweek. In contrast, only 70 per cent of migrant, or adescendant froma migrant. A the femalepotential labour forceis in the dummy indicates whether (1)or not (0)the active labour force.As expected, menhave a respondent leftschool in the year before the higher probability of being engaged in job interview. The year of interview is indicated search than women: fromthe unemployed in four categories: 1994; 1995; 1996; 1997. labour force,73 per cent of the malerespon- Lambda-1 is acontinuous variable ranging dents compared with only 52 per cent of the from0 to inŽnity. femalerespondents areengaged in job In the off-the-job search analyses, six in- search. For on-the-job search there areno dependent variables areincluded. Level of gender differences; 10 per cent of those in education, age and local underemployment the employed labour forceare engaged in job aremeasured in the sameway as before. search. Working experience is measured in atwo- category variable, indicating whether (1)or 4.1 Analysis of Participation not (0)respondents have ever had ajob of morethan 12 hours aweek. The type of Table 1gives the results of the analysis of household is categorised as: single unem- participation in the active labour force.For ployed; unemployed with working partner; both menand women, the probability of both partners unemployed; others. The con- being in the active labour forceincreases trol factorLambda-2 is acontinuous variable with level of education and decreases with ranging from0 to inŽnity. age. Tests showed only aslight effectof Inthe on-the-job search analyses, the same ethnicity on participation. The variable isnot variables asin the off-the-job search analyses included in the model to avoid multi- areincluded, together with the presence of collinearity in the second step. children and some additional job characteris- Having achild under the age of 5has a tics. The presence of children is categorised signiŽcant negative effecton the probability as: no children; youngest child under 6years of being in the active labour force.This was old; youngest child between 6and 12 years asexpected forwomen, but the factthat there 1742 MAARTENVAN HAM ET AL.

Table 1. Logistic regression of being in the activelabour force, by gender

Men Women

Standard Standard B error Exp(B) B error Exp(B)

Education Primary 0 1 0 1 Lower secondary 0.800*** 0.056 2.226 0.433*** 0.019 1.542 Upper secondary 1.274*** 0.056 3.573 1.129*** 0.019 3.090 High vocational 1.585*** 0.085 4.878 2.005*** 0.028 7.428 University 1.945*** 0.126 6.995 2.829*** 0.059 16.917 Age (years) , 25 0 1 0 1 25– 34 2 0.246*** 0.081 0.782 2 0.718*** 0.031 0.488 35– 44 2 0.809*** 0.080 0.445 2 1.384*** 0.030 0.251 45– 54 2 1.593*** 0.080 0.203 2 2.267*** 0.031 0.104 Child under 5years No 0 1 0 1 Yes 2 0.277*** 0.066 0.758 2 1.435*** 0.017 0.238 Partner No 0 1 0 1 Yes 1.225*** 0.051 3.404 2 0.840*** 0.019 0.432 Partner works No 0 1 0 1 Yes 2 0.240*** 0.061 0.787 0.047** 0.019 1.048 Educational levelof partner Primary 0 1 0 1 Lower secondary 0.634*** 0.074 1.885 0.118*** 0.023 1.125 Upper secondary 0.670*** 0.080 1.953 0.302*** 0.023 1.352 High vocational 0.500*** 0.127 1.643 0.382*** 0.031 1.466 University 0.019 0.211 1.020 0.282*** 0.043 1.326 Job levelof partner Elementary 0 1 0 1 Low 0.058 0.133 1.060 2 0.224*** 0.035 0.799 Middle 2 0.371*** 0.130 0.690 2 0.230*** 0.034 0.795 High 2 0.577*** 0.161 0.562 2 0.287*** 0.039 0.751 Academic 2 0.429* 0.259 0.652 2 0.486*** 0.048 0.615 Local underemployment 2 0.008 0.010 0.992 2 0.027*** 0.003 0.973 (percentage) Constant 3.413*** 0.472 3.674*** 0.143

Initial-2 log likelihood 24 590 190 409 Model-2 log likelihood 22 561 156 406 Improvement 2 029, df 5 19, p 5 0.00 34 003, df 5 19, p 5 0.00

