Changes in Registered Unemployment: Duration and Flows - CiteSeerX

Labour Market Bulletin 1995:1 Pages 21–40

Changes in Registered Unemployment: Duration and Flows MICHAEL FLETCHER1 This paper analyses recent trends in registered unemployment, focusing on the dynamics of the underlying flows on to and off the register. The data used are from a weekly duration profile database of the New Zealand Employment Service’s register of unemployed job seekers. The paper discusses trends in aggregate flows and exit probabilities and the impact these have had on the duration profile of the register; and analyses flows and duration trends by gender, age group, ethnicity and educational qualifications. The paper concludes with a consideration of the impact of the introduction of the Job Action programme on exit rates.

1 Introduction

S

TRONG GROWTH IN OUTPUT and employment has been accompanied, over the last two-and-a-half years, by a very rapid fall in the number of job seekers registering with the New Zealand Employment Service. By March 1995, total registered unemployment had fallen to 160,088, almost 30 percent below its 2 peak in December 1992. Changes in the number of people on the register—the stock of registered unemployed—result from the combined effects of changes in flows on to and off the register and the duration of unemployment spells. While most attention is usually focused on the numbers of people registered as seeking work at a given point in time, such static pictures are less informative than the underlying dynamics of register flows. This paper reports trends in registered unemployment up to March 1995 focusing on changes in the dynamics of the register. It is derived from databases developed in the Department of Labour over the last three years to analyse enrolments, exit flows and exit probabilities and to relate this information to the characteristics and labour market experiences of registered job seekers. The paper is in five parts. Section 2 summarises recent trends in the register. This is followed, in section 3, by an analysis of flows on to and off the register; the impacts of these flows on the duration profile of the register; and differences in flows and duration profiles for the various demographic and educational subgroups of job seekers. Finally, in section 4, the paper discusses how a flows analysis can be used to inform the assessment of employment policies, in this case the pilot Job Action programme for the very long-term unemployed which was phased in from September 1994. 1

Michael Fletcher is an adviser with the Labour Market Policy Group. The author thanks Bryan Chapple, Sylvia Dixon, Paul Gardiner, Veronica Korolev, Colin Lynch, Dave Maré, Paula Rebstock and Andy Reynolds for comments on earlier drafts, and Peter McCaughan and Andy Reynolds for programming assistance. 2 By the time of publication registered unemployment had fallen to 150,626 (31 August 1995).

21

22

Labour Market Bulletin 1995: 1

2 Recent trends in registered unemployment Registered unemployment fell from 195,221 to 160,088 in the year to March 1995—a decrease of 18.0 percent. 3 This followed a fall of 9.4 percent in the previous year and contributed to a 27.8 percent total decline since the peak of 229,391 in December 1992. This is the most rapid and prolonged decline in unemployment in New Zealand’s recent history and it comes after a period of almost continuous growth in the number of job seekers since the beginning of 1986 (see Figure 1). In broad terms, the decline in unemployment has been shared by all groups on the register, though there have been some significant differences within groups. In the year to March 1995 the number of men on the register fell by 19.3 percent, and the number of women fell by 14.0 percent (see Table 1 below). The proportion of men on the register fell slightly to 64.3 percent. Across age groups, the percentage change in the year to March 1995 has been greatest for 20–25 year olds (-23.4 percent), and least for under-20 year olds (-4.4 percent). At the end of March, almost half (47.5 percent) of all registered job seekers were aged between 25 and 44 years; 13.6 percent on the register were under 20 years old. During the same period, the number of Päkehä on the register decreased by 21.8 percent, the number of Mäori by 13.6 percent and the number of Pacific Islands peoples by 17.0 percent. A lower rate of fall in Mäori registered unemployment occurred despite the high rate of employment growth among Mäori recorded in the Household Labour Force Survey (9.8 percent compared with an overall increase of 5.0 percent in the year to March 1995). For all major ethnic groups, the reduction in numbers on the register has been greater for men than for women; although, comparing Mäori with Päkehä, the differences between men and women of the same ethnic group have been smaller than the differences between groups. The number of job seekers in the ‘other ethnic groups’ category actually rose by 791 or 7.2 percent, although this group—a proportion of whom are recent migrants—still have comparatively short average completed and uncompleted durations. Looking at the educational qualifications of job seekers reveals that registered unemployment continues to be dominated by those with few qualifications. 3

These numbers are from the NZES EMIS database, which is the official measure of registered unemployment. The analysis in this paper is derived from the Department of Labour’s database of registered unemployed job seekers by weekly duration. Owing to technical differences in the way re-enrolments are treated, totals from the two databases may differ slightly. There are also some differences in the way job seekers’ durations on the register are calculated. In particular, in the database used for this analysis, a job seeker who re-enrols within two weeks of leaving the register is deemed not to have left and his or her duration of registration is counted from the previous enrolment date. Similarly, departures from the register which are followed within two months by a re-enrolment coded by the Employment Adviser as ‘re-enrolled due to error’ are deemed not to have occurred and the job seeker’s duration is counted as continuous from the earlier enrolment.

