Common use of Data and Methods Clause in Contracts

Data and Methods. This study uses data from the British Household Panel Survey (BHPS). The BHPS is a panel survey initiated in 1991, when a nationally representative sample of 10,300 individuals from 5,500 UK households were selected and interviewed (▇▇▇▇▇▇ et al., 2010). These individuals have been re-interviewed annually on a wide range of topics, with additional households added to the panel from Scotland, Wales and Northern Ireland in 1999 and 2001. In addition to possessing a large sample surveyed over many time points, the BHPS is ideal for this project for two main reasons. The first key advantage of the BHPS is that it gathers information about moving desires and expectations from all adults living with a sample member. This enables the construction of variables indicating (dis)agreements in moving desires and expectations between partners living in couples. A second advantage of the BHPS is its comparatively low attrition rate (Berthoud, 2000). While movers are known to be more likely to drop out of the sample than non-movers, the BHPS typically records whether individuals have moved even if they were not re-interviewed (Buck, 2000). This enables us to retain these cases in our analyses of actual moving behaviour. This study makes use of a person-year file based on eight waves of the BHPS covering the years 1998-2006. Earlier waves could not be used as information on moving expectations was not gathered until 1998. Wave 11 (2001) cases were excluded as housing satisfaction information was not gathered during this survey sweep. Given the aims of this paper, the research population consisted of individuals who had an identified and opposite sex ‘lawful spouse’ or ‘live-in partner’ in their household. A very small number of person-years where the partners lived in an institution were excluded, as these couples are unlikely to have independent housing careers. Person- years where key household information was missing (such as housing tenure or income) were removed. Cases were also dropped where it was impossible to compute household level similarity or (dis)agreement variables, as only one partner had responded to the relevant survey question. Moving desires were coded using the response to the question ‘If you could choose, would you stay here in your present home or would you prefer to move somewhere else?’ Similarly, moving expectations were identified from the response to the question ‘Do you expect you will move in the coming year?’. A small proportion of respondents replying that they ‘did not know’ whether they desired or expected to move were treated as having no desire or expectation of moving. This is because these individuals appear not to have given moving much thought. In addition, analysis was restricted to couples that stayed intact between two consecutive waves. Couples were defined as ‘movers’ if both changed their address between t and t+1 and they remained in the same household and relationship. Likewise, couples were defined as ‘stayers’ if neither moved and they remained partners. This procedure takes into account that individuals may not change marital status but may change partner between waves (particularly if they cohabit). Longer observation intervals for identifying a move (for example over the subsequent 2 or 3 years) were rejected due to the phrasing of the survey questions, which explicitly obtained the respondent’s moving expectations over the next year. In addition, using longer observation windows would ignore that the respondent’s expressed desires and expectations may have changed at the intervening waves. If only one partner moved or both partners moved but to different households, the couple were assumed to have separated and these person-years were omitted (see ▇▇▇▇▇, 2008b for a similar sample selection procedure). After transforming the person-year file into a couple-year format, 30,617 couple-years remained, provided by 6,675 couples over an average of 4.6 waves. The first set of cross-tabulations linked various household level independent variables to the occurrence of disagreements in moving desires between partners. To investigate the effects of disagreements on the subsequent moving behaviour of couples, random effects (panel) logistic regression models were used (▇▇▇▇▇, 2003). The dependent variable in these models is a binary variable indicating whether the household moved over the subsequent survey year (0=no move, 1=move). The control variables in these models contain lagged values, with transition variables measuring the occurrence of life events (such as changes in employment status) between the observation of moving desires at t and moving behaviour at t+1. Table 1 provides a summary of all variables used in these analyses. Panel models are valuable as they account for the non-independence of observations, as couple-year cases are nested within couples. The descriptive results presented in Table 1 and Figure 1 confirm that it is important to consider the pre-move thoughts of both partners in couples. Partners often disagree about whether a move is desired (19.11% of cases) or expected (4.36% of cases). Figure 1 shows how partner (dis)agreement on moving desires and the actual mobility rate vary with the age of the older partner in the couple. Disagreements appear to occur fairly consistently across the life course, although younger couples are more likely to disagree than older couples. While total agreement rates remain fairly stable, the composition of this agreement shifts from desiring to move to not desiring to move as age increases. It is important to note that the actual mobility rate is consistently lower than the proportion of couples where one or both partners desire to move (sum of disagree and both desire). This suggests that many people may be unable to act upon their moving desires, or that moving desires are often unattainable and hence abandoned. Table 2 presents data on the associations between partner similarity and (dis)agreement on moving desires. The results provide only weak support for the idea that partners who are demographically and socioeconomically more similar to one another are less likely to disagree about whether moving is desirable. The age gap separating partners appears unrelated to the propensity for partners to disagree about whether moving is desirable, although couples separated by the largest age gaps are slightly more likely to disagree. Ethnically mixed couples are more likely to disagree than ethnically homogenous couples, despite the idea that only more committed individuals are willing to enter into such unions. A gap in educational levels between partners seems unrelated to (dis)agreement on moving desires. Both dual and single earner couples are more likely to disagree than couples where neither partner is employed. The results in the lower section of Table 2 provide preliminary support for the contention that disagreement about whether moving is desirable is more likely when partners also disagree about the quality of their dwelling or neighbourhood. Disagreements are most likely to occur if the partners already disagree about whether they are satisfied with their dwelling or dislike their neighbourhood. Further analysis (not shown) reveals that it is almost always the partner who is unhappy with their dwelling or neighbourhood who desires to move. This suggests that individual moving desires are stimulated by personal subjective evaluations of dwelling and neighbourhood conditions (Landale and Guest, 1985). This interpretation is further supported by the strong links between shared negative evaluations (particularly of the neighbourhood) and shared desires to move: more than 96% of couples who agree on disliking the neighbourhood also share a desire to move. Couples who disagree about their housing or neighbourhood conditions also often agree that moving is desirable. This suggests that people often take their partner’s happiness with their current location into account when expressing their own moving desires. This would not be visible in an individual level study. Table 3 presents descriptive results linking the level of shared commitments to the relationship to moving desire (dis)agreements. There is somewhat mixed support for the