Fertility and Commuting Behaviour in Germany Johannes Huinink, Michael Feldhaus Abstract: Fertility behaviour is closely related to other dimensions of the individual life course, which are strongly interrelated themselves. Regarding the impact of job-related spatial mobility, empirical fi ndings show a negative correlation between having children and commuting, particularly for women. Up to now, fertility inten- tions have not been thoroughly investigated in this respect. Longitudinal studies are lacking, too. In this paper, the effects of commuting arrangements of men and women on the intention of having a child within the next two years as well as the probability of realising this intention are addressed. The assumption is, that after accounting for other important factors (employment status, level of qualifi cation, type of consensual union, number of children, residential mobility), medium- and long-distance commuting is negatively related to the fertility intention of women and its realisation. For men, effects are assumed to be nonexistent or even slightly positive. Longitudinal data from the fi rst three waves of the German “Panel Analy- sis of Intimate Relationships and Family Dynamics” (pairfam) are used to test the hypotheses. Firstly, a cross-sectional, multivariate probit-regression (with correlated errors) on the intention to have a child within two years, on being childless and on medium- and long- distance commuting is applied. The model shows no signifi cant correla- tion between commuting and the intention to have a child; it does however show a correlation between medium- and long distance commuting and the probability of women to be childless. Secondly, a longitudinal difference model on changing fertility intentions between panel wave 1 and wave 3 is estimated. For women, a positive effect can be found of interrupting medium- and long-distance commuting or, surprisingly, continuing medium- and long-distance commuting on the intention to have a child within two years. Thirdly, for men and women who reported a fertil- ity intention in the fi rst wave, a longitudinal Heckman-selection probit-regression on the probability of having a child between wave 1 and wave 3 is estimated. It shows negative effects of medium- and long-distance commuting on having a child. Taken together, these fi ndings support the assumption that commuting plays a character- istically different role in different phases of the fertility-related decision process. Keywords: Spatial Mobility · Commuting · Fertility Intentions Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft Vol. 37, 3-4 (2012): 491-516 (Date of release: 06.12.2012) © Federal Institute for Population Research 2012 URL: www.comparativepopulationstudies.de DOI: 10.4232/10.CPoS-2012-05en URN: urn:nbn:de:bib-cpos-2012-05en2 • Johannes Huinink, Michael Feldhaus492 1 Introduction Job-related spatial mobility is an important dimension of the life course and is in- terrelated with other life domains, such as family and work. In general, job-related mobility can be the result of or the reason for other aspects of social mobility, i.e. all kind of biographical transitions in the individual life course. In regard to spatial mobility, two different types can be distinguished: residential and circular mobility (Schneider 2005). Residential mobility means a change of resi- dence within a town, a move within a country or to another country. Circular mobil- ity encompasses all kinds of transient spatial mobility. This includes daily commut- ing as well as other types of circular mobility, such as overnighting outside the own household for job reasons (Schneider/Meil 2008). Both types of spatial mobility, residential and circular, are interdependently linked with decisions and transitions in other life domains. They can also be interlinked with each other. For example, a move leads to a change of the distance to the workplace, if the job remains the same. Or, if the workplace changes, the question comes up whether to commute or to move (Kalter 1994). Spatial mobility is an important factor in individual- and relationship-related life plans (Courgeau 1990). In particular, job-related mobility can be of major relevance for family development and at the same time be infl uenced by family planning. For example, it seems plausible to assume that job-related commuting over a long dis- tance affects the desire to have another child and the timing of childbirth (Meil 2010a). Conversely, adapting mobility arrangements of a couple to the needs of family life can lead to a move or a change of the working place (Kulu/Milewski 2007). Residential mobility can be a consequence or part of the process of family forma- tion, when young parents realise that the housing situation is not an appropriate environment for their child. This again can bring about – under otherwise constant conditions – a change in job-related commuting. Therefore, spatial mobility helps to combine relationship- or family-related goals with goals in other life domains. Commuting can be an option to lower couple’s decision problems and to coordinate interests of both partners (Abraham/Auspurg/Hinz 2010). In this paper, we analyse some aspects of this complex relation between spa- tial mobility and family formation, respectively family extension. We analyse (1) whether even the intention of having a child within the next two years is correlated with the extent of commuting behaviour; (2) whether changes in the commuting behaviour come along with changes in regard to fertility intentions; and fi nally, (3) whether these have an impact on the probability of the realisation of an intended pregnancy. The empirical basis of our study are data from the Panel Analysis of Intimate Relationships and Family Dynamics (pairfam), hereafter called the German Family Panel (Huinink et al. 2011). In this panel project, which analyses couple and family dynamics, information on spatial mobility, fertility intentions as well as other impor- tant biographical characteristics (employment, relationship) is gathered. After presenting some relevant fi ndings from previous research, the theoretical discussion and the deduction of the hypothesis, we will introduce the operationali- Fertility and Commuting Behaviour in Germany • 493 sation of included variables and the methods applied in our analysis. In the context of empirical analyses, we will investigate the correlation between commuting be- haviour and the intention of having a child for women and men with the help of a cross sectional multiple equation model. In the second and third step, using data from the fi rst and third wave of pairfam, the impact of changes in the commuting behaviour on the fertility intention and on the realisation of a pregnancy are ana- lysed. In a résumé, we will summarise the fi ndings and discuss some limitations and future research. 2 Previous Research Empirical research on the relationship between fertility behaviour and spatial mobil- ity is quite new and focuses in particular on the correlation between residential mo- bility and family formation (Huinink/Wagner 1989; Kulu/Milewski 2007). Although our research focus is commuting behaviour, and residential mobility is just seen as a control variable, we will take a short look at previous research about the infl u- ence of residential mobility on family formation, because both types of mobility are strongly interrelated. After the fi rst ground-breaking works by Rossi (1955), only a few papers regard- ing the relationship between family formation, local opportunity structures, and residential mobility were published in the 1980s and 1990s (comp. Hervitz 1985; Courgeau 1985, 1990; Huinink/Wagner 1989; Strohmeier 1989; Wagner 1989). In these studies, some hypotheses were developed, which are being picked up again in current research (Clark/Davies Wizers 2007; Kulu 2005; 2006; Kulu/Vikat/Andersen 2007; Kulu/Boyle 2009; Kulu/Boyle/Andersen 2009). According to these hypotheses, residential mobility brings about crucial changes in general circumstances of life, which drive plans on family formation or family extension into the background and lead to a postponement of childbearing or a decrease in fertility. In turn, having chil- dren may lead to a residential move, particular of short distance, provided that this change of residence leads to an improvement of the conditions in which to raise the fi rst child or additional children (Kulu 2008; Rabe/Taylor 2010). Just the intention to have another child can also already trigger a move (Feijten/Mulder 2002). This is in line with previous research, showing that home ownership, particularly in combi- nation with a marriage, coincides with the realisation of a fertility intention (Davies Wizers 1998; Mulder/Wagner 2001; Kulu/Vikat 2007). Thus, it can be stated that there are interdependencies between residential mo- bility and fertility behaviour. Mobility does not only take place after the birth of a child, but often precedes an intended child birth. One can also assume – although this has not been empirically verifi ed up to now – that residential mobility is con- nected to fertility intentions or a realised birth, because parents want to cut back on the burden of time-consuming commuting and not only, because they wish to im- prove their living environment or would like to own a house. Therefore, in order to analyse the relationship between circular mobility and fertility intentions, residential • Johannes Huinink, Michael Feldhaus494 mobility has to be considered, too. It is important to account for residential mobility