Inter-occupational Labour Mobility in Canada, 1994-2005: Evidence from the SLID Xuyang Chen, Maxime Fougère, and Zhengxi Lin* Policy Research Human Resources and Social Development Canada Phone: (819)934-5305/(819)953-3432/(819)994-4472 Email:
[email protected] [email protected] [email protected] April 2008
* This is one of a series of empirical studies on inter-provincial, inter-industrial, and inter-occupational labour mobility in Canada. Views expressed herein are those of the authors and do not necessarily reflect those of HRSDC. We thank Charles Beach, Erwin Gomez-Gomez, and Zhichao Wang for valuable comments. We are solely responsible for any errors remaining.
Abstract This paper empirically investigates inter-occupational mobility in Canada, one of important adjustment mechanisms through which the labour market adapts to skills and occupational imbalances, on a year-over-year basis from 1994 to 2005. In addition to documenting stylized facts, it estimates a micro-econometric model of inter-occupational mobility, in which individuals’ occupation-specific skills or human capital discount are introduced as a conceptual measure of mobility cost. The data on hand show that interoccupational mobility rates are high but with substantial variations across age groups, skills levels, and occupational groups. The regression results demonstrate that workers’ decision to change occupation strongly depends on expected wage differentials and the occupation-specific skills discount. Regarding other determinants, young adults, the more educated, Employment Insurance and Social Assistance benefits recipients are more likely to change occupation, among other things. On the other hand, job tenure, immigrant status, union membership, and marriage tend to reduce the probability of changing occupation. Key words: JEL classification:
Inter-occupational Labour Mobility in Canada, 1994-2005: Evidence from the SLID 1. Introduction There is growing concern that sustained growth in the demand for labour in specific sectors, occupations, and regions of Canada, combined with increasing global competition for skilled workers and population ageing, will lead to skills imbalances/shortages in a variety of occupations, notably in health. In unregulated sectors, these skills shortages, if materialize, would lead to increased wage pressures and thus attract workers from other groups to move into. According to the literature, inter-occupational mobility is an important adjustment mechanism (Kambourov and Manovskii 2005) through which the labour market adapts to potential skills and occupational shortages. Hence, understanding key driving forces as well as barriers to inter-occupational mobility may help to identify policy levers in promoting a more flexible labour market. There is an abundant empirical literature on inter-occupational mobility in the US. In Canada, however, the research has been highly limited. To the best of our knowledge, there is little study available using Canadian data that provides insights on trends in inter-occupational mobility, on a year-over-year basis. Using the Survey of Labour and Income Dynamics (SLID) from Statistics Canada, this paper’s objectives are twofold. First, it documents stylized facts on interoccupational mobility in Canada over the past decade at a fairly disaggregate level. This helps to investigate inter-occupational mobility patterns, profile of inter-occupational movers, and economic returns to inter-occupational mobility. Secondly, it develops and estimates a micro-econometric model of interoccupational mobility. By empirically estimating effects of potential factors (personal, family, economic, social, and cultural) at the individual level, our model examines chiefly the effect of wage differentials and skills levels on individual's decision to change occupations, among other things. Data on hand show that inter-occupational mobility rates are high but with substantial variations across age groups, skills levels, and occupational groups. In general, younger workers, lower skilled workers and non-immigrants are more occupationally mobile. Among occupational groups, workers in Arts, Recreation and Sports are most likely to change occupations in contrast to health workers who are the least likely to change occupations. Economic returns to inter-occupational mobility are substantial --- movers enjoy wage increases more than twice of that for non-movers. Regression results suggest that expected wage differentials are a key driver of interoccupational mobility. Occupation-specific skills have a significant negative influence on inter-occupational mobility. Regarding other determinants, higher level of education increases the chance an individual changes occupation, while immigrant status, union membership, job tenure, and marriage reduce the probability of changing occupation.