commitment hypothesis, which proposed that possessing fewer joint commitments is associated with a greater propensity for partners to disagree about the desirability of moving. Disagreements are more likely among cohabiters than married couples, with cohabiters also much more likely to agree that moving is desirable. This may indicate that individuals select into marriage when they foresee that a shared future living arrangement is feasible, typically as relationship duration and stability increase. Disagreements also appear to be more common for couples with children, with the incidence of disagreement generally increasing with the age of the children (see also ▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇, 2009). This suggests that although families with children are fairly immobile, it is common for one or both partners to still desire to move. There is also some evidence that tenure commitments are linked to desire disagreements. Highly committed homeowning couples disagree in 18.55% of cases, whereas disagreements are slightly more common amongst renting couples (just over 21%). Again this suggests that individuals often enter committed states only when they perceive that the needs of both partners can be met through residence at a single location. Overall we have found little convincing evidence that levels of partner similarity are associated with moving desire disagreements. We did find that disagreements are most likely to occur when the partners disagree about the quality of their dwelling or neighbourhood. These findings provide initial support for conceptual models of residential mobility decision making (see Lu, 1999). There is also some support for hypothesis 2, suggesting that greater levels of commitment are associated with a reduced propensity to disagree about whether moving is desirable. Table 4 contains descriptive results testing the third and fourth hypotheses. The results indicate that taking the moving desires of both partners into account more accurately predicts whether couples subsequently move. The upper section of Table 4 links the desires of only the male partner to the couple’s moving behaviour over the next year. Ignoring the views of the female partner, these results show that 15.90% of couples where the male desires to move also actually move. The lower section of the table reveals however that the likelihood of the male partner’s desire to move being realised is heavily dependent upon the views of his partner. If only the male partner desires to move, then a move occurs in 7.57% of cases. If both partners desire to move then a move occurs in over 20% of cases. These findings support the hypothesis that moving desires are most likely to be realised if shared by both partners. This demonstrates that linking only one partner’s desires to the actual moving behaviour of the couple leads to inaccurate estimates of how strongly desires are associated with actual moves. Shared moving desires are much more likely to be realised than desires which are not shared. Table 5 contains the results from five panel logistic regression models estimating the likelihood of couples making joint moves. These models enable robust hypothesis testing, by controlling for the effects of background characteristics known to affect mobility. Our main interest is in how partner (dis)agreements in evaluations of housing and/or neighbourhood quality, moving desires and moving expectations, affect the moving propensity of couples. It is possible that interview conditions may have affected our measurements of (dis)agreements. It is likely that disagreements are less likely to be expressed if both partners are interviewed together. Further analysis revealed that partners are somewhat more likely to disagree if they completed the relevant section of the interview separately than if they were interviewed together. As partners were not interviewed separately in approximately 50% of cases, we may undercount disagreements in our analyses. To ensure that our results are robust, the models were rerun with a variable indicating the interview conditions included as an extra control. The model results were almost identical to the models without this control variable (not shown). Model 1 includes only housing dissatisfaction and neighbourhood assessments as independent variables. The model shows that couples are more likely to move if one or especially both partners are dissatisfied with their dwelling or dislike their neighbourhood. Consistent with ▇▇▇▇ and ▇▇▇▇▇▇’▇ (2010) findings, moves are more likely to occur if only the woman dislikes the neighbourhood than if only the man dislikes the neighbourhood. These parameters remain strong and significant when a range of control variables (but without moving desires and expectations) are added in Model 2. In general the control variables have the effects anticipated from the literature, apart from the negative coefficient of the cohabitation dummy (although this is not significant). The propensity to move decreases with age, and couples with children are less likely to move than those without (particularly if the children are school age or older). Changes in the number of children in the household do not appear significantly linked to mobility. High levels of education are associated with a higher probability to move, while single and particularly dual earner couples are less likely to move than couples where neither partner is employed. Interestingly, reductions in the number of people in employment are also associated with moving. This may be due to moves related to retirement. With higher levels of income the likelihood of moving increases. Private renters are more likely to move than those in other housing tenures, while room stress is also associated with a greater propensity to move. The longer people stayed in their current dwelling, the less likely they are to move. Further analyses (not shown here) demonstrate that there is little evidence of any significant regional or period effects on moving behaviour. Model 3 only includes the moving desires and expectations of both partners. The results support both hypotheses 3 and 4. Desiring to move is associated with a greater propensity to actually move, particularly if this desire is shared between partners. Shared moving expectations are very strongly linked to mobility, although moves are also likely if only expected by one partner (especially if the woman expects to move). The effects of moving desires and expectations remain stable when control variables are included in Model 4. Most of the control variable parameters are similar to those in Model 2, although there are some minor changes in significance levels (for instance education level becomes insignificant). Model 4 fits the data much better than Model 2, as shown by the considerably lower log likelihood value in Model 4. This suggests that desires and expectations are more strongly linked to actual moves than evaluations of dwelling and neighbourhood conditions. Finally, Model 5 contains all variables included in the previous models. Most of the control variables have similar effects to those estimated in the previous models. The most important finding is that some of the effects of housing satisfaction and all of the effects of disliking the neighbourhood become insignificant when desires and expectations are included in the same model. This indicates that subjective evaluations of dwelling and neighbourhood conditions are associated with moving desires and expectations, with these desires and expectations in turn conditioning the propensity to move (▇▇▇ et al., 1994; ▇▇▇▇▇, 1955). ▇▇▇▇ and ▇▇▇▇▇▇ (2010) have previously reported that whether the female partner dislikes the neighbourhood has a particularly strong effect on whether the household subsequently moves. While this is correct, our results demonstrate that this is also partially dependent on how disliking the neighbourhood affects the moving desires and expectations of both partners. Interestingly, after also controlling for moving desires and expectations, couples remain

