The Distribution of Pension Wealth and the Process of Pension Building: Augmenting Survey Data with Administrative Pension Records by Statistical Matching [Elektronische Ressource] / Anika Rasner. Betreuer: Gert G. Wagner

Workshop

Sujets

• Statistics

Informations

• Publié par: technische_universitat_berlin
• Publié le: 01 janvier 2012
• Nombre de lectures: 9
• Langue: English
• Poids de l'ouvrage: 6 Mo

Extrait

THE DISTRIBUTION OF PENSION WEALTH AND
THE PROCESS OF PENSION BUILDING
AUGMENTING SURVEY DATA WITH
ADMINISTRATIVE PENSION RECORDS
BY STATISTICAL MATCHING

vorgelegt von
Anika Rasner
aus Korbach

Von der Fakultät VII Wirtschaft und Management
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Wirtschaftswissenschaften
Dr. rer. oec.
genehmigte Dissertation

Promotionsausschuss:
Vorsitzender: Prof. Dr. Axel Werwatz
Berichter: Prof. Dr. Gert G. Wagner
Berichter: Prof. Dr. James W. Vaupel
Tag der wissenschaftlichen Aussprache: 30. November 2011
Berlin 2012

D 83
1


2

Acknowledgments
Completing a dissertation feels like a marathon, and I would not have
been able to complete this journey without the aid of countless people
over the past years. I must first express my gratitude towards my advisors
Prof. Dr. Gert G. Wagner and Prof. Dr. James W. Vaupel for their sup-
port, supervision and guidance.
Gratitude is also expressed to Dr. Joachim R. Frick and Dr. Markus
Grabka for this priceless collaboration in order to get the matches right
from the beginning to the end of this project.
Many thanks go to Dr. Jutta Gampe who became a great mentor and al-
ways had the best piece of advice and Dr. Olaf-Groh Samberg for his pa-
tience in helping me with loops and other seemingly insurmountable
obstacles. Thank you to Silvia Leek who is responsible for all the nice
graphs in this dissertation. I gratefully thank Prof. Don Taylor at the San-
ford School of Public Policy at Duke University for giving me the oppor-
tunity to work with restricted HRS data and Anne Fletcher for her great
support during my research stays in the summer of 2009 and 2010. My
stays at Sanford greatly benefited my studies that were made possible be-
cause of the generosity and hospitality of Prof. Sunny Ladd and Prof.
Marjorie McElroy who are always a source of inspiration and great role
models for me. Special thanks go to Edgar Kruse and Dr. Ralf K. Him-
melreicher who provided me access to administrative pension records and
supported the statistical matching project.
Thank you to all my friends and colleagues at the DIW Berlin and the
Max Planck Institute for Demographic Research for all your help and
support.
Lastly, and most importantly, I wish to thank my parents and brother for
their confidence and support. To them, I dedicate this thesis.