Michael Fletcher

23

FIGURE 1: Registered unemployment, 1980–1995 250,000

200,000

150,000

100,000

50,000

Actual

January-1995

January-1994

January-1993

January-1992

January-1991

January-1990

January-1989

January-1988

January-1987

January-1986

January-1985

January-1984

January-1983

January-1982

January-1981

January-1980

0

Seasonally adjusted

Source: Department of Labour, EMIS database

Fifty-one percent have no qualifications at all, and a total of 74 percent have no formal qualifications beyond School Certificate. The most notable trend in the year to March 1995 has been the rapid decrease in the number of registered unemployed with trades qualifications, which fell by 26.9 percent in the year. This decline was largely the result of much lower levels of new registrations by tradespeople, rather than a rapid increase in exits of already unemployed job seekers with trades qualifications. The number of unemployed whose highest qualification was School Certificate fell by 19.4 percent; and the number of unemployed with no qualification fell by 16.3 percent. The number of job seekers with university degrees fell by only 6.3 percent, though this group remains a relatively small proportion of the total register. One of the most significant trends over the last two years has been the fact that long-term unemployment—defined in New Zealand as those unemployed continuously for 26 weeks or more—has fallen more rapidly than has short-term unemployment. In the two years to March 1995, long-term unemployment fell by 32.4 percent, compared with an 18.6 percent fall in the number on the register for less than 26 weeks. By March 1995, long-term unemployment comprised under half the total register for the first time since August 1991. The number of job seekers registered for 52 weeks or more has declined by an even greater

24


TABLE 1: Composition of registered unemployment Number on register as at 31 March 1995 Total Gender Age

Ethnicity

Education

165,724 106,550 59,174 22,491 38,753 78,675 25,805 90,951 49,831 13,225 11,712 84,310 38,620 24,570 8,181 10,367

Male Female Under 20 20–24 25–44 45 plus Päkehä Mäori Pacific Islands Other None School Certificate Other school Trade Degree

Annual change Number Percent 35,084 -25,434 -9,655 -1,023 -11,819 -6,187 -6,375 -25,340 -7,818 -2,717 +791 -16,393 -9,375 -5,357 -3,008 -694

17.5 -19.3 -14.0 -4.4 -23.4 -17.1 -19.8 -21.8 -13.6 -17.0 +7.2 -16.3 -19.4 -17.9 -26.9 -6.3

Source: Department of Labour, weekly duration profile database

amount—33.6 percent—over the two years to March 1995. The number registered continuously for more than two years has been slower to start falling, but over the last year declined by 24.7 percent; compared with a 17.5 percent decrease in the total register (see Table 2). Although the incidence of long-term unemployment is now declining, it is still higher than it was in 1991 when the rate of unemployment was similar to what it is now. The dynamics behind these trends are discussed in more detail in the following sections. TABLE 2: Annual percentage change in registered unemployment by duration

Year to:

Total register

Under 26 weeks

26 plus weeks

52 plus weeks

104 plus weeks

March 1994 March 1995

-10.5% -17.5%

-8.8% -10.8%

-11.9% -23.3%

-13.0% -23.7%

-2.2% -24.7%


3 Flow analysis and changes in the duration profile The trends described above concern numbers on the register; that is, changes in the stock of job seekers. Although these data provide useful snapshots regarding changes in who is registered as unemployed, and are the most common focus of