*** indicates signiŽcant atthe 1per cent level; **indicates signiŽcant atthe 5per cent level; *indicates signiŽcant atthe 10 per cent level. is also an effectfor men was not. The effect ing apartner, whether the partner works, the is much stronger forwomen than formen. partner’s educational level, and job level. For Four variables wereentered into the model women, having apartner has anegative ef- to indicate apartner’s earning capacity: hav- fectand this is exacerbated ifthe partner’s LOCAL UNDEREMPLOYMENT 1743 job level is high. The educational level of the which measures an individual’s chances in partner yields au-shaped effect.People the labour market.Lambda-2 is used as an whose partner has amedium level of edu- independent variable in the search analyses. cation have ahigher probability of participat- The likelihood of being unemployed is ing than those with apartner whose level of increased by having alow level of education, education is either high or low. This indi- being aschool leaver, an immigrant, the cates that the effectof being atwo-wage- descendant of an immigrantor by living earnercouple is most prominent among alone. The ethnicity variable in particular couples with average earning capacity. For shows astriking effectthat is moresubstan- men, having apartner has apositive effecton tial than the educational variable. The poor participation, which is offset to some extent chances in the labour marketof the migrant ifthe partner works and in particular ifthe population cannot be attributed to an overall level of the partner’s job is high. skill-mismatch. To test the hypothesis on discouragement People interviewed in morerecent years in the participation decision, the local per- have alower probability of being unem- centage of underemployment is included as ployed. This Žnding can be explained by the an independent variable. For women, the re- factthat, fromthe mid 1990s, the economy sults areas expected: the local level of un- in the Netherlands has shown an upward deremployment has anegative effecton the tendency. For both men and women, participation decision of women. Women liv- Lambda-1 has asigniŽcant effecton the ing in areaswith ahigh local level of under- probability of being unemployed: this means employment state that they do not want to that (self)selection exists. The factthat the work moreoften than women in more parameterfor Lambda-1 is positive indicates favourable labour markets. For menthere is that people who stated that they wanted to no effect.The results show that women are work forat least 12 hours aweek, but who indeed moreeasily discouraged than menby had characteristics similarto those who have poor local labour marketconditions. chosen not to participate, have ahigh proba- The analysis of participation results in a bility of being unemployed. This means that correction factorknown as Lambda-1 which there is acategory of people who have used is used as an independent variable in the the decision not to participate as ameans of second model to control forselection effects. avoiding unemployment: they have been dis- The results fromthe analyses of participation couraged. show the plausibility of the effectof discour- agement on the decision to refrainfrom 4.3 Off-the-job Search working. Personal characteristics that indi- catepoor chances on the labour market(low Table 3presents the off-the-job search re- education, high age) and alack of job offers sults by gender. The researchpopulation con- in the local economy both have anegative sists of unemployed respondents who stated impacton the decision to participate. Itis that they would like to have ajob forat least shown, by entering the Lambda-1 score as an 12 hours aweek. independent variable in the unemployment model, that non-participation is away of Men.As expected, level of education has a avoiding unemployment (seenext section). positive effectfor men on the probability of being engaged in off-the-job search. Work experience also has apositive effecton job 4.2 Analysis of Unemployment search. Both Žndings conŽrm the idea that Table 2presents the results of the analysis of unemployed people have ahigher probability unemployment among those in the active of being engaged in job search as the level of labour force.The main function of this se- human capital rises. With rising age, menare cond analysis is to construct Lambda-2, less likely to be engaged in job search. This 1744 MAARTENVAN HAM ET AL.

Table 2. Logistic regression of being unemployed, by gender

Men Women

Standard Standard B error Exp(B) B error Exp(B)

Education Primary 0 1 0 1 Lower secondary 2 0.691*** 0.034 0.501 2 0.183*** 0.026 0.832 Upper secondary 2 0.980*** 0.039 0.375 2 0.439*** 0.028 0.645 High vocational 2 1.201*** 0.048 0.301 2 0.627*** 0.036 0.535 University 2 0.989*** 0.054 0.372 2 0.630*** 0.049 0.533 School leaver No 0 1 0 1 Yes 0.960*** 0.041 2.611 0.864*** 0.040 2.373 Age (years) , 25 0 1 0 1 25– 34 0.128*** 0.034 1.136 0.116*** 0.030 1.123 35– 44 0.097** 0.040 1.102 0.397*** 0.031 1.487 45– 54 0.044 0.050 1.045 2 0.068* 0.036 0.935 Immigrant or descendant No 0 1 0 1 Yes 1.425*** 0.022 4.157 0.664*** 0.022 1.943 Household situation Single 0 1 0 1 Couple, partner unemployed 2 1.214*** 0.033 0.297 2 0.552*** 0.030 0.576 Couple, partner employed 2 1.562*** 0.033 0.210 2 0.777*** 0.021 0.460 Other 2 0.705*** 0.032 0.494 2 0.864*** 0.036 0.422 Year of interview 1994 0 1 0 1 1995 2 0.109*** 0.024 0.896 2 0.067*** 0.020 0.935 1996 2 0.219*** 0.025 0.804 2 0.133*** 0.020 0.876 1997 2 0.423*** 0.026 0.655 2 0.284*** 0.021 0.753 Lambda-1 1.403*** 0.407 4.066 1.672*** 0.044 5.323 Constant 2 0.480*** 0.060 2 0.855*** 0.040