III

IV
Table of Content
ACKNOWLEDGMENTS III

TABLE OF CONTENT I
LIST OF FIGURES V
LIST OF FIGURES (APPENDIX) VII
LIST OF TABLES VIII
LIST OF TABLES (APPENDIX) X
LIST OF ABBREVIATIONS XIII
1 INTRODUCTION 1
1.1 Motivation 1
1.2 Problem Statement 4
1.3 Contributions 11
1.4 Outline of the Thesis 15
2 ANALYSIS WITHIN LIMITS: THE POWER OF
ADMINISTRATIVE DATA 16
2.1 Introduction 16
2.2 The German Statutory Pension Insurance 19
2.3 The Sample of Active Pension Accounts 20
2.4 Limits to Generalizability: What is the Population to
Generalize to? 22
2.5 Limits to Accuracy and Selectivity Concerns: The
Case of Pension Account Validation 32
2.6 Limits to Measurement: Assessing the Validity and
Reliability of Variables 39
2.6.1 Earning Points 41
2.6.2 Educational Attainment 44
2.6.3 Unemployment 46
2.7 Limits to Content: The Lack of Relevant Covariates 49
2.8 Conclusion 56
I
3 BEST OF BOTH WORLDS: PREPARATORY STEPS IN
MATCHING SURVEY DATA WITH ADMINISTRATIVE
PENSION RECORDS 59
3.1 Introduction 59
3.2 Issues of Data Confidentiality 62
3.3 The Data 64
3.3.1 Completed Insurance Biographies 2004 (SUF
VVL 2004) 64
3.3.2 The Socio-Economic Panel 66
3.3.3 Perfect Complements: The Best of Both
Worlds? 67
3.3.4 Potential Pitfalls: When Worlds Collide 69
3.4 Specification of the Sample Population 71
3.4.1 Specifying the Analysis Population in the
SOEP 72
3.4.2 Specifying the Analysis Population in the SUF
VVL 2004 74
3.5 Finding Matching Variables 75
3.5.1 Monthly Public Pension Benefit 76
3.5.2 Time Spent in Different Pension-Relevant
States 83
3.5.3 Gender 91
3.5.4 Region 91
3.5.5 Marital Status 92
3.5.6 Number of Children 93
3.5.7 Retirement Age 95
3.5.8 Migration History 97
3.5.9 Type of Health Insurance 98
3.5.10 Educational Attainment 100
3.6 Estimating Regression Equations 104
3.6.1 Which Variables Enter Which Model? 104
3.6.2 Regressions Diagnostics 108
3.7 Regression Results 115
3.7.1 Discussion 119
II
3.7.2 Effectiveness of Modifications 124
3.8 Testing the Feasibility of a Statistical Matching 126
3.8.1 In-Sample Predictions 126
3.8.2 Assigning Random Residuals 128
3.8.3 Out-Of-Sample Predictions 130
3.9 Conclusion and Outlook 134
4 STATISTICAL MATCHING OF ADMINISTRATIVE AND
SURVEY DATA – AN APPLICATION TO WEALTH
INEQUALITY ANALYSIS 136
4.1 Introduction 136
4.2 The System of Old Age Provision in Germany 138
4.3 Data 143
4.4 First Match: Linking SOEP and SAPA 144
4.4.1 Notation and Conditional Independence 144
4.4.2 Statistical Matching or Imputation: Four
Alternatives 145
4.4.3 Matching Variables 147
4.4.4 Second Match: Record Linkage of SAPA and
Divorce Statistics 149
4.5 Assessing the Quality of Matches: Which Technique
Performs Best? 152
4.5.1 Sample Specification and Evaluation Criteria 152
4.5.2 Results 153
4.5.3 Discussion 159
4.6 Wealth Inequality 163
4.6.1 Determining the Present Value of Pension
Entitlements 163
4.6.2 Social Security Wealth Assuming Different
Discount Rates 164
4.6.3 The Distribution of Total Individual Net
Worth and SSW 167
4.6.4 Net worth and Extended Wealth across
Occupational Groups 169
4.7 Conclusion 172
III
5 WOMEN’S MARITAL TRAJECTORIES AND THE
ACCUMULATION OF PENSION BENEFITS IN GERMANY
AND THE UNITED STATES 175
5.1 Introduction 175
5.2 Literature Review 178
5.3 The Policy Background 182
5.3.1 The Process of Pension Building 182
5.3.2 The Institutional Design of Pension Schemes 183
5.3.3 The Impact of Marital Transitions 185
5.3.4 Welfare States and Incentives 190
5.3.5 Research Hypothesis 193
5.4 Data & Analytic Approach 196
5.5 Marital Trajectories and Retirement Outcomes:
Comparing Retired Women in Germany and the
U.S. 201
5.5.1 Descriptive Results of Marriage Patterns 201
5.5.2 Clusters of Marital Trajectories 205
5.5.3 Retirement Outcomes and Marital Trajectories 207
5.6 Marital Trajectories & Pension Building in the Pre-
Retirement Cohorts 219
5.6.1 Marriage Patterns and Marital Clusters 219
5.6.2 Paths of Pension Building 223
5.7 Conclusion 226
6 CONCLUSION AND OUTLOOK 231
REFERENCES 238
APPENDIX 256



IV
List of Figures
Figure 1 Comparing Pension Records and Population Counts,
Birth Cohorts 1940-1992 25
Figure 2 Share of Years with Missing Information, by Gender,
Region and Migration Status 30
Figure 3 Receipt of Unemployment Compensation in SAPA data,
1970 - 2007 48
Figure 4 Distribution of Monthly Public Pension Benefits in SOEP
and SAPA data 51
Figure 5 Fictitious Employment History 84
Figure 6 Returns to Education based on SOEP and SUF VVL 104
Figure 7 The Distribution of Monthly Public Pension Benefit for
Different Demographic Groups in the SOEP 106
Figure 8 The Distribution of Monthly Public Pension Benefit for
Different Demographic Groups in the SUF VVL 2004 106
Figure 9 Distribution of Public Pension Benefit in SOEP & SUF
VVL 2004 across Deciles 110
Figure 10 Example for In-Sample Predictions, Women West 127
Figure 11 Example for In-Sample Prediction SOEP, Women West 128
Figure 12 In-Sample Predictions with Randomly Assigned Residuals
SOEP, Women West 129
Figure 13 95% Confidence Bands for Predictions with Randomly
Assigned Residuals, Women West 130
Figure 14 Example for Out-of-Sample Prediction, Women West 131
Figure 15 Comparison of Observations and In and Out of Sample
Prediction, Women West 132
Figure 16 Out-of-Sample Predictions with Randomly Assigned
Residuals SOEP, Women West 133
Figure 17 95% Confidence Bands for Out-of-Sample Predictions with
Random Residuals 134
Figure 18 Composition and Level of Monthly Old-Age Income
Before Taxes for Men and Women Aged 65 and Older 141
Figure 19 Statistical Matching Process at a Glance- SOEP, VSKT and
Divorce statistics 151
Figure 20 Kernel Density Plots for Individual Differences between
Observed and Matched Benefit Information - Total
Population 156
Figure 21 Residual Plot Testing for Conditional Independence 159
V
Figure 22 Present value of pension wealth entitlements by age for
different discount rates, Germany 2007 165
Figure 23 Net worth, present value of pension wealth entitlements
and extended wealth by age, Germany 2007 166
Figure 24 Average Public Pension Benefit across Marital Clusters in
Germany and the U.S. 208
Figure 25 Average Total