Michael Fletcher

attention, they do not tell anything directly about the dynamics of registered unemployment. For this purpose, it is more useful to analyse flows on to and off the register. Flow analysis allows us to focus on the changing patterns of enrolments and exits from the register which have resulted in stock changes, and which underlie changes in the duration profile of the register. It also reveals more about the interaction between changing labour market circumstances and the different labour market behaviour of sub-groups on the register. The following sections discuss aggregate flows on to and off the register and the impact of these on duration profiles, and differences in flows for various demographic subgroups of the registered unemployed. 3.1 Aggregate flows Figure 2 charts annual flows on to and off the unemployment register for the five years from 1990–91. These flows are much larger than the stock of job seekers on the register at any time. The chart shows that inflows remained fairly constant through to 1993–94, before falling in the last year. Outflows rose steadily from 1991–92, exceeding inflows by 25,000 in 1993–94 before dropping back slightly in 1994–95. At first sight, this pattern seems surprising. Typically, a pick-up in labour demand is reflected first in a decline in the number of lay-offs and redundancies, followed by an increase in new jobs and additional hiring. Thus one might have expected to see inflows on to the register declining before outflows started to increase. Assuming that the number of layoffs and redundancies has in fact declined during the recovery, the constant level of inflows could be due to an increase in the numbers of people entering the labour force and registering with NZES while they search for work; or it could be caused by higher turnover (or both). The data available do not allow us to distinguish between these explanations with certainty, but two points should be noted: • The rate of outflows (ie outflows as a proportion of the stock) continued to increase from 1991–92 right through to March 1995 (see Figure 3). The average weekly probability of exit for the first quarter of 1995 was 3.7 percent, compared with 3.1 percent for the same period in 1994; 2.6 percent in 1993; and 2.3 percent in 1992. This implies that the decline in the absolute number of outflows in the most recent year has been the result of a reduction in the total size of the register, rather than, say, a downturn in employment opportunities or a decline in the average ‘employability’ of those on the register. • The number of people who experienced repeated spells on the register rose, relative to the number who experienced only one spell. Comparing the calendar year 1994 with the year to June 1994 reveals that the number of people with only one (completed or uncompleted) spell in the 12 months fell by 10,341 (-3.4 percent), but the number who had two spells actually rose by 250 (0.3 percent).

25

26


FIGURE 2: Annual inflows and outflows from register 320,000

300,000

280,000

260,000

240,000

220,000

200,000 1990–91

1991–92

1992–93

1993–94

1994–95

March years Inflows

Outflows


This suggests that increased turnover, or ‘churning’, within the register may be part of the reason why enrolments have been slow to fall. Job seekers are finding it easier to leave the register—but the proportion re-enrolling reasonably soon afterwards has not fallen commensurately. Higher levels of turnover are consistent with a strengthening labour market, though further analysis is required to identify why people are re-enrolling (and/or whether this has changed as the cycle has strengthened). The gap between inflows and outflows appears to have widened since the beginning of 1995 (Figure 4). Seasonally adjusted outflows exceeded 27,000 per month after December 1994, reversing the slowdown in outflows which occurred in the first three quarters of 1994. The phasing in of Job Action since September 1994 contributed to this rise in outflows; but the numbers of Job Action participants leaving the register before April 1995 (3,244) were not substantial enough for this to be the major cause. The impact of Job Action is discussed in more detail in section 4 below. 3.2 Changes in the duration profile of registered unemployment The duration profile of the register is determined by current and past flows into and out of unemployment. It is of particular policy importance because of

Michael Fletcher

27

FIGURE 3: Percentage of job seekers leaving the register each week 5

4

Percent

3

2

1

6/01/95

8/07/94

7/01/94

9/07/93

8/01/93

10/07/92

10/01/92

12/07/91

11/01/91

13/07/90

12/01/90

14/07/89

0

Source: Department of Labour, weekly duration profile database Note: The annual downward spikes are caused by the Christmas/New Year holiday. The large spike in July 1991 (9.3 percent) is the result of an administrative ‘clean-up’ of the register which occurred at that time. 4

widespread evidence here and overseas that the probability of leaving unemployment declines as the duration of unemployment lengthens; and evidence that once the incidence of long-term unemployment rises it appears not to fall back to previous levels when labour demand increases. This is one of the major reasons why recent policy initiatives have been aimed at reducing the incidence of long-term unemployment. In New Zealand, long-term unemployment (defined as 26 weeks or more) rose from negligible levels at the beginning of the 1980s to over 50 percent of total unemployment by 1991. The proportion of the register with a duration of 52 weeks or more had reached 30 percent by the beginning of this decade, and peaked at 37 percent in June 1993. However, in the last two years the annual percentage fall in both the 26-week plus and 52-week plus categories has been greater than the percentage fall in short-term unemployment. In the year to March 1995, the number registered for 4

See for example OECD Employment Outlook (1993 and 1995), Gardiner et al (1994), and Grimmond (1993).