-1-
This paper contributes to the literature in two ways. First, we present stylized facts that characterize inter-occupational mobility in Canada on a year-over-year basis, which to our knowledge has not been done with Canadian data before. Second, we introduce individual’s occupation-specific skills or human capital discount as a conceptual measurement of mobility cost and infer that the occupation-specific skills gaps are wider as skill-level increases, and the human capital loss or skills discount may be larger for higher-skilled occupational movers. The rest of the paper proceeds as follows. Section 2 briefly reviews the literature on inter-occupational mobility. Section 3 presents stylized facts on inter-occupational mobility, including patterns of different age groups, skills levels, gender and occupational groups, transition probability matrices between different occupation groups, and economic returns to occupational movers and stayers. Section 4 presents the modelling approach and discusses the estimation results. And Section 5 closes the paper with some concluding remarks. 2. Inter-occupational mobility: What do we know from the existing literature? There is an extensive literature on inter-occupational mobility. As mentioned earlier, however, most of these studies use US data, with one notable exception. We hereby summarize key findings from the existing literature. 2.1 Inter-occupational Mobility as a Labour Market Adjustment Mechanism According to Kambourov and Manovskii (2006) and Markely and Parks II (1989), the extent to which inter-occupational mobility acts as an adjustment mechanism in the labour market might be considerably large. Using the Current Population Survey and based on a question asked of individuals on inter-occupational mobility, Markely and Parks II (1989) find that about 10% of workers in the United States changed occupations in 1986. Kambourov and Manovskii (2006) use the U.S. Population Survey of Income Dynamics and find that the inter-occupational mobility rate is about 13% at one-digit and 20% at the three-digit level during the 1990s. Several studies, such as Robertson and Symons (1990), Markely and Parks II (1989), and Miller (1984), explore the characteristics of occupational movers. Overall, the studies find that certain demographic characteristics, like education and family antecedents play an important role in determining occupational choices. These findings also indicate that human capital investment is occupation-specific and that individuals tend to change occupations to maximize the value of their human capital investment. 2.2 Determinants of Inter-occupational Mobility Based on the neoclassical theory (Sjaastad 1962), individuals treat mobility as an investment in human capital in a rational cost-benefit analysis. A few studies have examined determinants of inter-occupational mobility at the individual level (Kambourov and Manovskii 2006, Parrado and Wolff 1999, and Bojas 1981). Their findings suggest
-2-
that occupational changes are associated with wages and tenure. They also find that younger individuals are more likely to switch occupations. Finally, previous studies that have focused on the earnings profile or earnings trends do not find significant costs or barriers to inter-occupational mobility.1 In Canada, to the best of our knowledge, most empirical studies on labour mobility have focused on regional labour mobility. One exception is Green (1999) who examines inter-occupational mobility for immigrants and non-immigrants. Using Census data, he compares inter-occupational mobility patterns between immigrants and the native-born and concludes that immigrants are more occupationally mobile and hence contribute to a more flexible labour market. 3. Inter-occupational mobility: stylized facts 3.1 Data Source In this paper, we use the Survey of Labour and Income Dynamics (SLID) from Statistics Canada as source of data. The SLID collects survey information on income and labour market activity and provides longitudinal data on individuals over time. More specifically, the SLID provides disaggregated occupational classification code, and extensive information on family situation, education and demographic background. The survey also provides a whole range of information on transitions, durations, and repeat occurrences (longitudinal) of people's financial and work situations, which is very helpful to understand the process and determinants of occupational choices of workers. We focus on individuals aged 16-69, who worked in any two consecutive years between 1994 and 2005. All occupations are classified within four skills levels based on the National Occupational Classification (NOC) system. 3.2 Inter-occupational Mobility Patterns Figure 1 shows the trend of inter-occupational mobility in Canada between 1994 and 2005. From the NOC, we consider that an individual switches occupation if his/her valid occupational code of main job at the end of the current year is different from the end of the previous year. The results indicate that at the four-digit level, the rate of interoccupational mobility increased substantially between 1995 and 2000, declined sharply in 2001 and 2002, then remained flat thereafter. Overall, about 20% of Canadian workers change occupation at the four-digit level. This decreases to 13% when we look at onedigit level data.2 This implies that a significant proportion of movers have changed occupation within large occupational groups. These mobility rates are very much in line with findings for the United States (there is no existing Canadian literature currently available for comparison).3 1
Parrado and Wolff (1999) and Bojas (1981) find that tenure is associated with a lower probability of changing occupations, but the effect is negligible. 2 See appendix for detailed information of digit level in National Occupational Classification (NOC). 3 For example, see Kambourov and Manovskii (2005).
-3-
Figure 1 Inter-occupational Mobility Rates in Canada, 1994-2005 25%
Four-digit Level
20%
15%
10%
One-digit Level
5%
0% 1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Source: Authors’ calculation from SLID.
3.3 Inter-occupational Mobility Rates by Gender Compared with male workers, female workers had slightly higher interoccupational mobility rates at the four-digit level over the past decade. However, the story is different at the one-digit level. Before 2000, male workers seemed somewhat more likely to change occupational groups than women, suggesting that female occupational movers may prefer more vertical moves within large occupational groups while male movers are more willing to make horizontal moves across occupational groups. In fact, as we will see later in the empirical results, after controlling for other factors, our regression analysis suggests that the effect of gender is not statistically significant on the decision to change occupation. Figure 2 Inter-occupational Mobility Rates by Gender 0.3
Four-digit Level Mobility Rate
One-digit Level Mobility Rate 0.2
0.25
Female
Male 0.15
0.2
Female
Male
0.1
0.15 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Source: Authors’ calculation from SLID.
-4-
0.05 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
3.4 Inter-occupational Mobility Rates by Age Group Overall, younger workers have higher inter-occupational mobility rates at the fourdigit level. On average, workers under 25 years of age are three times more mobile than workers over 45 years of age. As Figure 3 shows, the widest gap in inter-occupational mobility rate is between age group under 25 and 25+. The gap between older age groups narrows quickly as age increases. This likely reflects the fact that younger people are usually not in a permanent work situation and hence are much more occupationally mobile. Further, if we treat inter-occupational mobility as a human capital investment, older workers face a relatively shorter time period to realize their returns.4 Accordingly, they are likely more reluctant to change occupation. In fact, since the work of Byrne (1975), the negative relationship found between inter-occupational mobility and age in the literature “has been deemed as a socioeconomic law”5. However, this negative relationship is observed in isolation without taking other factors into consideration. As such, it must be interpreted with care, which will be achieved when we turn to multivariate regression analysis subsequently. Figure 3 Inter-occupational Mobility Rate by Age Group 60% Under 25
50% 40% 25-34
30% 35-44
20% 10%
45+
0% 1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Source: Authors’ calculation from SLID.