in the analysis of the relationship between commuting behaviour and fertility. In comparison to research on the relationship between fertility behaviour and res- idential mobility, studies dealing with the connection between family development and commuting behaviour are considerably less frequent. Provision of adequate comprehensive and representative data is restricted. Data on commuting behaviour are made available by the German Micro Census every four years (Grau 2009). The German Socio Economic Panel (GSOEP) provides information on the commuting distance and frequency of commuting (Wagner 1989; Stutzer/Frey 2007). Huinink and Kley (Kley 2010) collected information in a three wave panel study on the com- muting behaviour of residents in the cities of Magdeburg and Freiburg. Another study offers information on commuting behaviour in the Berlin-Brandenburg region (Wiethölter et al. 2009). The fi rst larger study, which stressed the effects of circu- lar mobility on family dynamics in Germany was conducted by Norbert Schneider (Schneider et al. 2002). This was continued in the comparative, international project „Job Mobilities and Family Lives in Europe“ (JobMob) (Schneider/Meil 2008). Schneider et al. (2002) suggested a typology of circular mobility, which has prov- en useful in empirical research. Besides residential mobility in regard to circular mobility, it differentiates between the amount of time needed to commute to the work place and whether people have to spend nights outside their home for job reasons. One distinguishes between: Non-mobiles (persons who are either not em- ployed, working at home, or are employed and commute in less than 60 minutes) and long-distance commuters (LDCs; persons commuting for more than 60 minutes one way). In order to get a more fi ne-graded differentiation, it was suggested to dis- tinguish the group of persons commuting in 30 to 60 minutes from the non-mobiles (Rüger et al. 2011). We call them medium-distance commuters (MDCs). Persons that have to spend more than 60 nights a year outside their home are called vari- mobiles. Shuttlers are people who have a second household near the workplace. Long-distance relationships (LDRs), in which both partners have their own house- hold for job-related reasons, are also considered (Schneider et al. 2002; Limmer/ Schneider 2008). Empirical fi ndings from cross-sectional analyses using fi rst wave data from the JobMob project give fi rst empirical evidence of a correlation between commuting behaviour and parenthood for men and women. If there are children in the house- hold, then women ś commuting behaviour is signifi cantly restricted. This is not the case for men (Schneider/Meil 2008). Similar fi ndings are reported by Kley (2010: 8): Fathers do not commute less often than childless men. However, mothers with young children in the household are more often non-mobile than other women. Meil (2010a/b) analysed the relationship between fertility and job-related spatial mobility with JobMob-data using different indicators. Here, the subjective percep- tion whether current childlessness is due to job-related demands was enquired. Particular mobile women (LDCs, vari-mobiles, shuttlers) agreed with this item more often than non-mobile women (39 % to 29 %; men: 29 % to 22 %). Furthermore, mobile women emphasised more often that they were postponing the birth of a (next) child (23 % to 7 %; men: 11 % to 7 %). The differences between mobile and Fertility and Commuting Behaviour in Germany • 495 non-mobile men are considerably lower. These results are confi rmed by analysing the number of actually realised children: Women and men who were faced with higher demands of job-related mobility were childless at a signifi cantly higher rate and had fewer children. These differences were much stronger for women com- pared to men. This correlation between long-distance commuting and women’s fertility behaviour had already been assumed by Courgeau (1990). These empirical fi ndings are replicated with data of the German Family Panel (pairfam). This study used the main concepts of the JobMob-project (Rüger et al. 2011). These data also show a signifi cant interaction effect between gender and circular mobility on the probability of being childless. Women who commute medium- or long-distance are signifi cantly more often childless (Rüger et al. 2011). The connection between circular mobility and postponing fertility in women is also supported by non-standardised interviews (Schneider et al. 2002). Schneider et al. (2002) mention that for childless mobile women, circular mobility has the func- tion to achieve a good fi nancial and occupational position. These women intend to realise their desire for a child, if it is possible to reduce the exhausting long-distance commuting which is hindering parenthood. This is particular true for long-distance relationships and shuttlers. Given the fact that mothers still are more responsible for childcare and take on more of the burden of domestic work, they can meet the high- er requirements of circular mobility less well. In some interviews, mothers report that they are not willing to move for job-related reasons, consider long-distance commuting or become a shuttler. Some mothers mention explicitly that they would refuse an attractive job offer, because this would undermine the interplay between family and work of both partners (Schneider et al. 2002). Thus, circular mobility is not a choice at all costs. The diffi cult relationship between (full-time) employment and motherhood is complemented by another important aspect: the question of how demanding a job is due to circular mobility and how compatible it is with par- enthood. In this paper, we do not primarily focus on the question of the realisation of the desire for a child, but concentrate on the connection between commuting and the intention of having a child. Fertility intention here means the wish of individuals to have a child within the next two years. Findings from motivation research show that having an intention on the one hand and realising it on the other are two interrelat- ed, subsequent processes, which are infl uenced by similar but also different factors (Heckhausen 2003; Brandtstädter 2001; Ajzen/Fishbein 2005). Therefore, one has to assume that the effect of commuting on the fertility intention and its realisation is different. Meanwhile, there are quite a considerable number of studies analysing the fac- tors infl uencing fertility intentions. They show that the expected benefi t and cost of a fi rst or an additional child have a signifi cant impact on the intention to have chil- dren (Billari et al. 2009; Philipov et al. 2006; Dommermuth et al. 2011). Internalised norms and attitudes of signifi cant others are also relevant. The expected support provided by parents, friends or other relatives is helpful, too (Bernardi et al. 2007; Dommermuth et al. 2011). Maul et al. (2010) show that the fertility intention, meas- ured with the intention to have a child within the next two years, depends on the • Johannes Huinink, Michael Feldhaus496 salience (relevance/priority) of this life goal compared to other goals in other life do- mains. It is also relevant whether individuals believe that important prerequisites for parenthood are fulfi lled (e.g. secure employment, solid relationship, adequate fi nan- cial resources, childcare facilities). Similar fi ndings were retrieved through analyses by Buhr et al. (2011): For East German women, suffi cient fi nancial resources and fl exible childcare arrangements are particular predictive for the intention to have a child within the next two years. In West Germany, the compatibility of job and fam- ily is also important. Using data from the Generations- and Gender Survey, Pailhé (2009) fi nds that an insecure employment situation has a negative effect on the fer- tility intention of women. However, this effect disappears, if a woman already has a fi rst child. Moreover, she reports a positive effect on the fertility intention, if the employer offers fl exible working hours and childcare. However, as mentioned before, the intention may not be seen on the same level as the realisation of a child. This is particular true for the general desire for a child, which again, has to be distinguished from the intention. Findings from Quesnel- Vallée/Morgan (2003) show, for example, that people who wish to have more than two children, overestimate the number of realised children, while those who want to have less than two children, underestimate it. This supports the assumption that the number of desired children and fertility intentions are not stable over time and that intentions must not be realised by all means. In the next section, we will try to answer the question of the role commuting behaviour plays in this context. 