Age-specific migration behaviour in Japan using spatial interaction models

1
Age-specific migration behaviour in Japan using
spatial interaction models

• Keiji YANO
• Department of Geography,
• Ritsumeikan University,
• Kyoto, Japan
• (yano_at_lt.ritsumei.ac.jp)

Presented at Third International Population
Geographies Conference University of Liverpool,
United Kingdom, 19-21 June 2006
2
OBJECTIVES

• 1) To estimate age-disaggregated
  inter-municipality migration OD matrices in the
  late 1980s and the late 1990s
• 2) To examine changes in Japanese migration
  system and the determinants of destinations
  choice using spatial interaction modelling from
  the viewpoints of the changing economic
  conditions and the age variation

Movement away from global and towards local
statistics!
3
PLAN of PRESENTATION

1. Study area, study periods and summary of Japanese
   migration
2. the age-disaggregated inter-municipalities with
   the estimation method
3. Formulation of spatial interaction models
4. the three parameters calibrated by
   origin-specific competing destination models to
   17 (5 year interval) age groups migration data
   for 2 periods
5. Conclusions

4
Spatial units in 1990 No. of municipalities
3372
Honshu
Kyushu
Hokkaido
Shikoku
Spatial units in 2000 No. of municipalities 3368
Tokyo
Osaka
5
Total (Over 500)
6
Economic conditions in Japan
the late 1990s
the late 1980s
Rate of economic growth ()
Bubble economic boom
Economic stagnation
7
Summary of Japanese migration
Boom
Stagnation
8
Total
Inter-municipality
Within same municipality
From abroad
Ratio of age-disaggregated migrants in 1990 2000
9
2000
First baby boom
Second baby boom
1990
10
1985-1990
Within same municipality
1995-2000
11
Estimating age-disaggregated OD matrices
1995-2000 3,368 municipalities Total, Male,
Female 17 age groups
1985-1990 3,372 municipalities Total, Male,
Female 17 age groups
17 Age groups Age 5_ 9, Age10_14, Age15_19,
, Age80_85, Age85_
12
Iterative Proportional Fitting method(Entropy
Maximizing method)

• Tijk Aik Bjk Cij
• Aik Sj Tijk / Sj Bjk Cij
• Bjk Si Tijk / Si Aik Cij
• Cij Sk Tijk / Sk Aik Bjk

Constraints of marginal totals
Sk Tijk
Sj Tijk
Si Tijk
13
Applying IPF to each of 47 prefectures
14
Estimated age-disaggregated OD matrices
1995-2000 3,368 municipalities Total, Male,
Female 17 age groups
1985-1990 3,372 municipalities Total, Male,
Female 17 age groups
17 Age groups Age 5_ 9, Age10_14, Age15_19,
, Age80_85, Age85_
15
Total (Over 500)
16
Age 5_ 9
17
Age10_14
18
Second baby boom
Age15_19
19
Age20_24
20
Second baby boom
Age25_29
21
Age30_34
22
Age35_39
23
First baby boom
Age40_44
24
Age45_49
25
First baby boom
Age50_54
26
Age55_59
27
Age60_64
28
Age65_69
29
Age70_74
30
Age75_79
31
Age80_84
32
Age85_
33
Spatial Interaction Models
Movement away from global and towards local
statistics!

• Origin-specific Production Constrained Competing
  Destinations Model
• pij Pjai Ajgi dijbi / Sj Pjai Ajgi dijbi ,
• Aj Sk (Pk / dik), Accessibility of
  destination j
• Pj Population size of destination j
• dij Distance between origin i and destination j

34
Interpretation of parameters

• Origin-specific Production Constrained Competing
  Destinations Model
• pij Pjai Ajgi dijbi / Sj Pjai Ajgi dijbi ,

For each origin i Alpha ai Attractiveness
of destinations with large populations
Beta bi Distance deterrence (Distance-decay)
Gamma gi ( ) Attraction of large clusters of
destinations Agglomeration effect (
- ) Competing destinations Competing effect
35
Pj
Population (Attractiveness)
36
Aj Sk (Pk / dik)
Accessibility
37
Number of calibrations
For the late 1980s (1985-1990) 18 (Total 17
Age groups) 3,372 (origins) 60,696 (models)
For the late 1990s (1995-2000) 18 (Total 17
Age groups) 3,368 (origins) 60,624 (models)
So , 121,320 models are calibrated in total!
38
Results of calibrations
39
Goodness-of-fit (PDEV)
Good
Bad
Goodness-of-fit (R-square)
Good
Bad
40
Basic statistics of attractiveness parameters
(a(i))
41
ai
Age-disaggregated attractiveness parameters
(a(i) average)
42
Total
Spatial distribution of attractiveness parameters
(a(i))
43
Age 5_ 9
Spatial distribution of attractiveness parameters
(a(i))
44
Age20_24
Spatial distribution of attractiveness parameters
(a(i))
45
Age75_79
Spatial distribution of attractiveness parameters
(a(i))
46
Basic statistics of distance parameters (b(i))
47
bi
Age-disaggregated distance parameters (b(i))
average)
48
Total
Spatial distribution of distance parameters (b(i))
49
Age 5_ 9
Spatial distribution of distance parameters (b(i))
50
Age20_24
Spatial distribution of distance parameters (b(i))
51
Age75_79
Spatial distribution of distance parameters (b(i))
52
Basic statistics of accessibility parameters
(g(i))
53
gi
Age-disaggregated accessibility parameters
(g(i) average)
54
Total
Spatial distribution of accessibility parameters
(g(i))
55
Age 5_ 9
Spatial distribution of accessibility parameters
(g(i))
56
Age20_24
Spatial distribution of accessibility parameters
(g(i))
57
Age75_79
Spatial distribution of accessibility parameters
(g(i))
58
ai
gi
bi
Boom
Stagnation
59
ai
gi
bi
Boom
A little bit LESS ATTRACTIVE
Almost SAME
From AGGLOMERATION To COMPETITION
Stagnation
60
Thank you very much for your kind attention!
61
SUMMARY

• 1. Age differences
• 1) Young migrants prefer large population and
  accessible destinations.
• 2) Middle migrants prefer inaccessible
  destinations excluding specific non metropolitan
  areas (Tohoku Kyushu).
• 3) Elderly migrants basically dont prefer large
  population and accessible destinations with
  strong distance deterrence.
• 2. Relationships with a change of economic
  conditions
• 1) Preference for large destinations decrease
  during economic stagnation.
• 2) Distance deterrence effect is relatively
  stable independently of the changing economic
  conditions.
• 3) Agglomeration effect becomes weak during
  economic stagnation.

62
(No Transcript)
63
Goodness-of-fit (PDEV)
64
Goodness-of-fit (R-square)