4
Theoretically, the time horizon used in Bojas’ (1990) framework may also be explained as age/mobility relationship. 5 From Markely and Parks II (1989), they argue that age is the most salient determinant of voluntary interoccupational mobility.
-5-
3.5 Inter-occupational Mobility Rates by Skills Level Based on the NOC, all individual workers in the sample are assigned a skills level according to their previous occupational code. At the four-digit level, there are 521 occupations and they can be reclassified within four skills levels (A0, B, C and D) and ten large occupational groups or skills types (one -digit level)6. Compared with education attainment, the skills level is a more suitable measure of an individual’s current jobspecific human capital and is well recognized by employers. However, we are aware that a proportion of workers might be either under- or over-qualified. Figure 4 shows inter-occupational mobility rates by skills level. In general, lowerskilled workers have a higher inter-occupational mobility rate. This is possibly because the skills gap between low-skilled or unskilled occupations is much narrower. Moreover, skills or human capital are occupation-specific, so moving across higher skills levels may be related to a higher human capital discount or mobility cost which may reduce individuals’ incentives to change occupations. Finally, as the opportunity cost of moving is lower for lower-skilled workers, they are more likely to move. Figure 4 Inter-occupational Mobility Rates by Skills Level 40%
Level D
35% 30%
Level C
25% 20%
Level B
15% Level A0
10% 5% 0% 1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Source: Authors’ calculation from SLID.
6
A0 refers to management and occupations usually requiring university education, B refer to occupations usually requiring college education or apprenticeship training, C refers to occupations requiring secondary school or occupation-specific training and D refers to occupations that on-the-job training is usually provided.
-6-
3.6 Inter-occupational Mobility Rates by Immigrant Status Immigrants as a whole have significantly lower inter-occupational mobility rates than the native-born. As seen in Figure 5, the overall mobility rate among immigrants is about three quarters of that for the native-born. This is consistent with Lin (1996) who finds that immigrants are less regionally mobility, but conflicts with findings from Green (1999)7. However, Green uses Census data and his conclusion is based on a definition of mobility after five years, which is long enough for multiple mobility or return mobility to happen. Moreover, Green’s study focuses on male immigrants during the 1980s and our sample is for both male and female in the period of 1994 to 2005. A larger proportion of immigrants who came to Canada more recently are more highly-skilled, and accordingly likely less occupationally mobile. Finally, we acknowledge that inter-occupational mobility may be very different between recent and other immigrants.8 However, in this paper we only discuss the effect of immigrants as a whole on the decision to change occupation. Figure 5 Inter-occupational Mobility Rates by Immigrant Status 30% Non-immigrant
25% 20% 15%
Immigrant 10% 5% 0% 1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Source: Authors’ calculation from SLID.
3.7 Inter-occupational Mobility Rates across Occupational Groups Table 1 summarizes the inter-occupational mobility rate across occupational groups (one-digit level). Each entry in Table 1 shows the percentage of workers in an occupational group who move to other groups in the second year. Arts, Recreation 7
Green (1999) uses the Census to compare immigrant and native-born male occupational distributions in Canada in the 1980s, and he finds that immigrants are more occupational mobile than native-born. 8 Recent immigrants usually refer to those who migrated to Canada less than five years.
-7-
&Sports workers are most likely to change occupation. On average, they record more than 20% mobility rate, followed by individuals working in primary industry and processing/ manufacturing/utility workers. Conversely, health workers are least likely to change occupational group, with about 6% overall mobility rate. It is well known that jobs in health occupations are very skills-specific and highly regulated. It is also known that skills shortages are more likely to occur within this occupational group. Table 1 Inter-occupational Mobility Rates across Occupational Groups, 1994-2005 Occupational Group
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Management
11.9
14.4
14.5
12.7
13.9
17.4
12.0
12.3
9.9
12.1
12.3
Business&Finance
11.5
11.4
13.1
13.2
13.1
12.4
11.6
10.4
9.6
10.7
11.2
Natural&Applied Science
11.5
15.1
14.9
12.6
14.7
12.9
10.8
7.2
10.4
8.1
9.8
Health
5.0
7.2
8.3
8.3
4.0
8.3
4.8
5.8
5.7
4.4
5.9
Social Science,Edu., Gov. & Religion
8.0
10.8
12.4
14.9
11.1
11.1
10.0
10.0
9.8
9.5
7.9
Arts,Recreation &Sports
22.7
26.0
22.0
24.8
23.1
22.1
21.6
24.7
17.5
20.1
16.8
Sales and Service
14.9
14.4
15.5
15.2
14.9
16.2
14.9
14.1
13.4
14.1
13.2
Trade,Transport and Equipment ope.
10.7
12.3
11.1
11.4
11.6
11.1
9.6
9.3
9.4
9.3
9.3
Primary Industry
17.4
18.6
17.0
21.6
17.7
19.6
14.3
14.3
16.5
14.9
14.6
Processing, Manufacturing&Utilities
17.0
19.5
14.8
14.3
16.2
15.9
16.5
15.2
15.6
13.8
12.6
Source: Authors’ calculation from SLID.