3 Theoretical Considerations and Hypotheses From a life course perspective, spatial mobility, employment and family develop- ment are dimensions of a highly interdependent process of individual welfare pro- duction. This process can be described as a more or less coherent sequence of individual action and biographical transitions, which are infl uenced by subjective evaluations, aspirations, expectations, and individual motives to act in different do- mains of the life course. Subjectively anticipated material and non-material costs and benefi ts are assessed. This assessment process is infl uenced by the actor’s subjective view on social, economical, and institutional circumstances in the short and long run, the available resources, biographical experiences, and personal traits. It comes to a decision to pursue such goals or related sub-goals, the achievement of which promises to improve or maintain the subjective well-being most effi ciently (Lindenberg/Frey 1993; Huinink/Schröder 2008). Spatial mobility, employment and raising children are different goals of the wel- fare production over the life course. Considering not only respective events and activities but also the decision-making process preceding them, the question of the determinants for generating goal-related intentions arises. This is the starting point of our analysis. The hypotheses, which we want to develop theoretically and test empirically, concentrate on the connection between commuting behaviour and the intention to have a child as well as the realisation of a child. Further factors which are important in the analysis of the decision in favour of parenthood and the realisa- Fertility and Commuting Behaviour in Germany • 497 tion of parenthood over the life course – such as the level of education, residential mobility, employment history – are included in our empirical analysis as control variables, but not discussed in detail. In our account of previous research, we reported that individuals with a higher demand for mobility are more often childless. This is particularly true for women. Now, the question arises whether this correlation can already be found in the fertil- ity-related decision process, more precisely, whether the correlation already arises when looking at the fertility intention or whether it is only correlated with its realisa- tion. At least two main strands of arguments can be distinguished. The fi rst strand of argument supports what we call the resource argument. According to fi ndings in the context of a qualitative analysis in an earlier study conducted by Schneider et al. (2002), for women, commuting is one way to achieve the pursued occupa- tional and fi nancial basis, which allows her to realise a given desire for a child. This would mean that women with more demanding commuting arrangements more often show an intention to have child than those who do not commute or whose commuting time is short. The burden of more demanding commuting is taken on in order to fulfi l the prerequisites (fi nancial and occupational security) for becoming a parent. The mobility arrangement, which might be intended as temporary, would be an instrumental sub-goal for attaining family-related subjective well-being. Mobil- ity can be perceived as an indicator for occupational success, which in turn should be supportive of a fertility intention and its realisation (Buhr et al. 2011). To achieve both, occupational success as well as family formation and family life, people rather commute for longer distances than taking a second residence for job reasons. The second line of argument addresses the compatibility argument. It is argued that a high demand for job-related mobility hinders the compatibility of family and work. Even without commuting, combining family and work can already be a dif- fi cult task. Thus, costly mobility arrangements should correlate negatively with the opportunities of organising parental tasks. This should be in particular true for those individuals who are responsible for childcare, which still applies more for women than for men (Geisler/Kreyenfeld 2006; Wengler/Trappe/Schmitt 2008). Therefore, one can assume that particularly women who commute longer distances postpone the realisation of a family. These women do not only have less children or stay child- less, but also do not even have the intention (at least in the short-run). We expect that an intention or plan to have children only occurs when they reduce the burden of mobility. Particularly for women who need to work for fi nancial reasons, who have got limited alternatives regarding the kind of job they are employed in, or who have problems returning to their workplace after childbirth, for those, problems in com- bining family and work can be an obstacle to plan a pregnancy in the short-run. The life goals of an occupational career and family are competitive. While spatial mobil- ity can be a crucial requirement for achieving a job-related goal, it simultaneously can be detrimental for the realisation of a fertility intention. This is often postponed, until a reduction of commuting occurs. Only if commuting is given up, it comes to an intention and to a realisation of a pregnancy. • Johannes Huinink, Michael Feldhaus498 It is diffi cult to say which argument has more weight. Previous fi ndings show that the question of compatibility of family and work is more relevant than the re- source-related argument, when it comes to the realisation of a fertility intention. Especially women with more demanding mobility arrangements (such as medium- or long-distance commuting) should have weaker fertility intentions compared to those with less demanding arrangements, due to problems with the compatibility of work and family. This expected effect could be less pronounced regarding the intention than regarding the realisation of a pregnancy. Whereas a fertility intention can go along with an intention to change the mobility arrangement fi rst, in order to achieve the requirements of compatibility, the latter intention might not be realised and childbirth will be postponed unintendedly. For men, we assume no or even small positive effects of commuting on family formation, i.e. we assume that the resource argument should be more relevant. In addition to the resource- and the compatibility argument, effects of self-selec- tion have to be considered. Those individuals who do not want to have any children are more willing to take on higher job-related mobility. This results in a negative correlation between fertility intentions and more demanding commuting, which is based on plans for life formed earlier in the life course (Schröder/Brüderl 2008). In order to handle this problem, it is necessary to use panel data, which allow account- ing for such selection effects. According to our previous arguments, we have for- mulated the fi rst two hypothesis, hereby differentiating between a cross sectional and a longitudinal perspective: 1. In a cross sectional analysis, time-consuming commuting of women is nega- tively correlated with the probability of the intention to have children and the realisation of a pregnancy. This effect does not appear among men, or should even be positive. 2. In the longitudinal perspective, reducing commuting distances increases the probability of the fertility intention among women and vice versa. For men, this is not the case, or the other way around. Having an intention to achieve a certain goal does not guarantee that the goal is actually achieved (Ajzen/Fishbein 2005). Unexpected factors as well as a wrong perception of the life situation can be hindering or supportive for a pursuit of a goal. The overestimation of the behavioural control or self-effi cacy can cause a discrep- ancy between intention and behaviour. In our discussion of previous research, we showed that this could also be true in the case of fertility intentions (Meil 2010a/b). There is also empirical evidence that demanding commuting arrangements pro- duce stress and health problems, which might have negative effects on the fertil- ity intention, too (Schneider et al. 2002). Therefore, we assume that a demanding commuting arrangement is one of the factors, which might be perceived as a strong obstacle for transition to parenthood, only when couples want to realise their inten- tion, and that it leads to a postponement of a planned child. This brings us to our third hypothesis: Fertility and Commuting Behaviour in Germany • 499 3. In the longitudinal perspective, the probability of realising a pregnancy by women who intend to have a child is negatively correlated with commuting distances. For men, no or moderately positive effects should be found. 4 Data and Operationalisation 4.1 Data In the following analyses we use data from the „Panel Analysis of Intimate Relation- ships and Family Dynamics“ (pairfam). This is a representative, interdisciplinary, longitudinal study of intimate relationships and familial living arrangements in Ger- many, which collects data every year (Huinink et al. 2011; Nauck 2012). In our analy- sis, we include the respondents (anchors), who were between 25 and 27 or 35 and 37 years old at the time of the fi rst panel wave in 2008/09. In the fi rst and third wave of pairfam, important information on job-related spa- tial mobility was collected in order to get the best possible picture of the couples’ and families’ organisation of everyday life. Therefore, pairfam also allows analysing job-related spatial mobility as such as well as its circumstances and consequences using a prospective cohort design. Because of the huge level of information being gathered in this panel and restricted survey time, detailed data on mobility are only collected in every second wave. In the other waves, a reduced number of questions are asked every year. As we use data from the fi rst and third wave of the pairfam- panel, we are able to analyse effects of spatial mobility and its changes between these waves on childbirth and on the change of fertility intentions. Pairfam provides the opportunity to distinguish between living arrangements and mobility as well as multi-locality patterns in greater detail. In addition, information on couples’ dy- namics and intentions, such as the intention to have a fi rst child, can be retrieved over a long period of time. Besides relevant cross sectional data from the fi rst wave changes in the activity status, the number of children, the place of residence and the mobility patterns are of particular interest for the longitudinal analysis. 