Data and Methods. This study uses data from the British Household Panel Survey (BHPS). The BHPS is a panel survey initiated in 1991, when a nationally representative sample of 10,300 individuals from in 5,500 UK households were selected and interviewed (▇▇▇▇▇▇▇▇, 2000; ▇▇▇▇▇▇ et al., 2010). These individuals have been re-interviewed annually on a wide range of topics, with additional households added to the panel from Scotland, Wales and Northern Ireland in 1999 and 2001. In addition to possessing a large sample surveyed over many time points, the BHPS is ideal for this project for two main reasons. The first key advantage of the BHPS is that it gathers information about moving desires and expectations from all adults living with a sample member. This enables the construction of variables indicating (dis)agreements in moving desires and expectations between partners living in couples. A second advantage of the BHPS is its comparatively low attrition rate (Berthoud, 2000). While movers are known to be more likely to drop out of the sample than non-movers, the BHPS typically records whether individuals have moved even if they were not re-re- interviewed (Buck, 2000). This enables us to retain these cases in our analyses of actual moving behaviour. This study makes use of a person-year file based on eight waves of the BHPS covering the years 1998-2006. Earlier waves could not be used as information on moving expectations was not gathered until 1998. Wave 11 (2001) cases were excluded as housing satisfaction information was not gathered during this survey sweep. Given the aims of this paper, the research population consisted of individuals who had an identified and opposite sex ‘lawful spouse’ or ‘live-in partner’ in their household. A very small number of person-years where the partners lived in an institution were excluded, as these couples are unlikely to have independent housing careers. Person- Person-years where key household information was missing (such as housing tenure or income) were removed. Cases were also dropped where it was impossible to compute household level similarity or (dis)agreement variables, as only one partner had responded to the relevant survey question. Moving desires were coded using the response to the question ‘If you could choose, would you stay here in your present home or would you prefer to move somewhere else?’ Similarly, moving expectations were identified from the response to the question ‘Do you expect you will move in the coming year?’. A small proportion of respondents replying that they ‘did not know’ whether they desired or expected to move were treated as having no desire or expectation of moving. This is because these individuals appear not to have given moving much thought. In addition, analysis was restricted to couples that stayed intact between two consecutive waves. Couples were defined as ‘movers’ if both changed their address place of residence between t and t+1 and they remained in the same household and relationship. Likewise, couples were defined as ‘stayers’ if neither moved and they remained partners. This procedure takes into account that is more appropriate than just comparing marital status across waves to check for relationship changes, as individuals may not change marital status but may change partner between waves (particularly if they cohabit). Longer observation intervals for identifying a move (for example over the subsequent 2 or 3 years) were rejected due to the phrasing of the survey questions, which explicitly obtained the respondent’s moving expectations over the next year. In addition, using longer observation windows would ignore that the respondent’s expressed desires and expectations may have changed at the intervening waves. If only one partner moved or both partners moved but to different households, the couple were assumed to have separated and these person-years were omitted (see ▇▇▇▇▇, 2008b 2008a for a similar sample selection procedure). After transforming the person-person- year file into a couple-year format, 30,617 couple-years remained, provided by 6,675 couples over an average of 4.6 waves. The To address the first set of research aim, cross-tabulations linked various household level independent variables to the occurrence of disagreements in moving desires between partners. To investigate the effects of disagreements on the subsequent moving behaviour of couples, random effects (panel) logistic regression models were used (▇▇▇▇▇, 2003). The dependent variable in these models is a binary variable indicating whether the household moved over the subsequent survey year (0=no move, 1=move). The control variables in these models contain lagged values, with transition variables measuring the occurrence of life events (such as changes in employment status) between the observation of moving desires at t and moving behaviour at t+1. Table 1 provides a summary of all variables used in these analyses. Panel models are valuable in longitudinal research as they account for the non-independence of observations, observations (as couple-year cases are nested within couples). The descriptive results presented in Table 1 and Figure 1 confirm that it is important to consider the pre-move thoughts of both partners in couples. Partners often disagree about whether a move is desired (19.11% of cases) or expected (4.36% of cases). Figure 1 shows how partner (dis)agreement on moving desires and the actual mobility rate vary rates varies with the age of the older partner in the couple. Disagreements appear to occur fairly consistently across the life course, although younger couples are more likely to disagree than older couples. While total agreement rates remain fairly stable, the composition of this agreement shifts from desiring to move to not desiring to move as age increases. It is important to note that the actual mobility rate is consistently lower than the proportion of couples where one or both partners desire to move (sum of disagree and both desire). This suggests that many people may be unable to act upon their moving desires, or that moving desires are often unattainable and hence abandoned. Table 2 presents data on the associations between partner similarity and (dis)agreement on moving desires. The results provide only weak support for the idea that partners who are demographically and socioeconomically more similar to one another are less likely to disagree about whether moving is desirable. The age gap separating partners appears unrelated to the propensity for partners to disagree about whether moving is desirable, although couples separated by the largest age gaps are slightly more likely to disagree. Ethnically mixed couples are more likely to disagree than ethnically homogenous couples, despite the idea that only more committed individuals are willing to enter into such unions. A gap in educational levels between partners seems unrelated to (dis)agreement on moving desires. Both dual and single earner couples are more likely to disagree than couples where neither partner is employed. This is probably a proxy age effect, as non employed couples tend to be retired. The results in the lower section of Table 2 provide preliminary support for the contention hypothesis that disagreement about whether moving is desirable is more likely when partners also disagree about the quality of their dwelling or neighbourhood. Disagreements are most likely to occur if the partners already disagree about whether they are satisfied with their dwelling or dislike their neighbourhood. Further analysis (not shown) reveals that it is almost always the partner who is unhappy with their dwelling or neighbourhood who desires to move. This suggests that individual moving desires are stimulated by personal subjective evaluations of dwelling and neighbourhood conditions (Landale and Guest, 1985; ▇▇▇▇▇, 1955). , This interpretation is further supported by the strong links between shared negative evaluations (particularly of the neighbourhood) and shared desires to move: more than 96% of couples who agree on disliking the neighbourhood also share a desire to move. Couples who disagree about their housing or neighbourhood conditions also often agree that moving is desirable. This suggests that people often take their partner’s happiness with their current location into account when expressing their own moving desires. This would not be visible in an individual level study. Table 3 presents descriptive results linking the level of shared