Initial-2 log likelihood 101 255 129 589 Model-2 log likelihood 88 688 118 849 Improvement 12 567, df 5 16, p 5 0.00 10 739, df 5 16, p 5 0.00

*** indicates signiŽcant atthe 1per cent level; **indicates signiŽcant atthe 5per cent level; *indicates signiŽcant atthe 10 per cent level. is also as weexpected. The effectof house- effectof local underemployment on job hold situation shows that unemployed men search. To test whether poor labour market with apartner have ahigher probability of expectations resulting fromindividual being engaged in job search than single men. characteristics have adiscouraging effecton Itwas expected that people living inareas job search, Lambda-2 has been included with ahigh local level of underemployment in the search analysis. The higher the pre- would have the lowest probability of being dicted probability of being unemployed, the engaged in job search. However, the results lower was Lambda-2. As expected, the show that formen there is no signiŽcant coefŽcient of Lambda-2 was positive and LOCAL UNDEREMPLOYMENT 1745

Table 3. Logistic regression of off-the-job search,by gender

Men Women

Standard Standard B error Exp(B ) B error Exp(B )

Education Primary 0 1 0 1 Lower secondary 0.278*** 0.054 1.320 0.112*** 0.038 1.118 Upper secondary 0.553*** 0.059 1.740 0.211*** 0.046 1.235 High vocational 0.704*** 0.083 2.022 0.346*** 0.064 1.413 University 1.102*** 0.098 3.011 0.805*** 0.088 2.236 Working experience No 0 1 0 1 Yes 0.397*** 0.049 1.487 0.199*** 0.036 1.220 Age (years) , 25 0 1 0 1 25– 34 2 0.309*** 0.066 0.734 2 0.431*** 0.052 0.650 35– 44 2 0.636*** 0.073 0.530 2 0.538*** 0.055 0.584 45– 54 2 0.810*** 0.077 0.445 2 1.790*** 0.056 0.454 Household situation Single 0 1 0 1 Couple, partner employed 0.130** 0.077 1.139 2 0.542*** 0.032 0.582 Couple, partner unemployed 0.960** 0.066 1.101 2 0.551*** 0.041 0.576 Other 0.161*** 0.067 1.175 0.394*** 0.075 1.483 Local underemployment 2 0.012 0.008 0.988 2 0.005 0.006 1.005 (percentage)

Lambda-2 0.265*** 0.067 1.304 0.471*** 0.070 1.602 Constant 0.882*** 0.394 2 0.097 0.265

Initial-2 log likelihood 19 021 42 206 Model-2 log likelihood 18 413 40 191 Improvement 607, df 5 13, p 5 0.00 2 015, df 5 13, p 5 0.00