4.2 Operationalisation of the Variables and Strategy of the Data Analysis Commuting Behaviour For the defi nition and classifi cation of types of job-related commuting and multi- locality, we differentiate between the various patterns following the work of Sch- neider et al. (Schneider et al. 2002; Schneider/Meil 2008; Limmer/Schneider 2008; Rüger et al. 2011): We distinguish between medium- and long-distance commuters • Johannes Huinink, Michael Feldhaus500 on the one hand and short-distance commuters (reference category) on the other.1 The short-distance commuters need less than 30 minutes (one way), the medium- distance commuters need at least 30 minutes but less than 60 minutes, and long- distance commuters travel 60 minutes and more. Fertility Intention The fertility intention is measured with the question, whether the respondent in- tended to have a child within the next two years. The variable used to measure the intention is coded dichotomously, even though there are four options for the answer. When respondents answered „do not know“ or „have not thought about it yet“ we assume that they do not intend to have a child within the next two years. If respondents had stated earlier in the interview that they did not expect to have more children or did not think about having children, this question was not asked. In these cases, we also assumed no intention. The question did not apply to pregnant respondents, either. They were excluded from the analysis. In the fi rst wave of pair- fam, this question was not asked if respondents reported that they or their partner was not able to have children or that they were homosexual. These cases were also excluded from the analysis, even though we are aware of the fact that many homo- sexual respondents may want to have children. Conception Not the birth of the child but the conception is our indicator for the realisation of a fertility intention. In this way, we avoid interpreting a change in the commuting behaviour occurring after conception as a causal reason for childbirth. The time of conception is reconstructed by using the information about the date of birth of a child or about the fact that a respondent (or his partner) is pregnant at the time of the interview in the third wave. In the fi rst case, we subtract 9 months from the date of birth. Therefore, pregnancies, which were not successful, are not considered. In the second case, we assume that conception took place approximately 4 months before the date of the interview of the third wave. Control Variables We have included a number of control variables. These are an indicator of the re- spective respondent’s age group (25-27 or 35-37 years); a dichotomous indicator of the level of qualifi cation, i.e. whether the respondent has graduated from a (techni- cal) university or not; and information on the labour force participation. Regarding the latter, we distinguish between full-time employment or self-employment, part- 1 The so-called „vari-mobiles“ are assigned to the respective commuter-categories. „Shuttlers“ with a job-related second household were excluded from the analysis, because the sample size was too small. Fertility and Commuting Behaviour in Germany • 501 time employment and non- or unemployment. As to the partner, we included an indicator of occupational status, i.e. whether the partner is an entrepreneur with employees or working as professional, in upper level service sector positions or in leading managerial positions in the private sector. To characterise the relationship constellation, we distinguish whether the cou- ples lived in one common household or not. We also considered the number of chil- dren and include an indicator of the salience of a (or another) child. It is a measure of the importance, which at the time of interview a respondent attributed to a fi rst or another child compared to the commitment in work and education, leisure and hobbies, friends or a relationship (Maul/Huinink/Schröder 2010; Brüderl et al. 2010). As we assume that mobility is correlated with the size of the city the respondents lived in, we included one more indicator. It indicates whether the respondents live in a place with up to 5000 inhabitants (small town) or in a big city with 500.000 inhabit- ants and more, places with a size in-between being the reference category. Finally, we included the information whether the respondents live in East or West Germany at the date of the fi rst wave’s interview. Models and Methods of Analysis In addition to some descriptive information, we analysed the proposed interdepend- encies in three steps applying probit-regressions for men and women separately. In a fi rst step, we estimated a multivariate probit regression model with cor- related errors in a cross-sectional analysis (hypothesis 1) using the data of the fi rst wave of the panel. It allowed us to obtain unbiased estimates of the correlation between commuting behaviour and the fertility intention respective to childless- ness (Cappellari/Jenkins 2003). The dependent variables are the fertility intention, reported childlessness and medium- and long-distance commuting of the respond- ents. The latter two dependent variables are also predictors of the fi rst one, and the indicator of medium- and long-distance commuting is modelled as a predictor of childlessness. Effects of common unobserved sources of a correlation between fertility intention, reported childlessness and commuting behaviour are accounted for in the multivariate model. The equations with fertility intention and childlessness as dependent variables allowed us to test for the expectation of different effects of commuting on intention and realisation of fertility. Because of technical reasons in our cross-sectional analysis, we included only respondents who were employed at the time of the interview of the fi rst wave. If we had included the non-employed respondents, the third equation of the model could not be identifi ed because, by defi nition, non- or unemployed persons do no com- mute for job reasons. We also included only those respondents who had a partner not in education and not pregnant at the time. In Table 1, the sample size and in- formation on the distribution of the variables, which have been included in the fi rst model, are displayed. In the second step, we estimated the effect of a change in the commuting scheme on a change in the fertility intention between the fi rst and the third wave in a differ- ence-model (Allison 2009). Here, the dependent variable is an indicator expressing • Johannes Huinink, Michael Feldhaus502 what kind of change in the intention between the two waves was observed. One can see a switch from a reported intention in the fi rst wave to no intention in the second wave and vice versa. The difference-model controls for time-constant unobserved heterogeneity of the respondents. In addition, information about changes or events in other life domains is included in the model: the occurrence of a pregnancy, a change in the employment status, short-distance residential mobility (distance of less than 50 km) and long-distance residential mobility (distance of 50 km and more). Medium- and long-distance com- muters were assembled into one category as the size of the sample used in this model is small. It can be assumed that this is not a problem, because in the cross- sectional models we found that their effects go in the same direction. In the model, we distinguished between respondents who did not commute these distances over the whole period of time (reference group), those who did commute medium- and long-distance over the whole period of time, and those who changed their commut- ing behaviour in one direction or the other. We also included time-constant vari- ables (e.g. age group, living in West Germany), the coeffi cients of which express whether the effect of these variables on the fertility intention changed between wave 1 and 3. In the difference model, all cases were excluded that did not experience a change in the fertility intention between wave 1 and 3. Again, we restricted our sample by including only those respondents who had a partner not in education and not preg- Tab. 1: Distribution of the dependent and independent variables in the cross- sectional multi-equation model; men and women (percentages) Men Women N=1572 N= 1679 Fertility intention reported 34.5 33.0 Childless 37.7 38.9 One child 23.5 24.2 Long-distance commuter 9.3 5.5 Medium-distance commuter 23.7 20.4 (Technical) university degree 22.9 23.8 Full-time or self-employment 95.6 51.2 High occupational status of partner 8.0 25.0 Cohabiting 86.1 83.1 Older age group 64.8 58.5 Living in West Germany 81.2 80.0 Living in a small town 17.8 17.3 Living in a big city 9.1 9.1 Salience of childbirth (mean) 1.76 1.65 Source: pairfam wave1; own calculations Fertility and Commuting Behaviour in Germany • 503 nant at wave 1. Additionally, we considered only those respondents who lived with the same partner during the whole observation period. In this case, we allowed non- and unemployment in the fi rst wave in order to keep those persons in the analysis who started commuting after they had gotten a job. To account for effects of changes in the employment status, indicators of those changes between wave 1 and 3 were included in the model. In Table 2, the descriptives of the variables con- sidered in the difference-model