28


FIGURE 4: Registered unemployment inflows and outflows (seasonally adjusted) 36,000 33,000 30,000 27,000 24,000 21,000 18,000

Inflows

February-1995

August-1994

February-1994

August-1993

February-1993

August-1992

February-1992

August-1991

February-1991

August-1990

February-1990

August-1989

15,000

Outflows

Source: Department of Labour

26 weeks or more fell by 23.3 percent, and the number registered for over two years fell by 24.7 percent. Short-term unemployment fell by 10.8 percent. The impact of these trends on the duration profile of the register is summarised in Table 3 below. It shows the recent decline in the incidence of longterm unemployment as a percentage of the total register. The proportion of shortterm unemployment has again risen to over half of the total register, and the proportion of unemployment in the 26–51, 52–104 and 104–200-week categories have all declined below their 1992 levels. Notably, the 201-week plus duration category has continued to increase as a percentage of the total. Five-and-a-half years ago that category contributed only 0.6 percent to the total register.5 By March 1995 it represented 9.3 percent of the total. There has however been a slight numerical decrease in very long duration unemployment, suggesting that perhaps falling unemployment over an extended period may be beginning to feed through to this group. These trends are consistent with the relationship between the economic cycle and unemployment; and the incidence of long-term unemployment, seen in 5

The employment register was only computerised in its present form in 1985–86, at which time existing job seekers were ‘enrolled’ on it with zero weeks duration; so the true number of job seekers with a duration in excess of 200 weeks may be slightly higher.

Michael Fletcher

29

TABLE 3: Registered unemployment by duration, number and percentage of total 0–25 weeks

26–51 weeks

52–103 weeks

104–200 weeks

201 weeks plus

1990

84,494 53.5%

30,672 19.4%

31,222 19.8%

10,761 6.8%

910 0.6%

1991

103,461 52.2%

35,321 17.8%

32,941 16.6%

23,928 12.1%

2,729 1.4%

1992

99,676 45.1%

46,574 21.1%

42,824 19.4%

25,205 11.4%

6,738 3.0%

1993

102,433 45.7%

38,313 17.1%

38,979 17.4%

31,185 13.9%

13,350 6.0%

1994

93,436 46.5%

34,722 17.3%

29,108 14.5%

27,919 13.9%

15,623 7.8%

1995

83,343 50.3%

26,938 16.3%

22,668 13.7%

17,318 10.4%

15,457 9.3%

March :


many countries, in which a falling incidence of long-term unemployment lags 6 behind the turning point in the unemployment rate. This relationship is illustrated for New Zealand in Figure 5. As can be seen, the incidence of longterm unemploy-ment (defined in this case as the proportion of unemployed with 52 weeks or longer continuous registration) continued to rise through to mid 1993 even after the unemployment rate7 began to fall. The proportion of long duration unemployment began to decline from June 1994, five quarters after the peak in the unemployment rate. The changes to the duration structure of the register over the last two years are a combination of falling inflows into the longer duration categories, and of rising probabilities of exit across all durations. The first effect—which is the more substantial numerically—is the lagged impact of earlier improvements in short duration exit probabilities. Even with approximately constant enrolment levels from 1991 to 1994, higher exit rates for the short-term unemployed after 1992 mean fewer people are progressing to long-term unemployment. This effect is evident in Figure 6, which shows the annual average percentage change in 8 inflows into the duration categories. The time lag between the peak for each duration category is approximately equal to the duration difference between categories. For example, the annual change in flows into the 52–103-week 6

OECD Employment Outlook (1993 and 1995). Proxied here by the ratio of registered unemployment to the HLFS estimate of employment plus ‘joblessness’. 8 That is, the annual average percentage change in the numbers reaching 26, 52, and 104 weeks respectively. 7

30


FIGURE 5: The relationship between unemployment and the incidence of long-term unemployment 0.40 0.38

March 1993 March 1994

Proportion over 52 weeks

0.36 0.34

March 1995

0.32

March 1991

March 1992

0.30 0.28 March 1990 0.26 0.24 0.22 0.20 8.00%

9.00%

10.00%

11.00%

12.00%

13.00%

14.00%

Registered unemployed as % of broadly defined labour force Source: Department of Labour, weekly duration profile database and HLFS (See footnote 7 for labour force definition.)

category peaked and fell sharply approximately six months after the flow into the 26–51-week group. The 104-plus week group had a flatter peak but the same sharp fall which lagged behind the 52-week group by 12 months. Thus, the dynamics of the register duration profile are still heavily influenced by the period of low labour demand in 1991–92 when both short and long-duration exit rates were at their lowest. The second effect is that during 1994–95 exit rates rose relatively more for the long-term unemployed than for those with shorter durations (although they remained lower in absolute terms). Table 4 shows the percentage increase for different duration categories between the March 1993 year and the March 1995 year regarding the (unweighted) average probabilities of leaving the register within the subsequent six months. These changes imply that exit rates for the long-duration unemployed show greater cyclical volatility than do those for the short-term unemployed. Note that the particularly large increase for those with durations greater than 104 weeks reflects the additional impact of the phasing in of Job Action discussed in section 4 below.