Turning to patterns of occupational switches across these groups, we observe that occupational movers are unevenly distributed. Table 2 reports detail destinations and the transition probability for 1995. It can be seen that Sales and Services occupations are the most likely destination for movers from other occupational groups. On average, this group absorbs more than 30% of total occupational movers from other occupational groups. In contrast, health occupations likely have greater barriers, as workers are least likely to move to this group. Workers in this area usually need local licenses or qualification. Obtaining a license is costly and also time consuming. We have done a similar analysis each year over the sample period and the overall story does not change noticeably. So, without losing generality, we only present results for the first year.9 Table 2 Probability Transition Matrices of Occupational Mobility, 1994-199510
9
Please see appendix for 1999-2000 and 2004-2005. Results for other years are available upon request. Occupational groups: 0 - Management; 1- Business, finance and administration; 2- Nature and applied science; 3- Health; 4- Social science education government and religion; 5- Art, culture, recreation and sport; 6- Sales and service; 7- Trade, transportation and equipment operation; 8- Primary industry; 9 Processing, manufacturing and utilities.
10
-8-
From/To
0
1
2
3
4
5
6
7
8
9
0
88.13
2.37
1.11
0.65
1.2
0.52
4.02
1.24
0.09
0.68
1
1.94
88.46
1.27
0.3
0.94
0.33
4.81
1.17
0.44
0.35
2
2.48
1.99
88.5
0.72
0.68
0.51
2.08
1.66
1.07
0.28
3
0.55
1.35
0.23
95.02
0.74
0.04
1.67
0
0.19
0.21
4
0.82
1.6
0.13
0.58
92.02
0.94
2.8
0.17
0.24
0.69
5
1.83
4.95
1.61
0.22
3.47
77.26
7.3
2.65
0.47
0.24
6
2.61
4.4
0.87
0.75
0.93
0.52
85.15
2.1
0.92
1.76
7
0.99
1.2
0.83
0.18
0.49
0.13
2.89
89.28
1.55
2.45
8
0.88
1.35
0.26
0.19
0.33
0.49
5.23
5.46
82.58
3.22
9
0.36
1.74
1.72
0.12
0.19
0.53
5.25
5.68
1.36
83.05
Source: the Survey of Labour and Income Dynamics (SLID) 1994-2005
3.8 Economic Returns to Inter-occupational Mobility Moving from one occupation to another can pay off sweetly. As seen in Figure 6, wage increases of inter-occupational movers are more than twice of that for non-movers. On average, inter-occupational movers' nominal hourly wage from the main paid job increases by 17%, while the corresponding hourly wage increase for stayers is only about 7%.11 Another interesting finding is that between 1995 and 2005, the gap in hourly wage increase has widened between movers and stayers. This finding confirms that individuals tend to change occupations for better economic returns. It also suggests that expected wage differentials between occupations are a key driving force to determine interoccupational mobility. Figure 6: Hourly Wage Increase for Movers and Stayers
11
Hourly wage increase for stayers is similar to finding in literature using the Labour Market Activity Survey (LMAS), for example, Lin (1996).
-9-
2 5% 2 0%
Move rs
1 5% 1 0% 5%
S taye rs
0% 19 95
1 99 6
19 9 7
1 998
1 99 9
20 0 0
2 0 01
2 0 02
20 0 3
2004
2 0 05
Source: Authors’ calculation from SLID.
Figure 7 shows hourly wage increases for movers and stayers by skills levels. It is clear that low-skilled workers have notably higher average economic returns to mobility than high- and medium-skilled workers. This is consistent with the perception in the literature that low-skilled workers change occupations for higher income, and highskilled workers change occupations more for challenge. Moreover, if wage increase is the key driving force of inter-occupational mobility, this may partly explain why low-skilled people have higher mobility rate than high-skilled workers. Figure 7 Hourly Wage Increase for Movers and Stayers by Skills Levels Skill Level A0
25% 20%
Skill Level B
25% 20%
Movers
15%
Movers
15%
10%
10%
Stayers 5% 0% 1995
Stayers
5%
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Skill Level C
25%
0% 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Skill Level D
25%
Movers 20%
Movers
20% 15%
15% 10%
10% Stayers
Stayers
5%
5%
0% 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
0% 1995
Source: Authors’ calculation from SLID.