commitments to the relationship to moving desire (dis)agreements. There is somewhat mixed support for the commitment hypothesis, which proposed that possessing fewer joint commitments is associated with a greater propensity for partners to disagree about the desirability of moving. Disagreements are more likely among cohabiters than married couples, with cohabiters also much more likely to agree that moving is desirable. This suggests that age may indicate that individuals select into marriage when they foresee that a shared future living arrangement is feasiblebe driving these correlations, as cohabiters are typically as relationship duration younger than married couples (▇▇▇▇▇▇▇ and stability increaseVan Ham, 2010). Disagreements also appear to be more common for couples with children, with the incidence of disagreement generally increasing with the age of the children (while agreement that moving is desirable simultaneously drops) (see also ▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇, 2009). This suggests that although families with children are fairly immobile, it is common for one or both partners to still desire to move. There is also some evidence that tenure commitments are linked to desire disagreements. Highly committed homeowning couples disagree in 18.55% of cases, whereas disagreements are slightly more common amongst renting couples (just over 21%). Again this suggests Given that individuals often enter committed states only when they perceive selection into home ownership is facilitated by wealth accumulation with age, it may be that the needs older average age of both partners can be met through residence at a single locationhomeowners is driving these correlations (see Figure 1). Overall we have found little convincing evidence that levels of partner similarity are associated with moving desire disagreements. We did find support for the first hypothesis that disagreements are most disagreement about whether moving is desirable is more likely to occur when the if partners disagree about the quality of their dwelling or neighbourhood. These findings provide initial support for conceptual models of residential mobility decision making (see Lu▇▇, 1999). There is also some support for hypothesis 2, suggesting that greater levels of commitment are associated with a reduced propensity to disagree about whether moving is desirable. Table 4 contains descriptive results testing the third and fourth hypotheses. The results indicate that taking the moving desires of both partners into account more accurately predicts whether couples subsequently move. This is at the heart of this paper’s contribution to the literature. The upper section of Table 4 links the desires of only the male partner to the couple’s moving behaviour over the next year. Ignoring the views of the female partner, these results show that 15.90% of couples where the male desires to move also actually move. The lower section of the table reveals however that the likelihood of the male partner’s desire to move being realised is heavily dependent upon the views of his partner. If only the male partner desires to move, then a move occurs in only 7.57% of cases. If both partners desire to move then a move occurs in over 20% of cases. These findings support the hypothesis that moving desires are most likely to be realised if shared by both partners. This demonstrates The results also demonstrate that linking only one partner’s desires to the actual moving behaviour of the couple leads to inaccurate estimates of how strongly desires are associated with actual moves. Shared moving desires are much more likely to be realised than desires which are not shared. Table 5 contains the results from five panel logistic regression models estimating the likelihood of couples making joint moves. These models enable robust hypothesis testing, by controlling for the effects of background characteristics known to affect mobility. Our main interest is in how partner (dis)agreements in evaluations of housing and/or neighbourhood quality, moving desires and moving expectations, affect the moving propensity of couples. It is possible that interview conditions may have affected our measurements of (dis)agreements. It is likely that disagreements are less likely to be expressed if both partners are interviewed together. Further analysis revealed that partners are somewhat more likely to disagree if they completed the relevant section of the interview separately than if they were interviewed together. As partners were not interviewed separately in approximately 50% of cases, we may undercount disagreements in our analyses. To ensure that our results are robust, the models were rerun with a variable indicating the interview conditions included as an extra control. The model results were almost identical to the models without this control variable (results not shown). Model 1 includes only housing dissatisfaction and neighbourhood assessments as independent variables. The model shows that couples are more likely to move if one or especially both partners are dissatisfied with their dwelling or dislike their neighbourhood. Consistent with ▇▇▇▇ and ▇▇▇▇▇▇’▇ (2010) findings, moves are more likely to occur if only the woman dislikes the neighbourhood than if only the man dislikes the neighbourhood. These parameters remain strong and significant when a range of control variables (but without moving desires and expectations) are added in Model 2. In general the control variables have the effects anticipated from the literature, apart from the negative coefficient of the cohabitation dummy (although this is not significant). The propensity to move decreases with age, and couples with children are less likely to move than those without (particularly if the children are school age or older). Changes in the number of children in the household do not appear significantly linked to mobility. High levels of education are associated with a higher probability to move, while single and particularly dual earner couples are less likely to move than couples where neither partner is employed. Interestingly, reductions in the number of people in employment are also associated with moving. This may be due to moves related to retirement. With higher levels of income the likelihood of moving increases. Private renters are more likely to move than those in other housing tenures, while room stress is also associated with a greater propensity to move. The longer people stayed in their current dwelling, the less likely they are to move. Further analyses (not shown here) demonstrate that there is little evidence of any significant regional or period effects on moving behaviour. Model 3 only includes the moving desires and expectations of both partners. The results support both hypotheses 3 and 4. Desiring to move is associated with a greater propensity to actually move, particularly if this desire is shared between partners. Shared moving expectations are very strongly linked to mobility, although moves are also likely if only expected by one partner (especially if the woman expects to move). The effects of moving desires and expectations remain stable when control variables are included in Model 4. Most of the control variable parameters are similar to those in Model 2, although there are some minor changes in significance levels (for instance education level becomes insignificant). Model 4 fits the data much better than Model 2, as shown by the considerably lower log likelihood value in Model 4. This suggests that desires and expectations are more strongly linked to actual moves than evaluations of dwelling and neighbourhood conditions. Finally, Model 5 contains all variables included in the previous models. Most of the control variables have similar effects to those estimated identified in the previous models. The most important finding is that some of the effects of housing satisfaction and all of the effects of disliking the neighbourhood become insignificant when desires and expectations are included in the same model. This indicates that subjective evaluations of dwelling and neighbourhood conditions are associated with moving desires and expectations, with these desires and expectations in turn conditioning the propensity to move (▇▇▇ et al., 1994; ▇▇▇▇▇, 1955). ▇▇▇▇ and ▇▇▇▇▇▇ (2010) have previously reported that whether the female partner dislikes the neighbourhood has a particularly strong effect on whether the household subsequently moves. While Our results nuance this is correctfinding, our results demonstrate as it seems that this is also partially dependent depends on how disliking the neighbourhood affects is translated into the moving desires and expectations of both partners. Interestingly, after also controlling for moving desires and expectations, couples remainremain significan