are shown. In the third step of analysis, we estimated the probability of a pregnancy (until wave 3) among respondents who in the fi rst wave reported an intention to have a child within the next two years. Here, we included all respondents independently from their employment status but living with the same partner over the time period between wave 1 and 3. In order to control the sample selection, we applied a probit- regression with a Heckman correction (Dubin/Rivers 1990). Predictors are – similar Tab. 2: Distribution of the dependent and independent variables included in the difference-model; men and women (percentages) Men Women N= 209 N= 260 Time dependent variables (changes between wave 1 and 3) Fertility intention: switch from no to yes 34.5 32.3 switch from yes to no 65.5 67.7 Commencement of medium-/long-distance commuting 8.6 8.5 Constantly commuting medium-/long-distance 22.5 11.2 Medium-/long-distance commuting ended 9.1 8.9 Switch from full-time to non-employment 4.3 11.9 Switch from full-time to part-time employment 0.5 3.1 Part-time employment unchanged or switch from non- to part-time employment 2.4 20.4 Constantly non-employed or switch from part-time to non- employment 6.2 33.5 Short-distance move (< 50 km) 24.4 25.4 Long-distance move ( 50 km) 2.9 4.2 Moved in with partner 11.5 11.5 Occurring pregnancy between W1 und W3 37.3 43.5 Time-constant variables at wave 1 High occupational status of partner 11.5 24.2 Living in West Germany 81.3 81.9 Older age group 55.0 35.4 Salience of childbirth (mean) 2.35 2.63 Source: pairfam waves 1 and 3; own calculations • Johannes Huinink, Michael Feldhaus504 to the difference model – changes in the commuting arrangements and the employ- ment status as well as residential mobility taking place before a potential pregnancy or the date of the interview in wave 3.2 Because of the small sample size, we put all men with the employment status full-time or self-employed over the whole period of time into one category and those who were not (i.e. who left full-time or part-time employment, who worked in part-time employment over the whole period of time or had been constantly non- or unemployed) into another category. In Table 3 de- scriptives of the variables are displayed. Tab. 3: Distribution of the variables included in the model of the realisation of a pregnancy; for men and women (percentages) Men Women N= 303 N=376 Time-dependent variables (changes between wave 1 and 3) Became pregnant 36.6 38.0 Commencement of medium-/long-distance commuting 6.6 10.1 Constantly commuting medium-/long-distance 23.4 12.5 Medium-/long-distance commuting ended 6.9 7.5 Switch from full-time to non-employment 1.0 1.9 Switch from full-time to part-time employment 0.7 3.7 Part-time employment unchanged or switch from non-to part-time employment 1.7 26.3 Constantly non-employed or switch from part-time to non-employment 3.0 18.4 Short-distance move (< 50 km) 12.9 14.1 Long-distance move ( 50 km) 2.3 2.9 Time-constant variables at wave 1 Childless 51.8 47.9 Having got one child 33.7 35.9 (Technical) university degree 27.1 29.8 Cohabiting 86.8 86.7 High occupational status of partner 8.3 27.9 Older age group 60.1 37.2 Living in West Germany 80.9 80.6 Salience of childbirth (mean) 2.85 2.86 Source: pairfam waves 1 and 3; own calculations 2 One could also estimate an event history regression model with time dependent covariates. It should provide similar results. Fertility and Commuting Behaviour in Germany • 505 5 Findings We will now present the empirical fi ndings following the three steps of analysis and test the hypotheses we formulated in chapter 3. First Hypothesis In the fi rst hypothesis, we proposed a negative relationship between the fertility intention and extensive commuting for women, not so for men. We analysed in the fi rst cross-sectional regression model whether this correlation can be found after accounting for other factors, which might be predictive for a fertility intention. We estimated a recursive, multivariate probit-regression with correlated errors, for women and men separately. Negative (positive) coeffi cients mean a negative (posi- tive) correlation of the independent variable with the probability of the considered status in one of the dependent variables. Included are men and women of the age groups 25-27 and 35-37, who or whose partner was not pregnant, had a partner and had fi nished education at the time of the survey. We fi rst estimated the same model for men and women separately. In a second model, we additionally included an interaction effect between employment status and the kind of commuting on the fertility intention. As this has shown not signifi cant for men, it is only displayed for women. Results are displayed in table 4. The most important fi nding for men and women in model 1 is that commuting distance does not play a considerable role for the fertility intention. In regard to women, this was not expected. The fi nding seems to show that the question of compatibility is less crucial than assumed. In the model including the interaction of the employment status (full-time/self-employed) with commuting behaviour, it can be seen that women working part-time and commuting long distances show a sig- nifi cantly lower probability of the intention to have a child within the next two years than full-time employed women (comp. the main effect of long-distance commut- ing in model 2). If the main effects and the interaction effect of long-distance com- muting and full-time employment are considered, it becomes apparent that long- distance commuting and full-time employed women do not differ from full-time employment women who commute short-distance. Descriptive analyses support the assumption that the household of part-time employed women are less well-off economically than their full-time employed counterparts. For them, the economic pre-conditions for having a child are possibly not fulfi lled suffi ciently. This supports the assumption that the resource-argument is of some relevance. The coeffi cients of the control variables in the equation of the fertility intention are mainly as one would expect. Men without children or with one child, who cohab- it, and who show a high salience of having a(nother) child, have high fertility inten- tions. For women, the level of education ((technical) university degree) is positively signifi cant, which can be interpreted as supporting the resource argument. Fur- thermore, women with no or one child and women with a high salience for (further) fertility intend to have a(nother) child within the next two years with a signifi cantly • Johannes Huinink, Michael Feldhaus506 Tab. 4: Mulitvariate probit-regression on the probability of a fertility intention, childlessness and medium-/long-distance commuting (recursive multivariate model with correlated errors) Men Women Model 1 Model 2 Coeff. Sig.-Lev. Coeff. Sig.-Lev. Coeff. Sig.-Lev. DV: Fertility intention Long-distance commuting (LDC) 0.05 0.83 -0.25 0.29 -0.96 0.02** Medium-distance commuting (MDC) -0.17 0.37 -0.20 0.33 -0.24 0.31 (Ref.: short distance) No child 1.12 0.00*** 1.00 0.00*** 1.01 0.00*** One child 0.97 0.00*** 0.73 0.00*** 0.73 0.00*** (Ref.: more than one child) (Technical) university degree 0.14 0.12 0.19 0.04** 0.19 0.04** Cohabiting 0.24 0.08* 0.14 0.20 0.13 0.20 High occupational status of partner -0.10 0.51 0.13 0.14 0.13 0.13 Full-time/self-employed (FS) 0.33 0.07* 0.19 0.16 0.14 0.34 (Ref.: part-time employed) Older age group -0.05 0.65 -0.46 0.00*** -0.45 0.00*** Living in West Germany 0.06 0.51 -0.04 0.72 -0.05 0.68 Interaction LDS * FS 0.88 0.03** Interaction MDC * FS 0.05 0.80 Salience of childbirth 0.31 0.00*** 0.29 0.00*** 0.29 0.00*** Constant -2.16 0.00*** -0.69 0.09* -0.68 0.09* DV: Childless Comp. women, Model Long-distance commuting (LDC) -0.07 0.75 0.56 0.02** Medium-distance commuting (MDC) -0.19 0.31 0.48 0.01*** (Ref.: short distance or no commuting) (Technical) university degree 0.05 0.61 0.41 0.00*** Cohabiting -1.39 0.00*** -0.67 0.00*** High occupational status of partner 1.21 0.00*** 0.01 0.90 Full-time/self-employed (FS) 0.15 0.41 1.51 0.00*** (Ref.: part-time employed) Older age group -1.15 0.00*** -1.48 0.00*** Living in West Germany 0.25 0.01*** 0.73 0.00*** Constant 3.50 0.00*** 2.33 0.00*** DV: Medium-/long-distance commuting Comp. women, Model (Technical) university degree 0.13 0.11 0.28 0.00*** Cohabiting 0.01 0.91 0.03 0.73 High occupational status of partner -0.16 0.19 0.04 0.61 Full-time/self-employed (FS) 0.20 0.24 0.18 0.01*** (Ref.: part-time employed) Older age group 0.05 0.50 -0.18 0.01*** Living in West Germany -0.09 0.32 -0.17 0.04** Living in a small town 0.01 0.94 -0.11 0.25 Living in a big city 0.48 0.00*** 0.49 0.00*** (Ref.: .: Living in medium size city) Constant -0.77 0.00*** -0.28 0.21 Number of cases 1.572 1.679 1.679 LR chi2(27) 783.41 1070.64 0.00 1076.59 0.00 Test of rho21 = rho31 = rho32 = 0 chi2(3); Sig. 1.35 0.72 2.27 0.52 2.41 0.49 *** sig. at level 0,01; ** sig. at level 0,05; * sig. at level 0,10 Source: pairfam, wave 1 Fertility and Commuting Behaviour in Germany • 507 higher probability. We see a negative age effect, meaning that the fertility intention is weaker in the older age group. The fi ndings in the equations estimating the probability of childlessness and me- dium-/long-distance commuting are also in accordance to what one would expect. However, with one exception, we will not comment on them in detail. Commuting behaviour is a strong predictor of the probability of childlessness for women. As we found virtually no correlation