- 10 -
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
We have thus far itemized some stylized facts that are helpful in visualizing interoccupational mobility patterns and trends as well as the substantial variations across age groups, skills levels, and occupational groups. However informative these observations may be, they are examined in isolation, independent of one another. However, it has been shown that a host of factors interdependently affect the decision to change occupation. To better understand the inter-occupational mobility decision making process and identify key driving forces (both conducive and obstructive), we now turn to econometric modelling through which impacts of all observable factors are properly controlled for. 4. Determinants of inter-occupational mobility 4.1 Theoretical Framework of Inter-occupational Mobility Assuming that individuals view inter-occupational mobility as a human capital investment and under a rational cost-benefit analysis, it can be predicted that an individual tends to change from occupation o to occupation d if the expected wage gain associated with the change exceeds the cost of moving. Following Borjas (1990), the expected returns to mobility ER(m) can be expressed as: (1)
ER ( m ) =
∫
[ E (Y
d
( t ) − Y o ( t )) ] e
− rt
dt − COM
,
where Yo and Yd are wages of origin and destination, r the discount rate, t the time horizon and COM the cost of mobility. The first part of the equation on the right-hand side represents the benefit, while the second part the cost of mobility. The basic criterion for mobility is ER(m)>0. Identifying the cost of mobility is always a challenge in empirical studies. Different from the cost of regional mobility, which can be proxied by the moving distance or relative housing price, the cost of inter-occupational mobility usually refers to the opportunity cost. The literature suggests that tenure can be one possible measure to use, because people may lose their seniority when they change occupation. For example, Parrado and Wolff (1999) and Borjas (1981) find that tenure is associated with a lower probability of changing occupations, but their finding also indicates that this effect is quite small. Some studies, for example, Robinson and Tomes (1982) attempt to use locationspecific skills loss as the measurement of moving cost in their inter-provincial model. Kambourov and Manovskii (2005) argue that human capital is largely occupationalspecific, and that a substantial amount of this capital can be destroyed upon switching occupation. However, both studies fail to find a convincing measurement to quantify the effect and their finding is far from conclusive. In this paper, we introduce the individual’s occupation-specific skills or human capital discount as one conceptual measurement of mobility cost. An occupation-specific skill refers to skills required to work in a particular occupation, which is obtained either through specialized education, training or work experience. When an individual switches from occupation A to occupation B, the non-transferable part of skills, or specific skills
- 11 -
required to work in occupation A, but not in occupation B is considered the unused/ underused human capital or human capital loss. Similarly, the skill gap between occupation A and B refers to skills specifically needed for occupation B but not for occupation A. Our hypothesis is that occupation-specific skills gaps are wider as skill level increases and the human capital loss or skills discount is larger for higher-skilled occupational movers than for lower-skilled occupational movers. If this is true, the more skilled an individual is, the lower the probability for this individual to change occupation. Of course, we also keep tenure to jointly measure the mobility cost. Inter-occupational mobility may not only be motivated by financial factors but also influenced by personal interest. Therefore, in addition to expected wage gains, skills levels and tenure, we also add a set of personal and job characteristics, as well as dummy variables indicating whether the individual was an Employment Insurance (EI) or Social Assistance (SA) beneficiary prior to the move.12 Therefore, the model can be explicitly expressed as: (2) Mi = ƒ (Δ Wagei , Skill Leveli ,Genderi, Unioni, Educationi, Agei, Childi, Marriagei, EIi, SAi, Mother Tonguei, Tenurei, Immigranti, Yeari), Mi denotes the dependent variable which takes on the value of 1 if an individual’s occupation differs in two adjacent years and 0 otherwise. 4.2 Estimated Wages To build such a model of choice among occupations, we first need to estimate the expected wage for occupational movers and stayers and calculate the wage differentials to be used as an explanatory variable in the regression. In the estimation procedure, wage estimates obtained from the sample of movers and stayers may be biased due to selectivity problems (Heckman 1976). To correct for the selectivity problem, we adopt the Heckman two-step estimation method. We choose hourly wages of the individual’s main job rather than total earnings since inter-occupational mobility in this paper refers to the individual’s main job and earnings may combine income from multiple jobs, including income prior to and after mobility in the same year. We focus on age 25 and older. We exclude workers under 25 in our estimation process because of uncertainty in their inter-occupational mobility behaviour. We also restrict the sample to those who report valid hourly wage in two consecutive years.13 Table 3 reports the results of estimated wages for inter-occupational movers and stayers. The results from different years are very similar. Without particular preference, we only present results for 1999 and 2000 in the text. Results for other years are reported in the Appendix.
12 13
Please see appendix for description of explanatory variables. We exclude those unreasonable wage data, for example, hourly wage less than five dollar an hour.
- 12 -
Table 3 Expected wages for inter-occupational movers and stayers, 1999-2000 Dependent Variable: Hourly Wage V a ria b le W a g e (-1 ) Sex U n io n U n iv e rsity P o st S e c o n d a ry S k illa 0 S k illb S k illc A ge35_44 A ge45+ Im m ig ra n t F u ll T im e J o b D u ra tio n C o n sta n t Lam bda N 1. 2.
M o v e rs 0 .8 9 9 (0 .1 6 7 )* * * 1 .5 8 1 (0 .1 8 9 )* * * -0 .0 8 8 (0 .2 4 0 ) 3 .4 6 0 (0 .2 9 8 )* * * 1 .0 6 4 (0 .2 0 6 )* * * 1 .4 9 3 (0 .3 7 9 )* * * 0 .5 3 1 (0 .3 3 0 )* 0 .1 1 7 (0 .2 8 9 ) 0 .1 2 7 (0 .2 1 5 ) 0 .2 2 3 (0 .2 5 3 ) 0 .0 3 9 (0 .3 1 1 ) -0 .3 8 3 (0 .2 6 2 ) 0 .0 0 0 (0 .0 0 2 ) 1 .7 3 3 (0 .8 3 6 )* * 0 .8 4 9 (0 .8 5 0 ) 3004
S ta y e rs 0 .9 6 5 (0 .0 0 6 )* * * 0 .8 9 3 (0 .0 8 6 )* * * -0 .0 8 4 (0 .0 9 3 ) 1 .5 4 9 (0 .1 3 7 )* * * 0 .2 9 6 (0 .0 9 2 )* * * 1 .9 9 4 (0 .1 7 9 )* * * 0 .9 8 9 (0 .1 6 3 )* * * 0 .4 8 3 (0 .1 5 1 )* * * 0 .2 1 3 (0 .1 0 5 )* * 0 .2 0 2 (0 .1 1 5 )* 0 .0 2 8 (0 .1 2 7 ) 0 .0 5 1 (0 .1 2 5 ) -0 .0 0 1 (0 .0 0 1 ) 1 .3 5 0 (0 .4 2 9 )* * -1 .1 6 9 (0 .4 2 9 )* * 12800
Standard error in parenthesis. * Significant at 10 percent; ** significant at 5 percent; and *** significant at 1 percent.