Data and Methods. This study uses data from the British Household Panel Survey (BHPS). The BHPS is a panel survey initiated in 1991, when a nationally representative sample of 10,300 individuals from 5,500 UK households were selected and interviewed (▇▇▇▇▇▇ et al., 2010). These individuals have been re-interviewed annually on a wide range of topics, with additional households added to the panel from Scotland, Wales and Northern Ireland in 1999 and 2001. In addition to possessing a large sample surveyed over many time points, the BHPS is ideal for this project for two main reasons. The first key advantage of the BHPS is that it gathers information about moving desires and expectations from all adults living with a sample member. This enables the construction of variables indicating (dis)agreements in moving desires and expectations between partners living in couples. A second advantage of the BHPS is its comparatively low attrition rate (Berthoud, 2000). While movers are known to be more likely to drop out of the sample than non-movers, the BHPS typically records whether individuals have moved even if they were not re-interviewed (Buck, 2000). This enables us to retain these cases in our analyses of actual moving behaviour. This study makes use of a person-year file based on eight waves of the BHPS covering the years 1998-2006. Earlier waves could not be used as information on moving expectations was not gathered until 1998. Wave 11 (2001) cases were excluded as housing satisfaction information was not gathered during this survey sweep. Given the aims of this paper, the research population consisted of individuals who had an identified and opposite sex ‘lawful spouse’ or ‘live-in partner’ in their household. A very small number of person-years where the partners lived in an institution were excluded, as these couples are unlikely to have independent housing careers. Person- years where key household information was missing (such as housing tenure or income) were removed. Cases were also dropped where it was impossible to compute household level similarity or (dis)agreement variables, as only one partner had responded to the relevant survey question. Moving desires were coded using the response to the question ‘If you could choose, would you stay here in your present home or would you prefer to move somewhere else?’ Similarly, moving expectations were identified from the response to the question ‘Do you expect you will move in the coming year?’. A small proportion of respondents replying that they ‘did not know’ whether they desired or expected to move were treated as having no desire or expectation of moving. This is because these individuals appear not to have given moving much thought. In addition, analysis was restricted to couples that stayed intact between two consecutive waves. Couples were defined as ‘movers’ if both changed their address between t and t+1 and they remained in the same household and relationship. Likewise, couples were defined as ‘stayers’ if neither moved and they remained partners. This procedure takes into account that individuals may not change marital status but may change partner between waves (particularly if they cohabit). Longer observation intervals for identifying a move (for example over the subsequent 2 or 3 years) were rejected due to the phrasing of the survey questions, which explicitly obtained the respondent’s moving expectations over the next year. In addition, using longer observation windows would ignore that the respondent’s expressed desires and expectations may have changed at the intervening waves. If only one partner moved or both partners moved but to different households, the couple were assumed to have separated and these person-years were omitted (see ▇▇▇▇▇, 2008b for a similar sample selection procedure). After transforming the person-year file into a couple-year format, 30,617 couple-years remained, provided by 6,675 couples over an average of 4.6 waves. The first set of cross-tabulations linked various household level independent variables to the occurrence of disagreements in moving desires between partners. To investigate the effects of disagreements on the subsequent moving behaviour of couples, random effects (panel) logistic regression models were used (▇▇▇▇▇, 2003). The dependent variable in these models is a binary variable indicating whether the household moved over the subsequent survey year (0=no move, 1=move). The control variables in these models contain lagged values, with transition variables measuring the occurrence of life events (such as changes in employment status) between the observation of moving desires at t and moving behaviour at t+1. Table 1 provides a summary of all variables used in these analyses. Panel models are valuable as they account for the non-independence of observations, as couple-year cases are nested within couples. The descriptive results presented in Table 1 and Figure 1 confirm that it is important to consider the pre-move thoughts of both partners in couples. Partners often disagree about whether a move is desired (19.11% of cases) or expected (4.36% of cases). Figure 1 shows how partner (dis)agreement on moving desires and the actual mobility rate vary with the age of the older partner in the couple. Disagreements appear to occur fairly consistently across the life course, although younger couples are more likely to disagree than older couples. While total agreement rates remain fairly stable, the composition of this agreement shifts from desiring to move to not desiring to move as age increases. It is important to note that the actual mobility rate is consistently lower than the proportion of couples where one or both partners desire to move (sum of disagree and both desire). This suggests that many people may be unable to act upon their moving desires, or that moving desires are often unattainable and hence abandoned. Table 2 presents data on the associations between partner similarity and (dis)agreement on moving desires. The results provide only weak support for the idea that partners who are demographically and socioeconomically more similar to one another are less likely to disagree about whether moving is desirable. The age gap separating partners appears unrelated to the propensity for partners to disagree about whether moving is desirable, although couples separated by the largest age gaps are slightly more likely to disagree. Ethnically mixed couples are more likely to disagree than ethnically homogenous couples, despite the idea that only more committed individuals are willing to enter into such unions. A gap in educational levels between partners seems unrelated to (dis)agreement on moving desires. Both dual and single earner couples are more likely to disagree than couples where neither partner is employed. The results in the lower section of Table 2 provide preliminary support for the contention that disagreement about whether moving is desirable is more likely when partners also disagree about the quality of their dwelling or neighbourhood. Disagreements are most likely to occur if the partners already disagree about whether they are satisfied with their dwelling or dislike their neighbourhood. Further analysis (not shown) reveals that it is almost always the partner who is unhappy with their dwelling or neighbourhood who desires to move. This suggests that individual moving desires are stimulated by personal subjective evaluations of dwelling and neighbourhood conditions (Landale and Guest, 1985). This interpretation is further supported by the strong links between shared negative evaluations (particularly of the neighbourhood) and shared desires to move: more than 96% of couples who agree on disliking the neighbourhood also share a desire to