between time- consuming commuting and the fertility intention, childlessness is strongly and negatively correlated with it. This reinforces fi ndings of previous research. All models seem to be well-specifi ed and the effects of endogeneity seem to be small because the errors of the equations are not cor- related signifi cantly with each other. Second hypothesis The second hypothesis proposes that a change in the commuting arrangement should induce a change of the fertility intention – even though we cannot rule out that there is also an effect in the reverse direction. The commencement of medium- or long-distance commuting should trigger giving up such an intention and termi- nating commuting should foster it. As additional events, we consider changes in the employment status, residential mobility, and a change in regard to the cohabitation with a partner. In the difference model, the dependent variable expresses whether a respondent gives up a fertility intention or develops it between wave 1 and wave 3. The reference category in our model is giving up the intention. As cases with a con- stant intention are excluded from the analysis, we do not analyse whether a change or no change in the commuting arrangement have effects on the stability of the fertility intention.3 Mainly, in this analysis, we consider the independent variables of the previous model, but focus on time dependent attributes. Because of small sam- ple size, we put medium- and long-distance commuters into one category, as their effects show the same direction anyway. The results are displayed in table 5. It is not surprising that we did not fi nd a negative effect of the commencement of time-consuming commuting on the fertility intention of women. This is in accord- ance with the results of the fi rst analysis. If women gave up medium- or long-dis- tance commuting between wave 1 and wave 3, however, they reported an intention in the third wave more frequently compared to those who were constantly immobile or short-distance commuters. The coeffi cient is nearly signifi cant at the 10 per cent level. This can be interpreted differently. Either it could mean that giving up com- muting leads to a consideration of having a child, as we assumed. However, it could also mean that the intention to have a child triggers a discontinuation of time-con- suming commuting. As we do not have empirical information on the chronological order of these changes, we cannot decide which version is correct. 3 To do this in separate models, for respondents with or without a fertility intention at wave 1, we can estimate the probability of a change and control for sample selection using the Heckman- correction. • Johannes Huinink, Michael Feldhaus508 At fi rst sight, the signifi cant positive effect of continuous commuting between the wave 1 and 3 might be surprising. It means that in this group, compared to the reference category, the intention to have a child becomes more important as time goes by, after having been delayed before. This is another indication that the re- source argument might have a relevance for the commencement of the considera- tion to have a child. For men, no noteworthy effects of commuting were found. A change in the employment status showed no additional effects on the fertility intention, neither for men not for women. Residential mobility between wave 1 and wave 3 is positively correlated with the development of the intention to have a child. For women, these effects are not signifi cant. Again, the question of causality arises. Tab. 5: Probit-regression on the probability of developing or giving up a fertility intention between wave 1 and 3 (difference-model) Men Women Coeff. Sig.-Lev. Coeff. Sig.-Lev. Time-dependent variables Commencement of medium-/long-distance Commuting -0.14 0.76 0.11 0.75 Constantly commuting medium-/long-distance 0.03 0.90 0.79 0.01*** Medium-/long-distance commuting ended -0.56 0.21 0.51 0.14 (Ref.: Continuously short distance) Switch from full-time to non-employment 0.25 0.61 0.47 0.37 Switch from full-time to part-time employment -0.20 0.61 Part-time employment unchanged or switch from non- to part-time employment 0.71 0.32 0.08 0.76 Constantly non-employed or switch from part- time to non-employment 0.33 0.44 -0.17 0.51 (Ref.: Switch to or continuously full-time/self- employed) Short-distance move (< 50 km) 0.37 0.15 0.12 0.59 Long-distance move ( 50 km) 1.45 0.03** 0.30 0.52 Moved in with partner 0.96 0.01*** 0.19 0.54 Occurring pregnancy between wave 1 and 3 -1.26 0.00*** -1.21 0.00*** Time-constant variables High occupational status of partner 1.10 0.00*** -0.34 0.16 Living in West Germany 0.21 0.50 0.60 0.02** Older age group -0.93 0.00*** -0.67 0.00*** Salience of childbirth in wave 1 -0.14 0.03** -0.18 0.00*** Constant 2.07 0.00*** 1.39 0.01*** Number of cases 209.00 260.00 LR chi2(14) 92.38 91.45 Pseudo R2 0.34 0.28 *** sig. at level 0,01; ** sig. at level 0,05; * sig. at level 0,10 Source: pairfam, waves 1 and 3 Fertility and Commuting Behaviour in Germany • 509 Does an intention motivate residential mobility, or is it the other way round? An answer to this question needs a more precise modelling of the relationship, which has not been pursued in this study. Finally, a pregnancy which occurred between the two waves encourages the dismissal of further fertility intentions, as the desired child has already been realised in the meantime. The coeffi cients of the time-constant variables in this model are to be interpreted as interaction effects with time. A high occupational status of partners of male re- spondents results in a higher fertility intention in the third wave. This can be inter- preted as a delaying effect. The same is true for the West-effect in women, which is considerably smaller for men. For men and women, the effects of age and the sali- ence to have a child on the fertility intention become smaller in the third wave. The latter is comprehensible, as the salience could have changed between the waves for various reasons. The negative effect of being a member of the older age group means that the giving up of a fertility intention among these respondents becomes more and more probable as time goes by. Descriptive analyses support this. The discrepancy between the two age groups increases over time. Third hypotheses Finally, we analysed the interrelationship between commuting behaviour and the realisation of a fertility intention, which was stated in wave 1, between wave 1 and 3. We argued that the probability of the realisation should be negatively correlated with the time used for commuting. Commuting behaviour, the employment sta- tus – again including non-employment – and residential mobility were included in the model as time-dependent covariates. The model estimates are presented in table 6. In contrast to the model of changes in the fertility intention (table 5), we found signifi cant negative correlations between time-consuming mobility arrangements (commencement of or continuous medium-/long-distance commuting) and the probability of becoming pregnant. Only when women discontinue medium-/long- distance commuting, one obtains a positive coeffi cient, which is not signifi cant, however. This means that women who commute over these distances and who had intended to have a child within the next two years in the fi rst wave postponed the realisation, at least for the time being. In contrast to the intention, the realisation of a pregnancy is strongly hindered by the demands of commuting. Interestingly enough, we found a similar effect for the commencement of medium-/long-distance commuting among men. The coeffi cient is signifi cant. However, we will come back to that later. When full-time or self-employment is interrupted or not existent at all, it has a negative effect on the realisation of a pregnancy (of the partner), but this is only found to be signifi cant for men. Hence, it reinforces the well-known fi nding in fertil- ity research that the economic or fi nancial security is perceived as a prerequisite for having a(nother) child. We can also replicate evidence from the literature that a short-distance move is positively correlated with the realisation of a pregnancy – the • Johannes Huinink, Michael Feldhaus510 Tab. 6: Probit-regression on the probability of a pregnancy between wave 1 and 3 (Heckman-selection model) Men Women Coeff. Sig.-Lev. Coeff. Sig.