As seen in Table 3, wages for both stayers and movers have a strong positive relationship with their wages in the previous year. Wage increases for males are higher than for females for both stayers and movers. Also, as expected, higher educated and higher skilled workers can expect higher raise in the second year regardless of staying or moving. Other personal and old job characteristics do not contribute to explain the wage of movers. From these estimation results, we have calculated expected wage differentials between movers and stayers that we incorporate in our logit model of equation (2). 4.3 Logit Estimated Results In this section we model the determinants of the probability of inter-occupational mobility at the four-digit level. We pool all data from 1995 to 2005 and exclude all duplicate observations in the final sample. The SLID sample is composed of two panels
- 13 -
of respondents who are surveyed annually for a six-year period. A new panel is introduced every three years, so two panels always overlap. Our final sample consists of 45,943 adults after some exclusion. Table 4 reports the results of the Logit model on inter-occupational mobility for the period 1994 to 2005. All important variables have the expected sign and the results confirm most of our findings from the stylized facts discussed in the previous section. Wage differentials are positive and statistically significant. Conversely, there is strong evidence that skills levels have a negative influence on inter-occupational mobility. This confirms our assumption that skills discount or mobility costs may be larger for higherskilled occupational movers, and hence reduce the probability of mobility.
Table 4 Logit model of probability of inter-occupational mobility, 1994 to 2005 V a ria b le W a g e D iffe re n tia l Sex U n io n U n iv e rsity P o st S e c o n d a ry S k illa 0 S k illb S k illc A ge35_44 A ge45 Im m ig ra n t M a rry C h ild J o b D u a tio n A tte n d U n iv e rsity F re n c h 1. 2.
0 .1 5 5 (0 .0 2 4 )* * * 0 .0 3 0 (0 .0 2 6 ) -0 .3 0 7 (0 .0 3 1 )* * * 0 .1 0 6 (0 .0 5 3 )* * 0 .0 5 5 (0 .0 3 3 )* -0 .4 9 4 (0 .0 5 1 )* * * -0 .2 9 4 (0 .0 4 4 )* * * -0 .2 1 9 (0 .0 4 0 )* * * -0 .0 6 6 (0 .0 3 1 )* * -0 .1 3 3 (0 .0 3 6 )* * * -0 .1 5 2 (0 .0 5 6 )* * * -0 .1 5 6 (0 .0 3 7 )* * * -0 .0 0 7 (0 .0 3 1 ) -0 .0 0 7 (0 .0 0 0 )* * * 0 .5 4 0 (0 .0 5 8 )* * * 0 .0 1 2 (0 .0 6 2 )
V a ria b le O th e r T o n g u e S o c ia l A ssista n c e EI Y 1994 Y 1995 Y 1996 Y 1997 Y 1998 Y 1999 Y 2000 Y 2001 Y 2003 Y 2004 C o n sta n t N o f o b se rv a tio n s N o f D e p . V a r.= 1 L o g -lik e ly h o o d P e rc e n t c o n c o rd e n t
0 .1 3 7 (0 .0 5 6 )* * 0 .2 6 3 (0 .0 6 7 )* * * 0 .4 3 2 (0 .0 3 0 )* * * 0 .1 9 0 (0 .0 6 1 )* * * 0 .3 9 1 (0 .0 7 2 )* * * 0 .6 1 9 (0 .0 7 1 )* * * 0 .7 6 6 (0 .0 7 4 )* * * 0 .7 3 1 (0 .0 8 9 )* * 0 .7 8 8 (0 .0 7 2 )* * * 0 .1 1 6 (0 .0 6 2 )* -0 .0 3 9 (0 .0 8 0 ) 0 .0 1 5 (0 .0 6 7 ) 0 .0 0 0 (0 .0 9 2 ) -1 .1 6 (0 .0 9 4 )* * * 45943 8118 -1 9 2 4 3 .8 7 2 .8 %
Standard error in parenthesis. * Significant at 10 percent; ** significant at 5 percent; and *** significant at 1 percent.
- 14 -
There is no indication that the probability of inter-occupational mobility is different between males and females. Compared with elementary schooling, higher educated workers have a higher probability to change occupation.14 Being immigrant, marriage and job duration reduce the probability of mobility, while people attending university or under Employment Insurance or Social Assistance are more likely to change occupations. We also calculate marginal effects for the statistically significant variables. Table 4 shows that the marginal effect of the wage differentials on the decision to change occupation is relatively important. For example, a $1.00 increase in wage differential raises the probability of an individual to change occupation by 2 percentage points. Skills levels also have large effects. High-skilled workers (skill level A0) are more than five percent less likely to change occupations than low-skilled workers (skill level D). Compared with the skills level effect, a one-month increase in job duration does not increase individuals’ loyalty to their occupation by much. Although the variable is statistically significant, its marginal effect remains very small, less than one-tenth of a per cent, which is consistent with findings from the literature. After controlling for all other factors, the probability for an individual aged 35 to 44 to change occupation is less than one percent lower than that of individuals aged 25 to 34. Similarly, compared with age group 35 to 44, the probability decrease for age 45 and up is equally small.