*** indicates signiŽcant atthe 1per cent level; **indicates signiŽcant atthe 5per cent level; *indicates signiŽcant atthe 10 per cent level. signiŽcant formen. This means that men having apartner makes it less necessary to with ahigh probability of being unemployed search fora job. search less than menwith alow probability The results show that, just as formen, of being unemployed. This Žnding indicates there is no signiŽcant effectof local under- discouragement forunemployed menwith employment on job search forwomen. Ap- poor chances in the labour market. parently, local labour marketconditions do not lead to discouragement in job search by Women.For women, the effectsof level of the unemployed. Once people decide they education, work experience and age wereall want to participate in the labour market,they found to be in the expected direction and do not allow themselves to become discour- correspond with the effectsfound formen. aged by poor local labour marketconditions. Women with apartner have alower probabil- For women, the positive effectof Lambda-2 ity of being engaged in job search than single is also in line with the expected effect. women. Some women apparently Žnd that Women with ahigh probability of being 1746 MAARTENVAN HAM ET AL. unemployed search less than women with a commuting time.The category ‘commuting low probability of being unemployed. Unem- timeunknown’ consists mainly of respon- ployed women with poor chances arelikely dents with short-termcontracts. In contrast to be discouraged in job search. with what was expected, having irregular working hours was not found to have aposi- tive effecton job search. But, as expected, 4.4 On-the-job Search men search morewhen they have little job The logistic regression results forthe on-the- security. job search model arepresented in Table 4. Again, the local percentage of underem- The researchpopulation consists of em- ployment and Lambda-2 have been included ployed respondents who work forat least 12 to test whether poor labour marketexpecta- hours aweek. tions have adiscouraging effecton the prob- ability of being engaged in job search. As Men.Men with ahigher level of education expected on the basis of the discouraged weremore likely to be engaged in job search. worker hypothesis, formen the local percent- This is in accordance with the expectations age of underemployment has anegative ef- based on human capital theory. The probabil- fecton on-the-job search. For men, ity of being engaged in job search decreased Lambda-2 does not have asigniŽcant effect with age. Again, this is as expected because on job search; wedid not Žnd evidence for as age increases the pay-off period decreases an effectof poor labour marketexpectations forjob search and job change costs. Men resulting fromindividual characteristics on with achild between 0and 5years old search on-the-job search by men. the most and menwith children in the 12 – 17 age-group search the least. Apossible expla- Women.For women, the effectsof level of nation might be that menwith young chil- education and age arein the expected direc- dren feelmore responsible forthe family tion and correspond with the effectsfound income and so search forbetter-paid jobs. formen. For women, the effectof the pres- The effectmight also be aneffectof the age ence of children was as expected. Employed of the men themselves. As the age of the women without children search the most. children rises, so does the age of the parents When they have children, the probability of and as people get older they search less being engaged in job search increases with frequently. The effectof household situations the increasing age of the youngest child. shows that menwith apartner search less Being single has apositive effecton job frequently than single men. search. For men, the number of hours worked per As expected, forwomen the effectof week had anegative inuence on job search. hours worked per week ismuch smallerthan This is as expected; most men want afull- is the case formen. Women moreoften timejob. Aftercontrolling forlevel of edu- prefersmall (part-time) jobs, because they cation, every higher job level reached led often have to combine apaid job with dom- people to be less likely to search. This is estic work. The effectsof job level and com- according to what would be expected on the muting timeare as expected and correspond basis of the human capital theory. People with the effectsfound formen. However, the whose job level is not known search the effectof commuting timeis somewhat most. Many respondents in this category stronger on women than on men. This have not been asked fortheir job level, be- conŽrms the idea that women aremore sensi- cause they had short-termcontracts; since a tive to spatial restrictions than men. Surpris- short-termcontract offerslittle job security, ingly, women with irregularworking hours people with an unknown job level areoften search less frequently than women with reg- engaged in job search. As expected, job ular working hours. This is contrary to what search intensity increases with increasing was expected, but maybe explained by the LOCAL UNDEREMPLOYMENT 1747

Table 4. Logistic regression of on-the-job search,by gender Men Women

Standard Standard B error Exp(B) B error Exp(B)