-Lev. DV: Respondent/Partner of respondent becoming pregnant Time dependent variables Commencement of medium-/long-distance commuting -1.63 0.00*** -1.38 0.00*** Constantly commuting medium-/long-distance -0.16 0.17 -0.63 0.02** Medium-/long-distance commuting ended -0.13 0.55 0.11 0.67 (Ref.: Continuously short distance) Switch from full-time to non-employment -0.81 0.18 Switch from full-time to part-time employment -0.95 0.00*** -0.42 0.31 Part-time employment unchanged or switch from non-to part-time employment -0.02 0.90 Constantly non-employed or switch from part-time to non-employment -0.03 0.88 (Ref.: Switch to or continuously full-time/self- employed) Short-distance move (< 50 km) 0.41 0.08* 0.36 0.09* Long-distance move ( 50 km) 0.72 0.22 0.41 0.30 Time-constant variables No child in wave 1 -0.65 0.29 0.37 0.59 One child in wave 1 -0.37 0.48 0.53 0.25 (Ref.: more than one child) (Technical) university degree 0.09 0.62 0.06 0.70 Cohabiting -0.31 0.22 0.73 0.00*** High occupational status of partner -0.08 0.79 0.34 0.06* Older age group -0.05 0.77 -0.45 0.12 Living in West Germany -0.08 0.66 0.06 0.72 Salience of childbirth 0.04 0.74 0.04 0.78 Constant 0.65 0.64 -0.68 0.51 AV: Intention yes in wave 1 No child in wave 1 1.25 0.00*** 1.10 0.00*** One child in wave 1 0.95 0.00*** 0.71 0.00*** (Ref:. more than one child) High occupational status of partner -0.31 0.08* -025 0.01*** Older age group -0.13 0.23 -0.57 0.00*** Cohabiting 0.22 0.14 0.06 0.64 Salience of childbirth 0.32 0.00*** 0.25 0.00*** Constant -1.60 0.00*** -0.31 0.27 Number of cases 303 376 LR chi2(14/16) 27.04 0.01 56.67 0.00 LR test of rho21 = 0 chi2(1) 0.03 0.86 0.11 0.74 *** sig. at level 0,01; ** sig. at level 0,05; * sig. at level 0,10 Source: pairfam, waves 1 and 3 Fertility and Commuting Behaviour in Germany • 511 direction of causality is unclear, though. Cohabiting and a high occupational status of the partner also show signifi cant coeffi cients in the case of female respondents. As mentioned already, we estimated a Heckman-selection model, because only men and women who reported a fertility intention in the fi rst wave were included in the analysis. In the selection equation, we considered independent variables, which were relevant for the estimation of the fertility intention in the fi rst model displayed in table 4. These factors show the effects already known. The error terms of the main and the selection equation do not correlate signifi cantly with each other, though. 6 Summary and Conclusions The relationship between commuting behaviour and processes of family formation as well as extension has not been extensively analysed in previous research so far. Especially two defi cits can be identifi ed. Firstly, prospective longitudinal stud- ies are missing, which would allow analysing fertility intentions as they depend on time-dependent covariates. Secondly, a distinction between fertility intention and its realisation has rarely been made before. The aim of this article is to tackle these two points and thus add considerably to the current state of research. We focused on three research questions: How are fertility intentions correlated with commuting behaviour? Do changes in commuting behaviour lead to a change in fertility inten- tions? Do they also have an effect on the probability of the realisation of intended fertility? Our analyses provide evidence that commuting behaviour is only very weakly related to fertility intentions. However, the negative correlation between childless- ness and commuting distance known from the literature can be replicated for wom- en. Therefore, a remarkable difference between intention and realisation becomes apparent. In the cross-sectional model, contrary to our expectations, we did not fi nd a negative correlation between commuting distance and the intention to have a child within the next two years, neither for men nor for women. The question of the com- patibility of work and motherhood might not be as relevant at this point in the fertil- ity-related decision process as we had assumed. And the resource argument could to be given more weight - as least at the point of the commencement of an intention. In addition, the negative effect of commuting distance on childlessness could have been overestimated in the model, because in the cross-sectional analysis we could not control the effect of childlessness on commuting behaviour. In the difference model, we found that women who continuously commute show a shift towards a fertility intention with a signifi cantly higher probability. No effects could be found among men. More clearly than the cross-sectional analysis, the re- sults of this model support the idea that intentions might be affected by economic resources to a higher degree than we expected. In the last model, the relevance of commuting for the realisation of a fertility intention was analysed. For women, high commuting demands considerably affect the probability of a pregnancy. In contrast to previous evidence, a similar effect • Johannes Huinink, Michael Feldhaus512 shows up among men who started medium- or long-distance commuting. Might this be a biographical transition, in which at least for a certain period of time men consider the additional duty of fatherhood as not reasonable? However, one should assume that this is just a short-term effect and not a matter of a compatibility prob- lem for men, because continuous commuting between wave 1 and 3 does not have any notable effect. From the substantive point of view, the article emphasises the importance of a more detailed modelling of decision processes and the separate consideration of intentions and behaviour. In regard to fertility, we have been able to show interest- ing and systematic differences in the effects of other behavioural variables on these two aspects. From a methodological point of view, our fi ndings and the comparison between the results of the cross-sectional and the longitudinal models show the relevance of time-varying panel data. In the difference model, we could exclude the potentially biasing infl uences of time-constant unobserved heterogeneity. There- fore, we could test statements regarding a causal relationship between commuting behaviour and fertility in a more proper way. Certain issues of causality must be left open, though. This applies, for example, to the interrelationship between chang- ing fertility intentions and shifts in the commuting behaviour. In future research, one could also have a closer look at the connection between residential mobility, commuting arrangements, and family development, using the data we used in this study. References Abraham, Martin; Auspurg, Katrin; Hinz, Thomas 2010: Migration Decisions Within Du- al-Earner Partnerships: A Test of Bargaining Theory. In: Journal of Marriage and Fam- ily 72: 876-892 [doi: 10.1111/j.1741-3737.2010.00736.x]. Ajzen, Icek; Fishbein, Martin 2005: The Infl uence of Attitudes on Behavior. In: Albar- racín, Dolores; Johnson, Blair T.; Zanna, Mark P. (Eds.): The Handbook of Attitudes. New Jersey: Lawrence Elrbaum: 173-223. Allison, Paul D. 2009: Fixed Effects Regression Models. Thousend Oaks: Sage. Bernardi, Laura; Keim, Sylvia; von der Lippe, Holger 2007: Social Infl uences on fertility: A comparative mixed methods study in Eastern and Western Germany. In: Journal of Mixed Methods Research 1: 23-47 [doi: 10.1177/2345678906292238]. Billary, Francesco, C.; Philipov, Dimiter; Testa, Maria 2009: Attitudes, norms and per- ceived behavioural control: Explaining fertility intentions in Bulgaria. In: European Journal of Population 25: 439-465 [doi: 10.1007/s10680-009-9187-9]. Brandstädter, Jochen 2001: Entwicklung, Intentionalität, Handeln. Stuttgart: Kohlham- mer Verlag. Brüderl, Josef et al. 2010: pairfam Data Manual, Wave 1. Mannheim Buhr, Petra; Huinink, Johannes; Boehnke, Mandy; Maul, Katharina 2011: Kinder oder keine? Institutionelle Rahmenbedingungen und biographische Voraussetzungen für die Familiengründung und -erweiterung in Ost- und Westdeutschland. In: Brüderl, Josef; Castiglioni, Laura; Schumann, Nina (Eds.): Partnerschaft, Fertilität und inter- generationale Beziehungen. Würzburg: Ergon Verlag: 175-203. Fertility and Commuting Behaviour in Germany • 513 Cappellari, Laurenzo; Jenkins, Stephen P. 2003: Multivariate probit regres- sion using simulated maximum likelihood. The Stata Journal 3: 278-294 [doi: 10.1016/0304-4076(93)90049-B]. Clark, William A. V.; Davies Wizers, Susanne 2007: Family Migration and Mobility Se- quences in the United States: Spatial Mobility in the Context of the Life Course. In: Demographic Research 17,20: 591-622. Courgeau, Daniel 1985: Interaction between spatial mobility, family and career life-cy- cle: a French survey. In: European Sociological Review 1,2: 139-162. Courgeau, Daniel 1990: Migration, family, and career: A life course approach. In: Baltes, Paul B.; Featherman, David L.; Lerner, Richard M. (Eds.): Life-span development and Behaviour. Hillsdale, NJ: Lawrence Erlbaum: 219-255. Davies Wizers, Susanne 1998: Linking household transitions and housing transitions: a longitudinal analysis of renters. In: Environment and Planning A 30: 615-630 [doi: 10.1068/a300615]. Dommermuth, Lars; Klobas, Jane; Lappegard, Trude 2011: Now or later? The Theory of Planned Behavior and timing of fertility intentions. In: Advances of the Life Course Research 16: 42-53. Dubin, Jeffrey A.; Rivers, Douglas 1990: Selection Bias in Linear Regression, Log- it and Probit Models. In: Sociological Methods and Research 18: 360-390 [doi: 10.1177/0049124189018002006]. Feijten, Peteke; Mulder, Clara H. 2002: The timing of household events and hous- ing events in the Netherlands: a longitudinal perspective. In: Housing Studies 17,5: 773-792 [doi: 10.1080/026730302200000980 8]. Geisler, Esther; Kreyenfeld, Michaela 2006: Müttererwerbstätigkeit in Ost- und West- deutschland. In: Zeitschrift für Familienforschung 18,3: 333-360. Grau, Andreas 2009: Pendler: Die Mehrheit nimmt weiter das Auto. STATMagazin: Wiesbaden: Destatis. Green, Anne E.; Canny, Angela 2003: Geographical Mobility: Family Impacts. Bristol: Policy Press. Heckhausen, Heinz 2003: Motivation und Handeln. Berlin: Springer Verlag. Hervitz, Hugo M. 1985: Selectivity, Adaptation, or Disruption? A Comparison of Alterna- tive Hypotheses on the Effects of Migration on Fertility: The Case of Brazil. In: Interna- tional Migration Review 19,2: 293-317. Huinink, Johannes et al. 2011: Panel Analysis of Intimate Relationships and Family Dy- namics (pairfam) – Conceptual Framework and Design. In: Zeitschrift für Familien- forschung 23,1: 77-101. Huinink, Johannes; Feldhaus, Michael 2009: Family Research from the Life Course Per- spective. In: International Sociology 24,3: 299-324 [doi: 10.1177/0268580909102910]. Huinink, Johannes; Schröder, Torsten 