move. Couples who disagree about their housing or neighbourhood conditions also often agree that moving is desirable. This suggests that people often take their partner’s happiness with their current location into account when expressing their own moving desires. This would not be visible in an individual level study. Table 3 presents descriptive results linking the level of shared commitments to the relationship to moving desire (dis)agreements. There is somewhat mixed support for the commitment hypothesis, which proposed that possessing fewer joint commitments is associated with a greater propensity for partners to disagree about the desirability of moving. Disagreements are more likely among cohabiters than married couples, with cohabiters also much more likely to agree that moving is desirable. This may indicate that individuals select into marriage when they foresee that a shared future living arrangement is feasible, typically as relationship duration and stability increase. Disagreements also appear to be more common for couples with children, with the incidence of disagreement generally increasing with the age of the children (see also ▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇, 2009). This suggests that although families with children are fairly immobile, it is common for one or both partners to still desire to move. There is also some evidence that tenure commitments are linked to desire disagreements. Highly committed homeowning couples disagree in 18.55% of cases, whereas disagreements are slightly more common amongst renting couples (just over 21%). Again this suggests that individuals often enter committed states only when they perceive that the needs of both partners can be met through residence at a single location. Overall we have found little convincing evidence that levels of partner similarity are associated with moving desire disagreements. We did find that disagreements are most likely to occur when the partners disagree about the quality of their dwelling or neighbourhood. These findings provide initial support for conceptual models of residential mobility decision making (see Lu▇▇, 1999). There is also some support for hypothesis 2, suggesting that greater levels of commitment are associated with a reduced propensity to disagree about whether moving is desirable. Table 4 contains descriptive results testing the third and fourth hypotheses. The results indicate that taking the moving desires of both partners into account more accurately predicts whether couples subsequently move. The upper section of Table 4 links the desires of only the male partner to the couple’s moving behaviour over the next year. Ignoring the views of the female partner, these results show that 15.90% of couples where the male desires to move also actually move. The lower section of the table reveals however that the likelihood of the male partner’s desire to move being realised is heavily dependent upon the views of his partner. If only the male partner desires to move, then a move occurs in 7.57% of cases. If both partners desire to move then a move occurs in over 20% of cases. These findings support the hypothesis that moving desires are most likely to be realised if shared by both partners. This demonstrates that linking only one partner’s desires to the actual moving behaviour of the couple leads to inaccurate estimates of how strongly desires are associated with actual moves. Shared moving desires are much more likely to be realised than desires which are not shared. Table 5 contains the results from five panel logistic regression models estimating the likelihood of couples making joint moves. These models enable robust hypothesis testing, by controlling for the effects of background characteristics known to affect mobility. Our main interest is in how partner (dis)agreements in evaluations of housing and/or neighbourhood quality, moving desires and moving expectations, affect the moving propensity of couples. It is possible that interview conditions may have affected our measurements of (dis)agreements. It is likely that disagreements are less likely to be expressed if both partners are interviewed together. Further analysis revealed that partners are somewhat more likely to disagree if they completed the relevant section of the interview separately than if they were interviewed together. As partners were not interviewed separately in approximately 50% of cases, we may undercount disagreements in our analyses. To ensure that our results are robust, the models were rerun with a variable indicating the interview conditions included as an extra control. The model results were almost identical to the models without this control variable (not shown). Model 1 includes only housing dissatisfaction and neighbourhood assessments as independent variables. The model shows that couples are more likely to move if one or especially both partners are dissatisfied with their dwelling or dislike their neighbourhood. Consistent with ▇▇▇▇ and ▇▇▇▇▇▇’▇ (2010) findings, moves are more likely to occur if only the woman dislikes the neighbourhood than if only the man dislikes the neighbourhood. These parameters remain strong and significant when a range of control variables (but without moving desires and expectations) are added in Model 2. In general the control variables have the effects anticipated from the literature, apart from the negative coefficient of the cohabitation dummy (although this is not significant). The propensity to move decreases with age, and couples with children are less likely to move than those without (particularly if the children are school age or older). Changes in the number of children in the household do not appear significantly linked to mobility. High levels of education are associated with a higher probability to move, while single and particularly dual earner couples are less likely to move than couples where neither partner is employed. Interestingly, reductions in the number of people in employment are also associated with moving. This may be due to moves related to retirement. With higher levels of income the likelihood of moving increases. Private renters are more likely to move than those in other housing tenures, while room stress is also associated with a greater propensity to move. The longer people stayed in their current dwelling, the less likely they are to move. Further analyses (not shown here) demonstrate that there is little evidence of any significant regional or period effects on moving behaviour. Model 3 only includes the moving desires and expectations of both partners. The results support both hypotheses 3 and 4. Desiring to move is associated with a greater propensity to actually move, particularly if this desire is shared between partners. Shared moving expectations are very strongly linked to mobility, although moves are also likely if only expected by one partner (especially if the woman expects to move). The effects of moving desires and expectations remain stable when control variables are included in Model 4. Most of the control variable parameters are similar to those in Model 2, although there are some minor changes in significance levels (for instance education level becomes insignificant). Model 4 fits the data much better than Model 2, as shown by the considerably lower log likelihood value in Model 4. This suggests that desires and expectations are more strongly linked to actual moves than evaluations of dwelling and neighbourhood conditions. Finally, Model 5 contains all variables included in the previous models. Most of the control variables have similar effects to those estimated in the previous models. The most important finding is that some of the effects of housing satisfaction and all of the effects of disliking the neighbourhood become insignificant when desires and expectations are included in the same model. This indicates that subjective evaluations of dwelling and neighbourhood conditions are associated with moving desires and expectations, with these desires and expectations in turn conditioning the propensity to move (▇▇▇ et al., 1994; ▇▇▇▇▇, 1955). ▇▇▇▇ and ▇▇▇▇▇▇ (2010) have previously reported that whether the female partner dislikes the neighbourhood has a particularly strong effect on whether the household subsequently moves. While this is correct, our results demonstrate that this is also partially dependent on how disliking the neighbourhood affects the moving desires and expectations of both partners. Interestingly, after also controlling for moving desires and expectations, couples remain

