From the short view to the long view:Volatility and continuity
John Hills, Centre for Analysis of Social Exclusion, LSE
Longview Conference, Oxford, 21 July 2006
Initial income group
(tenths of individuals)
in 1991
Percentage of each initial income group
Ending up in each income group in 1992
Poorest 2 3 4 5 6 7 8 9 Richest
Poorest 46 21 15 5 4 5 3 1 0 2
2 23 39 20 11 4 1 1 1 0 1
3 12 19 28 22 8 3 3 2 2 1
4 7 9 19 27 20 9 5 2 0 2
5 2 4 11 15 30 22 7 5 2 1
6 3 5 5 10 17 25 18 10 5 2
7 3 1 1 4 11 20 36 14 6 3
8 2 1 1 2 2 11 19 34 17 6
9 4 2 2 2 2 6 8 23 41 13
Richest 2 1 1 1 1 2 3 7 24 58
BHPS data: Income mobility between 1991 and 1992
Initial income group
(fifths of individuals)
in 1991
Percentage of each initial income group
ending up in each income group in 2001
Poorest 2 3 4 Richest
Poorest 41 26 16 10 7
2 24 30 22 16 8
3 15 22 25 22 16
4 12 13 22 29 24
Richest 9 8 16 24 45
A longer view: Income mobility between 1991 and 2001
An even longer view: Parents’ incomes and daughters’ earnings, 1958 and 1970 cohorts
Daughter’s earnings group (early 30s)
Parents’ net income group
Bottom quarter Top quarter
(a) Daughters born 1958
Bottom quarter 26 18
Top quarter 18 35
(b) Daughters born 1970
Bottom quarter 33 15
Top quarter 13 40
Source: Blanden et al.
But what underlies these snapshots?• CASE/ NatCen income tracking project: survey of
weekly incomes of low- to medium income working families with children for whole of 2003-04 (sample drawn from WFTC recipients in winter 2002/3)
• Selected to give mix: lone parents/couples; 1-3 children; 1-2 earners; tenants/owners; higher/lower WFTC
• 192 agreed to take part; 180 started income reporting; 129 still in survey after six months; 110 still in at end of year; 93 records complete enough to use in this analysis
• No evidence of significant bias in attrition by initial characteristics, nor by variability in first part of year (for those completing six-month interview). However, where partnerships split or formed respondents tended to drop out.
Example case with regular weekly income
0
50
100
150
200
250
300
350
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51
Week
£/w
eek
Child benef it Net pay Tax credit in pay Other tax credit Other income
Example case with changing circumstances
0
100
200
300
400
500
600
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51
Week
£/w
eek
Child Benef it Income Support JSA Net pay Tax credit in pay Other tax credits
Stable with blips
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13
Period
£/f
ou
r w
ee
ks
D1
D2
D3
D4
Highly erratic cases
0
500
1000
1500
2000
2500
3000
1 2 3 4 5 6 7 8 9 10 11 12 13
Period
£/f
ou
r w
ee
ks H1
H2
H3
H4
Trajectory types
Cases
Highly stable (± 10% x 13) 7
Stable (± 10% x 11+; ± 20% x 1-2) 8
Broadly stable (± 15% x 11+; ± 25% x 1-2) 13
Stable with blips (± 15% x 10) 32
Rising 4
Falling 3
Erratic (± 25% x 10) 18
Highly erratic (all other cases) 8
Distribution of period income as percentage of case’s annual average income
0
5
10
15
20
25
Range of values
% o
f ob
serv
atio
ns
Income changes depending on length of income periods compared
-60
-40
-20
0
20
40
60
-60 -40 -20 0 20 40 60
% change periods 1-6 to periods 7-13
% c
ha
ng
e p
eri
od
s 3
-4 t
o p
eri
od
s 1
0-1
1
Income changes using different pairs of periods
-100
-75
-50
-25
0
25
50
75
100
-100 -75 -50 -25 0 25 50 75 100
% change periods 3-4 to periods 10-11
% c
ha
ng
e p
eri
od
s 5
-6 t
o p
eri
od
s 1
2-1
3
Implications for measuring income mobility: There may be a lot of noise!
• For this group of low to middle earners with children, income was (surprisingly?) variable over the year. Checks with administrative data suggests most of this was genuine, not a result of reporting lapses.
• Short-term volatility was greatest for those with lower incomes
• For measuring changes in income receipts, choice of periods compared can make a very large difference
• However, what people report as ‘normal’ or ‘usual’ income may be less variable – under examination
• And although this involved a large number of weekly income reports, it was a small number drawn from a particular population: other population groups may have less (or more?) volatility