Table 5 Marginal effects of significant variables V ariable W agedif
C hange in prob. V ariable 0.019 A ge35_44
C hange in prob. -0.0079
($1 ch ange)
U nion
-0.0359
A ge45
-0.0157
U niversity
0.0131
Im m igrant
-0.0175
Post Secondary
0.0067
M arry
-0.0194
Skilla0
-0.0545
Job D uration
-0.0008
(1 m onth change)
Skillb
-0.0341
A ttend U niversity
0.0772
Skillc
-0.0256
Social A ssistance
0.0346
EI
0.0570
14
Due to the fact of mismatch, education may lead to confusable inference without controlling for skills levels. For example, Parrado and Wolff 1999 conclude that there is no clear effect from schooling.
- 15 -
Finally, using the fitted logistic regression model, Figure 8 examines the model fit by comparing observed occupational movers and estimated expected movers frequencies within each decile of risk, defined by fitted value (probility) for Mi = 1.15 Overall, the model performs quite well and records a 72.8% concordance rate.
Figure 8 Model Performance Check A c t u a l v s P r e d ic t e d O u t c o m e s 6 0 % 5 0 % 4 0 % 3 0 % 2 0 % 1 0 % 0 %
1
2
3
4
5
6
7
8
9
1 0
M o d e l S c o r e D e c ile P r e d ic t e d
A c tu a l
A ve ra g e
5. Summary and conclusion There has been little work done on inter-occupational mobility in Canada. In this paper, we present stylized facts that clearly characterize inter-occupational mobility in Canada based on a relative large sample, and develop and estimate econometric models of inter-occupational mobility from 1994 to 2005. The data on hand shows that occupational mobility rates in Canada are quite high throughout the study period and at levels similar to the United States. Behind this overall trend, there are substantial variations across age groups, skills levels, and occupational groups. In general, older workers, higher-skilled workers and immigrants are less mobile inter-occupationally. Among occupational groups, workers in Arts, Recreation and Sports are most likely to change occupations in contrast to Health workers who are the least likely to move to other occupations. Measured in wage changes, economic returns to inter-occupational mobility are substantial. Our econometric analysis results are largely consistent with economic theory and empirical findings in the literature. First, as expected, wage differentials are found to be a key driver of inter-occupational mobility. Secondly, we introduce the individual’s occupation-specific skills or human capital discount as a conceptual measurement of mobility cost, and our results support the hypothesis that human capital loss or skills 15
We separate all observations into ten deciles with 4594 observations in each, and rank them from one to ten according to their estimated probability of Mi = 1 (mobility is yes), and compare the expected and the actual frequency of mobility.
- 16 -
discount may be larger for higher-skilled workers, and hence is negatively associated with inter-occupational movers. That is, other things being equal, inter-occupational mobility decreases with skills levels. Regarding other determinants, higher level of education increases the chance for an individual to change occupation, while age, immigrant status, union membership, job tenure and marriage reduce the probability of changing occupation. To close, labour market flexibility/adaptability is not only a feature of but also required by the highly innovative and internationally competitive knowledge-based global economy. Being one of labour market adjustment mechanisms, the role of interoccupational mobility in creating/promoting a flexible/adaptable labour market is getting increasingly important. This paper represents an attempt to empirically better understand inter-occupational labour mobility in Canada, which has been rarely done thus far. Our results demonstrate that market forces (i.e., inter-occupational wage differentials and human capital discount) are key drivers of inter-occupational labour mobility. These imply that the labour market, to a large degree, is capable of adjusting to changes that require the constant reallocation of human resources from areas of how demand to that of high demand. Our results also reveal that some occupational groups (most notably, health) that are widely believed to be facing potential labour/skills shortages are also the least likely to attract entry of workers from other groups. This may suggest the existence of institutional and/or human capital/skills-specific barriers. This implies that profession-targeting public policies may be beneficial in promoting greater labour market flexibility/adaptability. Reference Bijak J. (2006), “Forecasting International Migration: Selected Theories, Models, and Methods”, Central European Forum for Migration Research Working Paper, 4/2006. Borjas, G. (1990), “Friends or Strangers: The Impact of Immigrants on the US Economy”. Basic Books, New York. Borjas, G. (1981), “Job Mobility and Earnings over the Life Cycle,” Industrial and Labor Relations Review, Vol. 34, pp. 377-388. Dhrymes P. (1973), “Small Sample and Asymptotic Relations between Maximum Likelihood and Three Stage Least Squares Estimation” Econometrica 41 357-364. Green D. A. (1999) “Immigrant Occupational Attainment: Assimilation and Mobility over Time”,Journal of Labor Economics, Vol. 17, No. 1, 49-79. Harris J. and M. Todaro (1970), “Migration, Unemployment and Development: A TwoSector Analysis”, the American Economic Review, Vol. 60, No.1, 126-142.