Data and Methods. This study uses data from the British Household Panel Survey (BHPS). The BHPS is a panel survey initiated in 1991, when a nationally representative sample of 10,300 individuals from 5,500 UK households were selected and interviewed (▇▇▇▇▇▇ et al., 2010). These individuals have been re-interviewed annually on a wide range of topics, with additional households added to the panel from Scotland, Wales and Northern Ireland in 1999 and 2001. In addition to possessing a large sample surveyed over many time points, the BHPS is ideal for this project for two main reasons. The first key advantage of the BHPS is that it gathers information about moving desires and expectations from all adults living with a sample member. This enables the construction of variables indicating (dis)agreements in moving desires and expectations between partners living in couples. A second advantage of the BHPS is its comparatively low attrition rate (Berthoud, 2000). While movers are known to be more likely to drop out of the sample than non-movers, the BHPS typically records whether individuals have moved even if they were not re-interviewed (Buck, 2000). This enables us to retain these cases in our analyses of actual moving behaviour. This study makes use of a person-year file based on eight waves of the BHPS covering the years 1998-2006. Earlier waves could not be used as information on moving expectations was not gathered until 1998. Wave 11 (2001) cases were excluded as housing satisfaction information was not gathered during this survey sweep. Given the aims of this paper, the research population consisted of individuals who had an identified and opposite sex ‘lawful spouse’ or ‘live-in partner’ in their household. A very small number of person-years where the partners lived in an institution were excluded, as these couples are unlikely to have independent housing careers. Person- Person-years where key household information was missing (such as housing tenure or income) were removed. Cases were also dropped where it was impossible to compute household level similarity or (dis)agreement variables, as only one partner had responded to the relevant survey question. Moving desires were coded using the response to the question ‘If you could choose, would you stay here in your present home or would you prefer to move somewhere else?’ Similarly, moving expectations were identified from the response to the question ‘Do you expect you will move in the coming year?’. A small proportion of respondents replying that they ‘did not know’ whether they desired or expected to move were treated as having no desire or expectation of moving. This is because these individuals appear not to have given moving much thought. In addition, analysis was restricted to couples that stayed intact between two consecutive waves. Couples were defined as ‘movers’ if both changed their address between t and t+1 and they remained in the same household and relationship. Likewise, couples were defined as ‘stayers’ if neither moved and they remained partners. This procedure takes into account that individuals may not change marital status but may change partner between waves (particularly if they cohabit). Longer observation intervals for identifying a move (for example over the subsequent 2 or 3 years) were rejected due to the phrasing of the survey questions, which explicitly obtained the respondent’s moving expectations over the next year. In addition, using longer observation windows would ignore that the respondent’s expressed desires and expectations may have changed at the intervening waves. If only one partner moved or both partners moved but to different households, the couple were assumed to have separated and these person-years were omitted (see ▇▇▇▇▇, 2008b for a similar sample selection procedure). After transforming the person-year file into a couple-year format, 30,617 couple-years remained, provided by 6,675 couples over an average of 4.6 waves. The first set of cross-tabulations linked various household level independent variables to the occurrence of disagreements in moving desires between partners. To investigate the effects of disagreements on the subsequent moving behaviour of couples, random effects (panel) logistic regression models were used (▇▇▇▇▇, 2003). The dependent variable in these models is a binary variable indicating whether the household moved over the subsequent survey year (0=no move, 1=move). The control variables in these models contain lagged values, with transition variables measuring the occurrence of life events (such as changes in employment status) between the observation of moving desires at t and moving behaviour at t+1. Table 1 provides a summary of all variables used in these analyses. Panel models are valuable as they account for the non-independence of observations, as couple-year cases are nested within couples. The descriptive results presented in Table 1 and Figure 1 confirm that it is important to consider the pre-move thoughts of both partners in couples. Partners often disagree about whether a move is desired (19.11% of cases) or expected (4.36% of cases). Figure 1 shows how partner (dis)agreement on moving desires and the actual mobility rate vary with the age of the older partner in the couple. Disagreements appear to occur fairly consistently across the life course, although younger couples are more likely to disagree than older couples. While total agreement rates remain fairly stable, the composition of this agreement shifts from desiring to move to not desiring to move as age increases. It is important to note that the actual mobility rate is consistently lower than the proportion of couples where one or both partners desire to move (sum of disagree and both desire). This suggests that many people may be unable to act upon their moving desires, or that moving desires are often unattainable and hence abandoned. Table 2 presents data on the associations between partner similarity and (dis)agreement on moving desires. The results provide only weak support for the idea that partners who are demographically and socioeconomically more similar to one another are less likely to disagree about whether moving is desirable. The age gap separating partners appears unrelated to the propensity for partners to disagree about whether moving is desirable, although couples separated by the largest age gaps are slightly more likely to disagree. Ethnically mixed couples are more likely to disagree than ethnically homogenous couples, despite the idea that only more committed individuals are willing to enter into such unions. A gap in educational levels between partners seems unrelated to (dis)agreement on moving desires. Both dual and single earner couples are more likely to disagree than couples where neither partner is employed. The results in the lower section of Table 2 provide preliminary support for the contention that disagreement about whether moving is desirable is more likely when partners also disagree about the quality of their dwelling or neighbourhood. Disagreements are most likely to occur if the partners already disagree about whether they are satisfied with their dwelling or dislike their neighbourhood. Further analysis (not shown) reveals that it is almost always the partner who is unhappy with their dwelling or neighbourhood who desires to move. This suggests that individual moving desires are stimulated by personal subjective evaluations of dwelling and neighbourhood conditions (Landale and Guest, 1985). This interpretation is further supported by the strong links between shared negative evaluations (particularly of the neighbourhood) and shared desires to move: more than 96% of couples who agree on disliking the neighbourhood also share a desire to move. Couples who disagree about their housing or neighbourhood conditions also often agree that moving is desirable. This suggests that people often take their partner’s happiness with their current location into account when expressing their own