2008: Skizzen zu einer Theorie des Lebensver- laufs. In: Diekmann, Andreas; Eichner, Klaus; Schmid, Peter; Voss, Thomas (Eds.): Rational Choice: Theoretische Analysen und empirische Resultate. Wiesbaden: VS Verlag für Sozialwissenschaften: 291-308. Huinink, Johannes; Wagner, Michael 1989: Regionale Lebensbedingungen, Migration und Familienbildung. In: Kölner Zeitschrift für Soziologie und Sozialpsychologie 41,4: 669-689. Kalter, Frank 1994: Pendeln statt Migration. In: Zeitschrift für Soziologie 23,6: 460-476. • Johannes Huinink, Michael Feldhaus514 Kley, Stefanie 2010: Multilokalität als Strategie zur Nutzung von Chancen. In: Soeffner, Hans-Georg (Eds.): Unsichere Zeiten: Herausforderungen und gesellschaftliche Trans- formationen. Wiesbaden: VS Verlag: CD-Rom Veröffentlichung. Kulu, Hill 2005: Migration and Fertility: Competing Hypotheses Re-examined. In: Euro- pean Journal of Population 21,1: 51-87 [doi: 10.1007/s10680-005-3581-8] Kulu, Hill 2006: Fertility of Internal Migrants: Comparison between Austria and Poland. In: Population Space and Place 12,3: 147-170 [doi: 10.1002/psp.406]. Kulu, Hill 2008: Fertility and spatial mobility in the life course: evidence from Austria. In: Environment and Planning A, 40,3: 632-652 [doi: 10.1068/a3914]. Kulu, Hill; Boyle, Paul J. 2009: High Fertility in City Suburbs: Compositional or Con- textual Effects? In: European Journal of Population 25,2: 157-174 [doi:10.1007/ s10680-008-9163-9]. Kulu, Hill; Boyle, Paul B.; Andersen, Gunnar 2009: High suburban fertility: Evidence from four Northern European countries. In: Demographic Research 21,31: 915-944. Kulu, Hill; Milewski, Nadja 2007: Family change and migration in the life course: An in- troduction. In: Demographic Research 17,19: 567-590 Kulu, Hill; Vikat, Andres 2007: Fertility differences by housing type: an effect of housing conditions or of selective moves? In: Demographic Research 17: 775-802. Kulu, Hill; Vikat, Andres; Andersen, Gunnar 2007: Settlement size and fertility in the Nordic countries. In: Population Studies 51,3: 265-285 [doi: 10.1080/00324720701571749]. Limmer, Ruth; Schneider, Norbert F. 2008: Studying Job-Related Spatial Mobility in Eu- rope. In: Schneider, Norbert F.; Meil, Gerardo (Eds.): Mobile Living Across Europe I. Relevance and Diversity of Job-Related Spatial Mobility in Six European Countries. Opladen/Farmington Hills: Barbara Budrich Publishers: 13-46. Lindenberg, Sigwart; Frey, Bruno 1993: Alternatives, frames, and relative pric- es: a broader view of rational choice. In: Acta Sociologica 36: 191-205 [doi: 10.1177/000169939303600304]. Mayer, Karl Ulrich 1990: Lebensverläufe und sozialer Wandel – Anmerkungen zu einem Forschungsprogramm. In: Mayer, Karl Ulrich (Eds.): Lebensverläufe und sozialer Wan- del. In: Kölner Zeitschrift für Soziologie und Sozialpsychologie, Sonderheft 31: 7-21. Maul, Katharina; Huinink, Johannes; Schröder, Torsten 2010: Elternschaftsentschei- dungen im Spannungsfeld berufl icher und familiärer Planungen. In: Walper, Sabine; Wendt, Eva-Verena (Eds.): Partnerschaften und die Beziehungen zu Eltern und Kindern. Würzburg: Ergon Verlag: 215-241. Meil, Gerardo 2010a: Geographic Job Mobility and Parenthood Decisions. In: Zeitschrift für Familienforschung 22,2: 171-195. Meil, Gerardo 2010b: Job Mobility and Family Life. In: Schneider, Norbert; Collet, Beate (Eds.): Mobile Living Across Europe II. Causes and Consequences of Job-Related Spatial Mobility in Cross-National Comparison. Opladen: Barbara Budrich Publishers: 215-235. Mulder, Clara H.; Wagner, Michael 2001: The connection between family formation and fi rst-time home ownership in the context of West Germany and the Netherlands. In: European Journal of Population 17,2: 137-164 [doi:10.1023/A:1010706308868]. Nauck, Bernhard; Brüderl, Josef; Huinink, Johannes; Walper, Sabine 2012: Beziehun- gs- und Familienpanel (pairfam). GESIS Datenarchiv. Köln. ZA5678 Datenfi le Version 3.0.0 [doi: 10.4232/pairfam.5678.3.0.0]. Fertility and Commuting Behaviour in Germany • 515 Pailhé Ariane 2009: Work-family balance and childbearing intentions in France, Germa- ny and the Russian Federation. In: UNECE (Eds.): How generations and gender shape demographic change. Towards Policies Based on Better Knowledge. United Nations/ Geneva. Philipov, Dimiter; Speder, Zsolt; Billari, Francesco C. 2006: Soon, Later, or Ever? The Impact of Anomie and Social Capital on Fertiltiy Intentions in Bulgaria and Hungary. In: Population Studies 60: 289-308 [doi: 10.1080/00324720600896080]. Quesnel-Vallée, Amelie; Morgan, Philip S. 2003: Missing the target? Correspondence of fertility intentions and behavior in the U.S. In: Population Research and Policy Review 22: 497-525 [doi: 10.1023/B:POPU.0000021074.33415.c1]. Rabe, Birgitta; Taylor, Mark 2010: Residential Mobility, Quality of Neighborhood and Life Course Events. In: Journal of the Royal Statistical Society A, 173,3: 531-555 [doi: 10.1111/j.1467-985X.2009.00626.x]. Reuschke, Darja 2010: Multilokales Wohnen. Wiesbaden: VS-Verlag. Rossi, Peter H. 1955: Why families move: a study in the social psychology of urban resi- dential mobility. Glencoe, Ill.: Free Press. Rüger, Heiko; Feldhaus, Michael; Becker, Katharina; Schlegel, Monika 2011: Zirkuläre berufsbezogene Mobilität in Deutschland: Vergleichende Analysen mit zwei repräsentativen Surveys zu Formen, Verbreitung und Relevanz im Kontext der Part- nerschafts- und Familienentwicklung. In: Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft 36,1 [doi: 10.4232/10.CPoS-2011-05de]. Schier, Michaela 2009: Räumliche Entgrenzung von Arbeit und Familie. Die Herstellung von Familie unter Bedingungen von Multilokalität. In: Informationen zur Raumentwick- lung 1/2: 55-66. Schneider, Norbert F. 2005: Einführung: Mobilität und Familie. Wie Globalisierung die Menschen bewegt. In: Zeitschrift für Familienforschung 17,2: 90-95. Schneider, Norbert F.; Limmer, Ruth; Ruckdeschel, Kerstin 2002: Mobil, fl exibel, gebun- den. Familie und Beruf in der mobilen Gesellschaft. Frankfurt/M.: Campus. Schneider, Norbert F.; Meil, Gerardo (Eds.) 2008: Mobile Living Across Europe I. Rel- evance and Diversity of Job-Related Spatial Mobility in Six European Countries. Op- laden: Barbara Budrich Publishers. Schröder, Jette; Brüderl, Josef 2008: Der Effekt der Erwerbstätigkeit von Frauen auf die Fertilität: Kausalität oder Selbstselektion? In: Zeitschrift für Soziologie 37: 117-136. Statistisches Bundesamt 2005: Leben und Arbeiten in Deutschland. Wiesbaden. Stutzer, Alois; Frey, Bruno S. 2007: Commuting and life satisfaction in Germany. In: In- formationen zur Raumentwicklung 2/3: 179-189. Strohmeier, Klaus P. 1989: „Movers“ and „Stayers“. Räumliche Mobilität und Familienen- twicklung. In: Herlth, Alois; Strohmeier, Klaus P. (Eds.): Lebenslauf und Familienent- wicklung. Opladen: Leske+Budrich: 165-188. Wagner, Michael 1989: Räumliche Mobilität im Lebensverlauf: Eine empirische Untersu- chung sozialer Bedingungen der Migration. Stuttgart: Enke Verlag. Wengler, Annelene; Trappe, Heike; Schmitt, Christian 2008: Partnerschaftliche Arbeit- steilung und Elternschaft. Analysen zur Aufteilung von Hausarbeit und Elternaufgaben auf Basis des Generations and Gender Survey. Materialien zur Bevölkerungswissen- schaft 127. Bundesinstitut für Bevölkerungsforschung. Wiesbaden. Wiethölter, Doris; Bogai, Dieter; Carstensen, Janette 2009: Pendlerbericht Berlin- Brandenburg 2009. In: IAB-Regional 3/2010. • Johannes Huinink, Michael Feldhaus516 Translated from the original text by the authors, for information only. The reviewed and author’s authorised original article in German is available under the title “Fertilität und Pendelmobilität in Deutschland”, DOI 10.4232/10.CPoS-2012-05de or URN urn:nbn:de:bib-cpos-2012-05de8, at http:// www.comparativepopulationstudies.de. Date of submission: 25.10.2011 Date of Acceptance: 22.02.2012 Prof. Dr. Johannes Huinink ( ), Dr. Michael Feldhaus. Universität Bremen, Institut für empirische und angewandte Soziologie (EMPAS), Bremen, Germany. E-Mail: huinink@empas.uni-bremen.de, feldhaus@empas.uni-bremen.de URL: http://www.soziologie.uni-bremen.de © Federal Institute for Population Research 2013 – All rights reserved Published by / Herausgegeben von Prof. Dr. Norbert F. Schneider Federal Institute for Population Research D-65180 Wiesbaden / Germany Managing Editor / Verantwortlicher Redakteur Frank Swiaczny Editorial Assistant / Redaktionsassistenz Katrin Schiefer Language & Copy Editor (English) / Lektorat & Übersetzungen (englisch) Amelie Franke Copy Editor (German) / Lektorat (deutsch) Dr. Evelyn Grünheid Layout / Satz Beatriz Feiler-Fuchs E-mail: cpos@bib.bund.de Scientifi c Advisory Board / Wissenschaftlicher Beirat Jürgen Dorbritz (Wiesbaden) Paul Gans (Mannheim) Johannes Huinink (Bremen) Marc Luy (Wien) Clara H. Mulder (Groningen) Notburga Ott (Bochum) Peter Preisendörfer (Mainz) Board of Reviewers / Gutachterbeirat Martin Abraham (Erlangen) Laura Bernardi (Lausanne) Hansjörg Bucher (Bonn) Claudia Diehl (Göttingen) Andreas Diekmann (Zürich) Gabriele Doblhammer-Reiter (Rostock) Henriette Engelhardt-Wölfl er (Bamberg) E.-Jürgen Flöthmann (Bielefeld) Alexia Fürnkranz-Prskawetz (Wien) Beat Fux (Zürich) Joshua Goldstein (Rostock) Karsten Hank (Köln) Sonja Haug (Regensburg) Franz-Josef Kemper (Berlin) † Michaela Kreyenfeld (Rostock) Aart C. Liefbroer (Den Haag) Kurt Lüscher (Konstanz) Dimiter Philipov (Wien) Tomáš Sobotka (Wien) Heike Trappe (Rostock) Comparative Population Studies – Zeitschrift für Bevölkerungswissenschaft www.comparativepopulationstudies.de ISSN: 1869-8980 (Print) – 1869-8999 (Internet)