Michael Fletcher

31

FIGURE 6: Annual average percentage change in inflows into duration categories 60% 50% 40% 30% 20% 10% 0% -10% -20% -30% -40% 5/07/91 0–25 weeks

3/07/92 26–51 weeks

2/07/93 52–103 weeks

1/07/94 104 weeks plus


3.3 Trends in flows for different groups on the register The discussion so far has focused on aggregate flows for all job seekers. However, there are important differences in the dynamics affecting various subgroups on the register which help explain the trends outlined in section 1. These differences reflect variations in the labour market circumstances of sub-groups and also their interaction with past patterns of flows which determine current duration profiles. The following sections analyse recent inflows and exit rates for the major demographic groups represented on the register. Gender As reported above, the proportion of men on the register fell from 65.7 percent on 1 April 1994 to 64.3 percent on 31 March 1995. The slightly higher net outflows (outflows minus inflows) for men are largely the result of a greater rise in exit probabilities for men than for women; though enrolments have also fallen more for women than for men (see Table 5). Women still have higher average exit rates than men across most of the duration profile—but the gap is smaller now than it has been in previous years.

32


TABLE 4: Average probability of leaving the register by duration of unemployment (March year 1993 and March year 1995) Initial duration

March 1993

0–25 weeks 26–51 weeks 52–77 weeks 78–103 weeks 104–130 weeks

March 1995

3.1 3.1 2.2 1.4 1.0

Percentage increase

4.0 4.0 3.1 1.9 1.9

27.3 29.3 40.4 37.7 90.0


The average duration of men and women leaving the register during the year to March 1995 has risen slightly compared with the year ending June 1994.9 This reflects the higher percentage fall in long-term unemployed. Average uncompleted durations for those still on the register has fallen, by 2.4 weeks for men and 0.2 weeks for women. As a group, men still have substantially longer durations of registered unemployment than do women. TABLE 5: Flows and duration of registered unemployed by gender Inflows (annual % change)

Men Women

Average completed duration (weeks)

Average uncompleted duration (weeks)

Year to Year to 30 June 31 March 1994 1995

as at as at 1 April 31 March 1994 1995

Average weekly exit probability (%)

March quarter 1995

March quarter 1994

- 4.66 - 1.37

2.8 3.7

March quarter 1995 3.4 4.3

39.0 27.9

40.3 28.0

66.1 43.0

63.7 43.2


Age Typically, exit probabilities (and inflow rates) are lower on average for older age groups than for young job seekers, who tend to have greater job mobility and more frequent spells of unemployment. Inflows of young job seekers are also more seasonal, with a large influx at the end of the academic year. During the year to March 1995, exit probabilities rose for all age groups— especially in the fourth quarter. However, the rise was less for the 15–20-yearold age group. This, coupled with the relatively high inflows of young people 9

The previous analysis was based on the June year.

Michael Fletcher

33

after the end of the 1994 academic year, meant that the stock of young people on the register fell by only 4.4 percent in the year to 31 March 1995; compared with a fall of 19.4 percent for job seekers over 20. Average completed and uncompleted durations remained shorter for young job seekers.

TABLE 6: Flows and duration of registered unemployed by age group Inflows (annual % change) March quarter 1995 Under 20 20–24 25–44 45+

+2.1 -6.9 -1.4 -10.4

Average weekly exit probability (%) March quarter 1994 5.0 3.2 2.8 2.2

March quarter 1995 5.1 4.1 3.4 2.7

Average completed duration (weeks) Year to 30 June 1994 21.6 35.6 37.1 42.0

Year to 31 March 1995 20.6 34.9 38.5 46.0

Average uncompleted duration (weeks) as at as at 1 April 31 March 1994 1995 21.9 47.5 65.8 79.9