Education Primary 0 1 0 1 Lower secondary 0.234*** 0.051 1.263 0.095 0.065 1.100 Upper secondary 0.680*** 0.053 1.974 0.424*** 0.072 1.528 High vocational 1.097*** 0.062 2.996 0.820*** 0.087 2.270 University 1.215*** 0.069 3.370 1.299*** 0.100 3.664 Age (years) , 25 0 1 0 1 25– 34 2 0.051 0.038 0.951 2 0.273*** 0.038 0.761 35– 44 2 0.340*** 0.042 0.712 2 0.462*** 0.051 0.630 45– 54 2 1.053*** 0.048 0.349 2 1.149*** 0.056 0.317 Children under 18 years old No children 0 1 0 1 Youngest under 6years 0.101*** 0.029 1.107 2 0.332*** 0.045 0.717 Youngest 6– 12 years 0.043 0.036 1.044 2 0.132*** 0.044 0.877 Youngest 13– 17 years 2 0.138*** 0.042 0.872 2 0.041 0.046 0.960 Household situation Single 0 1 0 1 Couple, partner unemployed 2 0.342*** 0.048 0.710 2 0.441*** 0.055 0.643 Couple, partner employed 2 0.107** 0.050 0.899 2 0.622*** 0.040 0.537 Other 2 0.434*** 0.047 0.648 2 0.511*** 0.058 0.600 Hours per week 13– 20 0 1 0 1 21– 35 2 0.676*** 0.061 0.509 2 0.036 0.033 0.965 36– 40 2 0.869*** 0.054 0.419 2 0.291*** 0.036 0.748 . 40 2 0.824*** 0.066 0.439 2 0.072 0.083 0.931 Job level Elementary 0 1 0 1 Low 2 0.341*** 0.041 0.711 2 0.266*** 0.046 0.767 Middle 2 0.435*** 0.042 0.647 2 0.586*** 0.049 0.557 High 2 0.596*** 0.051 0.551 2 0.775*** 0.061 0.461 Academic 2 0.701*** 0.064 0.496 2 0.879*** 0.084 0.415 Unknown 0.077 0.057 1.080 0.111 0.067 1.118 Commuting time (minutes) 0– 30 0 1 0 1 31– 45 0.089** 0.035 1.093 0.096** 0.042 1.101 46– 60 0.133*** 0.031 1.142 0.152*** 0.038 1.165 . 61 0.305*** 0.030 1.357 0.468*** 0.037 1.597 Unknown 0.065** 0.029 1.068 0.219*** 0.039 1.244 Irregular hours No 0 1 0 1 Yes 0.023 0.021 1.023 2 0.098*** 0.025 0.907 Permanent contract No 0 1 0 1 Yes 2 1.215*** 0.031 0.297 2 0.785*** 0.032 0.456 Local underemployment 2 0.017*** 0.005 0.983 2 0.002 0.006 0.998 (percentage) Lambda-2 2 0.032 0.053 0.968 0.331*** 0.091 1.392 Constant 2 0.572** 0.238 2 1.762*** 0.296 Initial-2 log likelihood 77 779 51 919 Model-2 log likelihood 71 624 48 213 Improvement 6 155, df 5 29, p 5 0.00 3 705, df 5 29, p 5 0.00