- 17 -
Heckman J. (1976), “The Common Structure of Statistical Models of Truncation, Sample Selection and Limited Dependent Variables and a Simple estimation for Such Models”, Annals of Economic and Social Measurement 5, 397-433. Kambourov, G. and I. Manovskii (2004), "Occupational Mobility and Wage Inequality". IZA Discussion Paper No.1189. Kambourov, G. and I. Manovskii (2005), “Rising Occupational and Industry Mobility in the United States: 1968-1997”. IZA Discussion Paper No. 1110 Lin, Z. (1996), “Foreign-Born vs Native-Born Canadians: A Comparison of Their InterProvincial Labour Mobility”. 11F0019MPE No. 114, Ottawa: Statistics Canada. Markey, J. P. and W. Parks II (1989), “Occupational Change: Pursuing a different kind of Work,” Monthly Labor Review, September. Miller, Robert A. (1984), “Job Matching and Occupational Choice,” The Journal of Political Economy, Vol. 92, No. 6. (Dec.), 1086-1120. Mincer, J. and J. Boyan (1981), “Labour Mobility and Wages,” in Studies in Labor Markets, ed. S. Rosen, Chicago: University of Chicago Press, pp. 21-63. Osberg, L., D. Gordon, and Z. Lin (1994), “Inter-Regional Migration and InterIndustry Labour Mobility in Canada: A Simultaneous Approach”, Canadian Journal of Economics, vol. 27, no. 1, pp. 58-80. Robertson, D. and J. Symons (1990), “The Occupational Choice of British Children,” Economic Journal, Vol. 100, pp. 828-41 Robinson C. and N. Tomes (1982) “Self-selection and Inter-provincial Migration in Canada”, Canadian Jounal of Economics, Vol. 15, no. 3, 474-502. Sjaastad, L. A. (1962). “The costs and returns of human migration”. Journal of Political Economy, 70(5): 80–93. Appendix: Table 1: Variables and definitions
- 18 -
V a ria b le D e p e d e n t V a ria b le W a g e D iffe re n tia l Sex U n io n U n iv e rsity P o st S e c o n d a ry S k illa 0 S k illb S k illc A ge35_44 A ge45 Im m ig ra n t M a rry C h ild J o b D u a tio n A tte n d U n iv e rsity F re n c h O th e r T o n g u e S o c ia l A ssista n c e EI
D e fin itio n = 1 if c h a n g e o c c u p a tio n a t fo u r-d ig it le v e l = e x p e c te d w a g e o f m o v e rs -e x p e c te d w a g e o f sta y e rs = 1 if M a le = 1 if u n io n m e m b e r = 1 if e d u c a tio n > = u n iv e rsity = 1 if u n iv e rsity > e d u c a tio n > h ig h sc h o o l = 1 if sk ill le v e l = " A 0 " = 1 if sk ill le v e l = " B " = 1 if sk ill le v e l = " C " = 1 if 4 5 > a g e > = 3 5 = 1 if a g e > = 4 5 = 1 if im m ig ra n t = 1 if m a rrie d = 1 if h a v in g c h ild /c h ild re n = m o n th s o f e m p lo y m e n t in c u rre n t jo b = 1 if c u rre n tly a tte n d in g u n iv e rs ity = 1 if m o th e r to n g u e is F re n c h = 1 if m o th e r to n g u e is n o t E n g lish o r F re n c h = 1 if re c e iv e S o c ia l A ssista n c e b e n e fit = 1 if re c e iv e e m p lo y m e n t in su ra n c e b e n e fit
Table 2: Probability Transition Matrices of Occupational Mobility, 1999-2000 From/To
0
1
2
3
4
5
6
7
8
9
0
82.29
5.77
1.47
0.58
1.30
1.27
5.11
1.50
0.31
0.41
1
2.43
87.55
1.23
0.47
1.16
0.56
4.48
1.24
0.25
0.63
2
1.36
4.14
87.24
0.15
0.90
1.12
1.79
1.75
0.67
0.89
3
1.08
1.07
0.63
92.00
2.05
0.12
2.49
0.20
0.02
0.34
4
1.77
2.83
0.96
0.30
88.93
0.74
3.59
0.27
0.33
0.28
5
1.50
3.44
2.19
0.49
2.69
78.17
8.47
1.87
0.63
0.55
6
2.29
4.67
1.21
1.12
1.23
0.87
83.89
2.37
0.70
1.65
7
1.00
1.42
1.06
0.11
0.45
0.35
3.23
88.66
1.21
2.50
8
1.85
1.33
1.28
0.41
0.90
0.58
4.84
5.73
80.31
2.77
9
0.50
2.34
1.18
0.25
0.31
0.33
4.44
5.83
1.12
83.68
Table 3: Probability Transition Matrices of Occupational Mobility, 2004-2005
- 19 -
From/To
0
1
2
3
4
5
6
7
8
9
0
87.85
3.69
1.60
0.11
0.68
0.25
3.65
1.29
0.15
0.84
1
1.84
88.79
0.82
0.51
1.04
0.48
4.12
1.22
0.35
0.84
2
1.97
1.00
90.23
0.05
0.98
1.87
1.39
1.12
0.57
0.83
3
0.39
1.89
0.29
94.13
0.62
0.38
1.81
0.11
0.20
0.18
4
1.15
1.52
1.34
0.72
92.15
0.38
1.72
0.59
0.30
0.13
5
0.77
2.87
0.48
0.32
3.63
83.15
5.13
1.83
1.45
0.38
6
1.83
3.69
0.87
0.94
1.33
1.01
86.79
2.15
0.43
0.95
7
0.79
1.19
0.60
0.00
0.24
0.19
3.08
90.70
1.67
1.54
8
0.70
0.96
0.55
0.08
0.50
0.26
4.40
6.15
85.37
1.03
9
1.50
1.89
0.70
0.15
0.39
0.00
2.51
4.62
0.89
87.35
- 20 -