21.6 44.5 62.6 80.8


The flow dynamics for the over-45 year old age group have been very different from that of the young unemployed during the past year. Enrolments of older job seekers have fallen sharply, with the inflow in the March quarter 1995 being 10.4 percent below the same quarter in 1994. At the same time, there has been only a modest increase in older job seekers’ exit rates, which remain substantially below those of other age groups. In the current strengthening labour market, it appears that older workers are less likely to become unemployed; but those who have been unemployed for some time are still finding it comparatively hard to re-enter the workforce. This is reflected in the fact that the over-45 age group is the only group for which average uncompleted durations have continued to increase in the year to March 1995. By this time, the average period on the register for this group was almost 19 months—six months longer than the average for the whole register. Thirty percent of all job seekers registered for 200 weeks or more are people over the age of 45. Ethnicity Comparing the inflows and exit rates of different ethnic groups highlights considerable differences which are not evident from the changes in the numbers on the register. Among Päkehä, inflows have fallen rapidly in recent months and exit rates have also risen reasonably strongly. Among Mäori, the rise in exit rates has been almost as strong; but there has been little reduction in inflows, resulting in a lower relative decline in the total number of Mäori on the register.

34


Pacific Islands job seekers have exhibited yet a different pattern. Exit probabilities for this group have risen more than for any other—their average weekly exit rate for the March 1995 quarter was 27 percent higher than for the same quarter in 1994. However, inflows of Pacific Islands peoples have also risen. As a consequence, the average duration of Pacific Islands job seekers on the register, which was higher than for any other ethnic group, fell from 66.5 weeks at the end of March 1994 to 61.5 weeks in March 1995, and is now slightly below that of Mäori. TABLE 7: Flows and duration of registered unemployed by ethnic group Inflows (annual % change) March quarter 1995 Päkehä -9.2 Mäori -0.3 Pacific Islands +6.4 Other +34.1

Average weekly exit probability (%) March quarter 1994 3.3 2.8 2.6 3.3

March quarter 1995 3.9 3.3 3.3 3.8

Average completed duration (weeks) Year to 30 June 1994 31.8 39.2 40.1 29.5

Year to 31 March 1995 33.4 39.4 41.0 30.1

Average uncompleted duration (weeks) as at as at 1 April 31 March 1994 1995 54.9 65.2 66.5 48.0

53.2 62.1 61.5 40.7


Flows among the ‘other ethnic groups’ category are different from any of the other three categories. This group—which includes a number of recent migrants—has a high average exit probability and the shortest average completed duration on the register. However, comparing March quarter 1995 with March quarter 1994 reveals that enrolments have risen by more than 34 percent. The reason for this increase is not clear. It should be noted that, despite the increase, this category still comprises only seven percent of the total register. For all ethnic groups the greatest increases in exit probabilities occurred among those unemployed on the register for between 104 and 199 weeks. As discussed below, the introduction of Job Action targeted at this duration category appears to have been the principal reason for this. Educational qualifications The great majority of the registered unemployed are poorly qualified, with 74 percent having either no qualifications or nothing beyond School Certificate. These job seekers have a lower propensity to leave the register and a substantially longer average duration of unemployment. Over the last year, however, exit probabilities have increased for these job seekers, as for all

Michael Fletcher

35

qualifications categories, and there has been a decline in the average uncompleted duration even among those with no formal qualifications. Two recent trends are of particular interest. First, the number of job seekers with trade qualifications has fallen much more sharply than that of any other group. This group was 27 percent smaller in March 1995 than it was one year earlier. This is consistent with other evidence suggesting that it is trade skills which are in shortest supply. Most of this decline was the result of a fall in new registrations, rather than an increase in the exit rate. Comparing the March 1995 quarter with the same quarter one year earlier reveals that the inflows for combined groups fell 3.1 percent. In comparison, inflows of people with trade qualifications fell by 16.4 percent. At the same time, though, the tighter labour market for trades skills has not been reflected in a comparably large rise in exit probabilities for tradespeople on the register. It is possible that those left on the register have redundant skills, or it may be that they have—or are perceived by employers to have—other disadvantages which make them less able to find work even in a strongly improving labour market. The second notable trend is that the number of job seekers with university degrees has fallen much less sharply than that of any other group. This number fell by only 6.3 percent over the year to March 1995. Moreover, the March quarter inflow of new registrations actually rose by 9.8 percent compared with the same quarter a year earlier. It is not clear what is causing this trend. It may reflect a relative over-supply of graduates; or simply the fact that this group was less seriously affected by the earlier recession and so has benefited relatively less from the pick-up in labour demand. If this latter hypothesis is correct, the higher inflows may reflect increased turnover. As a group job seekers with degrees still TABLE 8: Flows and duration of registered unemployed by educational qualifications Inflows (annual % change) March quarter 1995 None School Cert Other school Trade Degree