*** indicates signiŽcant at the 1per cent level; **indicates signiŽcant at the 5per cent level; *indicates signiŽcant at the 10 per cent level. 1748 MAARTENVAN HAM ET AL. factthat irregular working hours may be morethan 12 hours per week. The category moreconvenient when domestic work and ‘out of employment’could thereforebe split paid employment have to be combined. As into the group that did not participate and the formen, having apermanent employment unemployed. Both the unemployed and the contract has anegative inuence on job employed wereasked whether they had been search. active in job searching in the 4weeks pre- The local level of underemployment does ceding the interview. Threemodels were not have aneffecton job search forwomen. speciŽed: one forthe probability of partici- Apparently, poor local labour marketcondi- pating, another forthe probability of unem- tions do not discourage women in their on- ployment given participation; athird forthe the-job search. However, the coefŽcient for probability of engaging in on-the-job or off- Lambda-2 is positive and signiŽcant: the the-job search. lower awoman’s predicted probability of The results indicate the existence of a being unemployed, the higher her probability discouraged worker effectin the stage of of being engaged in job search. In other deciding to participate. For both men and words, women with poor chances in the women, personal characteristics that indicate labour marketsearch less frequently than poor chances in the labour marketwere nega- women with good chances inthe labour mar- tively related tothe decision toparticipate in ket, possibly because of discouragement. the labour force.Discouragement atthis stage appeared to be gender-related. Not only werethe effectsof poor chances much 5. Summary and Discussion stronger forwomen; they werealso put off In this contribution, wehave elaborated the fromparticipation in places with ahigh level concept of the discouraged worker effectand of local underemployment. For men, the ef- reported our empiricaltesting. The discour- fectof local underemployment level was in- aged worker effecthas been deŽned as the signiŽcant. decision to refrainfrom job search as aresult In the analysis of the chances of unem- of poor chances in the labour market.Two ployment, acorrection factorwas entered to sources of discouragement wereidentiŽ ed: a account forthe selectivity of the group par- lack of individual qualiŽcations or ascribed ticipating in the labour force.The substantive characteristics that makea worker less com- interpretation of this correction factor petitive in the job market;and alack of showed that people who refrainfrom partici- suitable job offersresulting fromthe level of pating would have had ahigh chance of underemployment in the local economy. In being unemployed ifthey had put their elaborating the concept, wehypothesised that labour on offer.Not participating is astrat- discouragement can enter the job search pro- egy foravoiding unemployment chosen by cess attwo stages. The Žrst stage is the women in particular. decision to participate in the labour force. The results of the discouragement effectin We have tested the hypothesis that people in job search among those in the active labour general and women in particular who have forceare slightly less convincing. For the poor chances in the (regional) labour market unemployed, itcould be shown that personal moreoften refrainfrom participating. The characteristics (low education, older age, second stage is the decision to become ac- lack of work experience) werenegatively tively engaged in job search once one is related to job search. Inclusion of the correc- active in the labour market. tion factorthat indicates the overall chance In the empiricaltests, wehave used direct of being unemployed showed that poor measures of participation and job search, us- chances in the labour markethave astrong ing data fromthe Labour Force Surveys impacton the intensity of job search. This 1994– 97. In this survey, people wereasked conclusion is particularly relevant forthe whether or not they werewilling to work for occupational achievement of migrants and LOCAL UNDEREMPLOYMENT 1749 their descendants. Astheir chances of unem- els of local underemployment, regional ployment aremuch higher than those of the differences ineconomic performanceand un- indigenous population with the same deremployment arelow in this country. The qualiŽcations, the intensity of their job Žndings might be radically differentin other, search islower, further hampering social mo- largercountries. The reason why personal bility of this population. No effectwas found characteristics aremore dominant might also fromthe local level of underemployment. be an effectof the rapidly decreasing levels For the employed, the overall probability of unemployment. In atight labour market, of job search is much lower. Again, personal there is aproblem of the unemployed rather characteristics (including the level of the pre- than of unemployment. The category of the sent job, job security and the number of unemployed is becoming increasingly selec- hours worked) account forthe majorpart of tive. Only those people with really poor the differentiation in job search. Yetamong chances in the labour marketremain unem- mena high local level of underemployment ployed in agrowing economy. This selectiv- also led to reduced search activity. ity in unemployment goes beyond alack of In general terms,we have found discour- educational achievement. Also, aftercon- agement effectsat both stages of the search trolling forformal education, the chances of process. The dominant source of discourage- the migrant population turned out to be ex- ment is an individual’s lack of qualiŽcations ceptionally low. This indicates that ascribed or other personal and ascribed characteristics characteristics mayplay arole, both through that reduce the chances in the labour market. poor chances and through discouragement in We found mixed evidence of adiscourage- reaching occupational achievement. Italso ment effectarising froma lack of suitable indicates that the high level of local under- job offersin the local economy. The decision employment in the largertowns is morethan by women to participate and the decision of just