moving desires. This would not be visible in an individual level study. Table 3 presents descriptive results linking the level of shared commitments to the relationship to moving desire (dis)agreements. There is somewhat mixed support for the commitment hypothesis, which proposed that possessing fewer joint commitments is associated with a greater propensity for partners to disagree about the desirability of moving. Disagreements are more likely among cohabiters than married couples, with cohabiters also much more likely to agree that moving is desirable. This may indicate that individuals select into marriage when they foresee that a shared future living arrangement is feasible, typically as relationship duration and stability increase. Disagreements also appear to be more common for couples with children, with the incidence of disagreement generally increasing with the age of the children (see also ▇▇▇▇▇▇▇▇ and ▇▇▇▇▇▇, 2009). This suggests that although families with children are fairly immobile, it is common for one or both partners to still desire to move. There is also some evidence that tenure commitments are linked to desire disagreements. Highly committed homeowning couples disagree in 18.55% of cases, whereas disagreements are slightly more common amongst renting couples (just over 21%). Again this suggests that individuals often enter committed states only when they perceive that the needs of both partners can be met through residence at a single location. Overall we have found little convincing evidence that levels of partner similarity are associated with moving desire disagreements. We did find that disagreements are most likely to occur when the partners disagree about the quality of their dwelling or neighbourhood. These findings provide initial support for conceptual models of residential mobility decision making (see Lu, 1999). There is also some support for hypothesis 2, suggesting that greater levels of commitment are associated with a reduced propensity to disagree about whether moving is desirable. Table 4 contains descriptive results testing the third and fourth hypotheses. The results indicate that taking the moving desires of both partners into account more accurately predicts whether couples subsequently move. The upper section of Table 4 links the desires of only the male partner to the couple’s moving behaviour over the next year. Ignoring the views of the female partner, these results show that 15.90% of couples where the male desires to move also actually move. The lower section of the table reveals however that the likelihood of the male partner’s desire to move being realised is heavily dependent upon the views of his partner. If only the male partner desires to move, then a move occurs in 7.57% of cases. If both partners desire to move then a move occurs in over 20% of cases. These findings support the hypothesis that moving desires are most likely to be realised if shared by both partners. This demonstrates that linking only one partner’s desires to the actual moving behaviour of the couple leads to inaccurate estimates of how strongly desires are associated with actual moves. Shared moving desires are much more likely to be realised than desires which are not shared. Table 5 contains the results from five panel logistic regression models estimating the likelihood of couples making joint moves. These models enable robust hypothesis testing, by controlling for the effects of background characteristics known to affect mobility. Our main interest is in how partner (dis)agreements in evaluations of housing and/or neighbourhood quality, moving desires and moving expectations, affect the moving propensity of couples. It is possible that interview conditions may have affected our measurements of (dis)agreements. It is likely that disagreements are less likely to be expressed if both partners are interviewed together. Further analysis revealed that partners are somewhat more likely to disagree if they completed the relevant section of the interview separately than if they were interviewed together. As partners were not interviewed separately in approximately 50% of cases, we may undercount disagreements in our analyses. To ensure that our results are robust, the models were rerun with a variable indicating the interview conditions included as an extra control. The model results were almost identical to the models without this control variable (not shown). Model 1 includes only housing dissatisfaction and neighbourhood assessments as independent variables. The model shows that couples are more likely to move if one or especially both partners are dissatisfied with their dwelling or dislike their neighbourhood. Consistent with ▇▇▇▇ and ▇▇▇▇▇▇’▇ (2010) findings, moves are more likely to occur if only the woman dislikes the neighbourhood than if only the man dislikes the neighbourhood. These parameters remain strong and significant when a range of control variables (but without moving desires and expectations) are added in Model 2. In general the control variables have the effects anticipated from the literature, apart from the negative coefficient of the cohabitation dummy (although this is not significant). The propensity to move decreases with age, and couples with children are less likely to move than those without (particularly if the children are school age or older). Changes in the number of children in the household do not appear significantly linked to mobility. High levels of education are associated with a higher probability to move, while single and particularly dual earner couples are less likely to move than couples where neither partner is employed. Interestingly, reductions in the number of people in employment are also associated with moving. This may be due to moves related to retirement. With higher levels of income the likelihood of moving increases. Private renters are more likely to move than those in other housing tenures, while room stress is also associated with a greater propensity to move. The longer people stayed in their current dwelling, the less likely they are to move. Further analyses (not shown here) demonstrate that there is little evidence of any significant regional or period effects on moving behaviour. Model 3 only includes the moving desires and expectations of both partners. The results support both hypotheses 3 and 4. Desiring to move is associated with a greater propensity to actually move, particularly if this desire is shared between partners. Shared moving expectations are very strongly linked to mobility, although moves are also likely if only expected by one partner (especially if the woman expects to move). The effects of moving desires and expectations remain stable when control variables are included in Model 4. Most of the control variable parameters are similar to those in Model 2, although there are some minor changes in significance levels (for instance education level becomes insignificant). Model 4 fits the data much better than Model 2, as shown by the considerably lower log likelihood value in Model 4. This suggests that desires and expectations are more strongly linked to actual moves than evaluations of dwelling and neighbourhood conditions. Finally, Model 5 contains all variables included in the previous models. Most of the control variables have similar effects to those estimated in the previous models. The most important finding is that some of the effects of housing satisfaction and all of the effects of disliking the neighbourhood become insignificant when desires and expectations are included in the same model. This indicates that subjective evaluations of dwelling and neighbourhood conditions are associated with moving desires and expectations, with these desires and expectations in turn conditioning the propensity to move (▇▇▇ et al., 1994; ▇▇▇▇▇, 1955). ▇▇▇▇ and ▇▇▇▇▇▇ (2010) have previously reported that whether the female partner dislikes the neighbourhood has a particularly strong effect on whether the household subsequently moves. While this is correct, our results demonstrate that this is also partially dependent on how disliking the neighbourhood affects the moving desires and expectations of both partners. Interestingly, after also controlling for moving desires and expectations, couples remainof