-1.6 -0.7 -4.4 -16.2 +9.8

Average weekly exit probability (%) March quarter 1994

March quarter 1995

2.6 3.3 3.8 3.6 3.8

3.2 3.9 4.6 4.2 4.5

Average completed duration (weeks) Year to 30 June 1994 40.2 32.2 27.9 30.5 25.5


Year to 31 March 1995 41.4 33.2 28.2 31.2 25.1

Average uncompleted duration (weeks) as at as at 1 April 31 March 1994 1995 70.1 53.2 39.7 49.2 33.6

66.7 50.9 38.1 47.9 31.1

36


fare relatively better than other job seekers; with more than two-thirds being on the register for less than 26 weeks, and the average uncompleted duration of this group being less than half that of those with no educational qualifications. Regions The register data have been disaggregated into NZES’s four regions. Unfortunately this is too broad a categorisation to isolate geographical variations in register trends. A more detailed analysis using either Employment Centre or district disaggregation is necessary. The regional breakdown does however show that, in broad terms, improvements in the register have been greater in the South Island and the Wellington region than in the central North Island and northern regions. The central region, which extends from the Waikato and East Coast to Wanganui, fared least well with a smaller (13.5 percent) fall in the number of job seekers and a smaller decline in the average duration of those on the register. The northern region, which fell by 17.5 percent, nevertheless still has the longest average uncompleted duration: 58.6 weeks. The register fell by 19.1 and 19.6 percent in the Wellington and southern regions respectively and job seekers remaining on the register in these regions have the shortest average duration, of just under 52 weeks.

4 The impact of Job Action on register flows This section provides an example of how employment policies can impact on register flows and the extent to which analysis of these flows can be used as a tool for monitoring the impact of policies. In September 1994 the Employment Service began piloting a new programme, named Job Action. Job Action is a formalised process for assisting long-term unemployed job seekers, who reach 104 weeks continuous registration, through a more individualised and intensive 10 delivery of assistance. The key objectives of Job Action are to improve the labour market outcomes for participants; to develop alongside each participant an individualised commitment to look for work; and to ensure that eligible participants are appropriately registered with NZES. Job Action has four distinct components: an initial one-to-one assessment interview with an NZES Employment Adviser; a one-week workshop aimed at improving job seekers’ motivation and job search skills; an individual ‘job action’ plan developed during the workshop and then agreed between the job seeker and NZES; and a programme of continuing case management and regular contact between job seeker and Employment Adviser to assist in implementing the job action plan. Each individual’s job action plan is tailored to specific needs 10

A slightly different programme, Youth Action, has been introduced to assist young job seekers who are, or are at risk of becoming, long-term unemployed.

Michael Fletcher

and may include a number of steps including participation in training and employment programmes and systematic job search. An evaluation plan was included as part of the implementation of Job Action and the full results of this evaluation are to be released shortly. The evaluation, which included quantitative and qualitative surveys of participants and comparison groups, has shown that Job Action is having a generally positive effect on registered unemployment and on participants’ behaviour and experiences. In particular, Job Action is having a substantial impact on the numbers on the register both as a result of a number of participants leaving the register for training or work, and the identification of people who are inappropriately on the register. The database used to analyse register flows for this paper is not adequate by itself for evaluating employment policies such as Job Action. However, it is useful for monitoring the effects of policies on the register as a whole and to illustrate how policies can affect exit rates. In the case of Job Action this is made easier by the fact that the policy was introduced in three phases, with different Employment Centres beginning Job Action at different times. Job Action was targeted initially at the flow of job seekers reaching 104 weeks’ duration. The first phase of centres began selecting participants on 5 September 1994, the second group on 5 January 1995 and the remainder on 1 April.11 Once centres were delivering Job Action to the full flow of job seekers reaching 104 weeks, some began including job seekers with a duration of more than 104 weeks. By 31 March 1995, 59 percent of the 12,009 job seekers who had 12 participated in Job Action were people whose registration duration exceeded 104 weeks at the time of first participating in the programme. Comparing the weekly exit probability curve for the five-week average to 31 March 1995 with that of one year earlier shows the emergence of a new ‘spike’ starting at about 104 weeks’ duration, caused by the increase in exit rates associated with Job Action (see Figure 8).13 Averaging over the first five weeks of the March year, we see that the probability of exit for people with a duration of between 104 and 117 weeks was 1.66 percent (per week). By the end of the year, this had risen to 3.64 percent. 11

Each of the three phases included both urban and provincial centres from different regions. 12 ‘Participants’ in this case includes all those people NZES contacted (or sought to contact) for an assessment interview. 13 The three other ‘spikes’ in the exit probability curve are the result of administrative actions. The first, at eight to nine weeks’ duration, occurs because job seekers not on worktested benefits are removed from the register if they have not contacted NZES within eight weeks of registering. The second two, at 26 and 52 weeks’ duration, are the result of Work Focus Interview follow-ups (the higher exit rate mostly reflecting people removed from the register for not attending the interview).