a‘skills-mismatch’. on-the-job search by men arenegatively Obviously, the poor results on the discour- inuenced by ahigh local level of underem- aging effectof the local economy could also ployment. Apparently, women outside the arisefrom the limitations in our analyses. labour forceand working menhave some- First, although wehad the unique oppor- thing in common that makes them more tunity of using adirect question on job likely than other categories to be discouraged search, this is no guarantee that wehad a by local labour marketconditions. Itmay be sharp measurementof discouragement. Not that both groups have an alternative to search all those who stated that they had not and can ‘afford’to be discouraged. For searched in the four weeks preceding the women, being afull-timehousewife is so- interview might have been discouraged. cially accepted, especially when (young) Some might have just returned froma hol- children arepresent. Men in the employed iday, or have been ill. Secondly, the way we labour forcealso have areasonable alterna- measured local labour marketconditions tive to search. They already have ajob so might not be the best approach. The optimal they can stay put until the labour market solution would be to construct avariable to becomes morefavourable. indicate the number of vacancies in acertain arearelative to the number of underem- ployed people in that area.Unfortunately, 5.1 Implications and Limitations data on vacancies arehard toobtain, seldom The Žnding that local levels of underemploy- available ata low spatial level and of ques- ment only contribute incidentally to the dis- tionable reliability. Athird limitation of our couraged worker effectcould be aparticular analyses might be the way wemeasured characteristicof the Netherlands. Even discouragement by individual characteristics. though both peripheral ruralareas and inner- With our data, wecan only show that there is city neighbourhoods have above-average lev- astatistical relationship between job search 1750 MAARTENVAN HAM ET AL. and ahigh predicted probability of being control and constraints, Geography, 76, pp. 17– unemployed. We have no idea of the extent 26. BURDETT,K.(1978) Atheory of employee job to which people arereally awareof their own searchand quit rates, American Economic Re- poor chances in the labour market. view, 68, pp. 212– 220. Given these possible shortcomings, future COHEN, J. and COHEN, P. (1975) Applied Multiple researchcould improve on the present effort Regression/Correlation Analysis for the Be- by using data that overcome some of these havioural Sciences .NewYork: John Wiley & Sons. limitations. The use of data collected forthe COOKE, T. J. and ROSS,S.L.(1999) Sample purpose of researchon job search might give selection bias in models of commuting time, better insights. However, such adata-set Urban Studies ,36, pp. 1597 –1611. would have to be large enough to be able to DAVANZO,J.(1981) Microeconomic approaches incorporate variables on spatial differences in to studying migration decisions, in: G.F. DE JONG and R. W. GARDNER (Eds) Migration the local labour marketsituation. While Decision Making: Multidisciplinary Ap- quantitative researchhelps to gain morein- proaches to MicrolevelStudies in Developed sight into the statistical relationship between and Developing Countries , pp. 90– 129. New job search and local labour marketcondi- York: Pergamon Press. tions, qualitative methods could help us to DEVINE, T. J. and KIEFER,N.M.(1991) Empirical Labour Economics: The Search Approach. understand labour marketbehaviour in more NewYork: Oxford University Press. detail. Questions could address why people FESTINGER, L. (1957) ATheory of Cognitive Dis- do or do not search, how often they search sonance.Stanford, CA:Stanford University and wherethey search. Such qualitative re- Press. search could lead to abetter understanding of FISHER, M. M. and NIJKAMP,P. (1987) Spatial labour marketanalysis: relevanceand scope, in: the labour marketbehaviour of women, ex- M. M. FISHER and P. NIJKAMP (Eds) Regional plain some of the current confounding Labour Markets , pp. 1–36. Amsterdam:North Žndings and lead to new hypotheses. Holland. We have, however, shown that future re- GORDON, P., KUMAR, A. and RICHARDSON, H. W. search should include the stage of deciding (1989) Gender differencesin metropolitan travel behaviour, Regional Studies , 23, not to participate in the labour forceat all. pp. 499– 510. Poor chances affectthe decision to partici- HA¨ GERSTRAND,T.(1970) What about people in pate and the people atrisk of searching fora regional science?, Papers of the Regional Sci- (better)job area selective group. enceAssociation , 24, pp. 7– 21. HAM, M. VAN, HOOIMEIJER, P. and MULDER, C. H. (2001) Urban form and job access:disparate References realitiesin the Randstad, Tijdschrift voor Economische en Sociale GeograŽe , 92, pp. ADAMS,R.L.A.(1973) Uncertainty in nature, 231– 246. cognitive dissonance, and the perceptual distor- HANSON, S. and PRATT,G.(1988) Spatial dimen- tion of environmental information: weather sions of the gender division of labour in alocal forecastsand NewEngland beach trip deci- labour market, Urban Geography , 9, pp. 180– sions, Economic Geography , 49, pp. 287– 297. 202. BECKER,G.(1962) Human capital: atheoretical HANSON, S. and PRATT,G.(1990) Geographic and empiricalanalysis, Journal of Political perspectives on the occupational segregation of Economy, 70, pp. 9– 46. women, National Geographic Research , 6, BECKER, G. (1976) The Economic Approach to pp. 376– 399. Human Behavior. Chicago, IL: The University HANSON, S. and PRATT,G.(1991) Job searchand of Chicago Press. the occupational segregation of women, Annals BECKER, G. (1991) ATreatise on the Family , of the Association of American Geographers , enlarged edn. Cambridge MA:Harvard Univer- 81, pp. 229– 253. sity Press. HANSON, S. and PRATT,G.(1992) Dynamic de- BLAU,D.M.(1991) Searchfor non-wage job pendencies: ageographic investigation of local characteristics:a testof the reservation wage labor markets, Economic Geography , 68, hypothesis, Journal of Labor Economics , 9, pp. 373– 405. pp. 187–205. HANSON, S., KOMINIAK, T. and CARLIN, S. (1997) BOWLBY,S.(1990) Women, work and the family: Assessing the impactof location on women’s LOCAL UNDEREMPLOYMENT 1751