37

38


FIGURE 7: Average weekly hazard curve (total register)

Percent leaving per week

15

10

5

196

183

170

157

144

131

118

105

92

79

66

53

40

27

14

1

0

Duration on register (weeks) March-1994

March-1995


Job Action’s impact on exit probabilities is more clearly evident when comparing ‘phase one’ centres with those that introduced Job Action later. The average weekly probability of exit for job seekers with a duration of between 104 and 117 weeks in ‘phase one’ centres was 101.5 percent higher in the September 1994 quarter than in the same quarter in the previous year. This was almost twice the increase for job seekers of the same duration registered in other centres (see Table 3). In the following quarter, when the second phase of centres introduced Job Action, the annual increase for that group was 85 percent in ‘phase two’ centres, compared with 52 percent in ‘phase three’ centres.be cautious about this be cautious about this interpretation. Assuming that individual job seekers have differing probabilities of exit, it could be argued that the initial interview and workshop process is likely to have selected those with higher probabilities of exit. If this were the case, probabilities of exit two or three months later would be lower than before the workshops; unless the case management component of Job Action was also having a positive impact. The evaluation surveys provide some support for this hypothesis in that they found strong evidence that Job

Michael Fletcher

39

TABLE 9: Average exit probabilities for job seekers with a duration of 104–117 weeks Exit Phase Phase Phase

December quarter 1994 probability Annual (%) % change

one centres two centres three centres

2.16 1.67 1.83

Exit

103.9 59.2 54.4

March quarter 1995 probability Annual (%) % change 2.94 2.90 2.18

101.5 85.0 51.8


Action participants were more actively engaged in job search than comparable non-participants. The duration profile database can also be used to track cohorts of job seekers as they become eligible for Job Action at 104 weeks’ duration. This shows that each cohort’s average exit probability rises for about 10–12 weeks after the 104week duration mark, before falling back to approximately its earlier level. This suggests that the continuing higher exit rate evident in Figure 8 for durations greater than about 116 weeks may be a result of the participation of job seekers with a duration of that length in Job Action, rather than a persistent higher exit rate for job seekers who entered Job Action at 104 weeks. We should, however,

5 Conclusion This paper has analysed recent trends in registered unemployment, focusing on the dynamics of the underlying flows on to and off the register. In the period under examination unemployment has been falling rapidly, due to strong growth in output and employment. Until quite recently, the decline in registered unemployment has been a consequence of rising outflows, rather than a reduction in the number of people registering, or re-registering, with the New Zealand Employment Service. In the last year, however, the inflow on to the register has also begun to decline, while the rate—but not the absolute number— of exits from the register has continued to increase. The pattern of flows has also had an effect on the duration profile of the register as a whole. The proportion of short-term unemployed (less than 26 weeks) has begun to increase in the last two years, as long-term unemployment has fallen more rapidly than total unemployment. Looking more closely at the trends shows that this has been a result of the lagged effect of earlier improvements in exit rates for shorter-term unemployed; and of the relatively large increases in exit rates for the long-term unemployed (especially those with a duration of over 104 weeks) which have occurred during 1994–95.

40


The paper also analysed flows and duration trends by gender, age group, ethnicity and educational qualifications. There has been considerable variation in the enrolment and exit rate trends within sub-groups, indicating differences in dynamics which are not evident from analysis of point-to-point changes in numbers on the register. Finally, the paper discussed the impact of the introduction of Job Action on flows and the duration profile of the register. References Gardiner, P; Fletcher, M; Mersi, P; and Reynolds, A (1994) An analysis of the dynamics of the registered unemployed: exit probabilities and repeat spells, Department of Labour, unpublished paper. Grimmond, D (1993) Labour force dynamics in New Zealand, NZIER Research Monograph No 60, Wellington, NZ Institute of Economic Research. OECD (1993 and 1995) Employment outlook, Paris, OECD.