labor marketoutcomes: amethodological ex- marketanalysis, in: O.A SHENFELTER and R. planation, Geographical Analysis , 29, pp. 281– LAYARD (Eds) Handbook of Labor Economics , 297. pp. 849– 919. Amsterdam:North-Holland. HECKMAN,J.(1979) Sample selection bias asa OMMEREN, J. VAN (1996) Commuting and Relo- speciŽcation error, Econometrica , 47, pp. 153– cation of Jobs and Residences:A Search Per- 161. spective.Amsterdam:Vrije Universiteit HEY, J. D. and MCKENNA,C.J.(1979) To move or Amsterdam. not to move, Economica, 46, pp. 175– 185. ONG, P. and BLUMENBERG,E.(1998) Job access, IMMERGLUCK,D.(1998) Job proximity and the commute, and travel burden among welfare urban employment problem: do suitable nearby recipients, Urban Studies , 31, pp. 77–93. jobs improve neighbourhood employment OPHEM, H. VAN (1991) Wages, nonwage job char- rates?, Urban Studies , 35, pp. 7– 23. acteristics,and the searchbehaviour of em- JOHNSON,W.R.(1978) Atheory of job shopping, ployees, Reviewof Economics and Statistics , Quarterly Journal of Economics , 92, pp. 261– 73, pp. 145–151. 277. PARSONS,D.O.(1973) Quit ratesover time:a JOHNSTON-ANUMONWO,I.(1992) The inuence of searchand information approach, American household type on gender differencesin work Economic Review , 63, pp. 390– 401. trip distance, Professional Geographer , 44, PARSONS,D.O.(1991) The job searchbehaviour pp. 161– 169. of employed youth, Reviewof Economics and JONG, T. DE and FLOOR,H.(1993) Flowmap: een Statistics, 73, pp. 597–604. programma voor het weergeven en analyseren PRATT, G. and HANSON,S.(1991) Time,space, van interactiegegevens [Flowmap: asoftware and the occupational segregation of women: a package for displaying and analysing interac- critique of human capital theory, Geoforum, 22, tion data], Planning, methodiek en toepassing , pp. 149– 157. 44, pp. 16– 31. PRESTON, V. and MCLAFFERTY,S.(1999) Spatial JOVANOVIC,B.(1979) Job matching and the the- mismatchresearch in the 1990s: progress and ory of turnover, Journal of Political Economy , potential, Papers in Regional Science , 78, 87, pp. 972– 990. pp. 387– 402. KAIN,J.(1968) Housing segregation, negro em- ROUWENDAL,J.(1999) Spatial job searchand ployment, and metropolitan decentralization, commuting distances, Regional Scienceand Quarterly Journal of Economics , 82, pp. 175– Urban Economics , 29, pp. 491–517. 197. SIMPSON, W. (1992) Urban Structure and the KEITH, K. and MCWILLIAMS,A.(1999) The re- Labour Market:Worker Mobility, Commuting turns to mobility and job searchby gender, and Underemployment in Cities . Oxford: Industrial and Labor Relations Review , 52, Clarendon Press. pp. 460– 477. SJAASTAD,L.A.(1962) The costs and returns of KIEFER, N. M. and NEUMANN,G.R.(1989) human migration, Journal of Political Econ- Search Models and Applied Labour Economics . omy, 70, pp. 80– 93. Cambridge: Cambridge University Press. STIGLER,G.J.(1961) The economics of infor- LIPPMAN, S. A. and MCCALL,J.J.(1976) The mation, Journal of Political Economy , 69, economics of job search:a survey, Economic pp. 213– 225. Inquiry, 14, pp. 155–189. STIGLER,G.J.(1962) Information in the labor MADDEN,J.F.(1981) Why women work closerto market, Journal of Political Economy , 70, home, Urban Studies , 18, pp. 181– 194. pp. 94– 105. MINCER,J.(1978) Family migration decisions, TOPEL, R. H. and WARD,M.P.(1992) Job mo- Journal of Political Economy , 86, pp. 749–773. bility and the careersof young men , Quarterly MORTENSEN,D.T.(1986) Job searchand labor Journal of Economics ,107, pp. 439 – 479.