The Impact of Survey Nonresponse on Estimates of
Health Behavioural Differences between People with
and without a Migration Background
E.J. Lugard (582185)
Supervisor: dr. M.P.B. van den Broek
Master of Science Thesis
Health Economics, Policy and Law
Erasmus University Rotterdam
June 2025
Word count: 8286
,Acknowledgements
I would like to express my sincere gratitude to my supervisor, dr. M.P.B. (Thijs) van den
Broek, for his guidance and dedication throughout the supervision of my thesis. The final
product as it lies before you would not have been possible without his insightful support.
Statement on the use of Artificial Intelligence
In the writing of this thesis, OpenAI’s ChatGPT was used as a sparring partner for the
development of Stata code and the rephrasing of several paragraphs. All AI-generated
information was critically reviewed and verified using scientific resources. The
argumentations, interpretations and conclusions presented in this thesis are entirely my own.
2
,Abstract
This study investigates the impact of survey nonresponse on estimates of health behavioural
differences between people with and without a migration background. Nonresponse is a
common issue in survey-based research and is more prevalent among people with a migration
background, potentially biasing studies focused on this population. Accurate estimates of
health behaviour are crucial in informing policies aimed at reducing health inequalities.
Using the SIM2015 dataset, late response was used as a proxy for nonresponse
following the continuum-of-resistance model. Respondents were classified as early or late
respondents, and the sample was randomly divided into two groups, excluding late
respondents in one group. Logistic regression examined the impact of excluding late
respondents on the association between migration background and two health behaviours:
BMI and sports participation. A sub-analysis repeated the analysis including the control
variables sex, age and education to examine their effect on the estimates.
Consistent with existing evidence, this study found late response to be more prevalent
among people with a migration background. Contrary to theoretical expectations, excluding
late respondents did not significantly affect the estimated association between migration
background and the measured health behaviours in either the main or sub-analysis. The
findings were robust to variation in the late response threshold and group randomization.
Even though the findings of this study provide no reason to believe higher
nonresponse among people with a migration background significantly affects estimates of
health behavioural differences between people with and without a migration background,
such effects may exist in other settings and under different analytical conditions. Therefore,
future research should include a broader range of health behaviours and replicate this analysis
in different settings to assess the consistency and generalizability of the findings.
3
, Table of Contents
Acknowledgements..................................................................................................................... 2
Abstract ...................................................................................................................................... 3
1. Introduction ....................................................................................................................... 6
2. Theoretical framework....................................................................................................... 8
2.1 Health behaviour ....................................................................................................................8
2.2 Migration background, nonresponse, and health behaviour ..................................................8
2.3 Impact of nonresponse on estimates of health behavioural differences ............................... 11
2.4 Nonresponse and the continuum-of-resistance model ......................................................... 11
3. Methods ........................................................................................................................... 13
3.1 Sample ..................................................................................................................................13
3.2 Measures and data preparation .............................................................................................13
3.3 Analysis ................................................................................................................................16
3.4 Sensitivity analysis ...............................................................................................................17
4. Results .............................................................................................................................. 18
4.1 Descriptive statistics .............................................................................................................18
4.2 Baseline comparison of groups ............................................................................................21
4.3 Main analysis........................................................................................................................22
4.4 Sub-analysis with control variables......................................................................................23
4.5 Sensitivity analysis ...............................................................................................................24
5. Discussion ........................................................................................................................ 31
5.1 Key findings .........................................................................................................................31
5.2 Interpretation of findings ......................................................................................................31
5.3 Study strengths .....................................................................................................................32
5.4 Study limitations ..................................................................................................................32
5.5 Policy implications ...............................................................................................................33
5.6 Recommendations for future research..................................................................................34
4
Health Behavioural Differences between People with
and without a Migration Background
E.J. Lugard (582185)
Supervisor: dr. M.P.B. van den Broek
Master of Science Thesis
Health Economics, Policy and Law
Erasmus University Rotterdam
June 2025
Word count: 8286
,Acknowledgements
I would like to express my sincere gratitude to my supervisor, dr. M.P.B. (Thijs) van den
Broek, for his guidance and dedication throughout the supervision of my thesis. The final
product as it lies before you would not have been possible without his insightful support.
Statement on the use of Artificial Intelligence
In the writing of this thesis, OpenAI’s ChatGPT was used as a sparring partner for the
development of Stata code and the rephrasing of several paragraphs. All AI-generated
information was critically reviewed and verified using scientific resources. The
argumentations, interpretations and conclusions presented in this thesis are entirely my own.
2
,Abstract
This study investigates the impact of survey nonresponse on estimates of health behavioural
differences between people with and without a migration background. Nonresponse is a
common issue in survey-based research and is more prevalent among people with a migration
background, potentially biasing studies focused on this population. Accurate estimates of
health behaviour are crucial in informing policies aimed at reducing health inequalities.
Using the SIM2015 dataset, late response was used as a proxy for nonresponse
following the continuum-of-resistance model. Respondents were classified as early or late
respondents, and the sample was randomly divided into two groups, excluding late
respondents in one group. Logistic regression examined the impact of excluding late
respondents on the association between migration background and two health behaviours:
BMI and sports participation. A sub-analysis repeated the analysis including the control
variables sex, age and education to examine their effect on the estimates.
Consistent with existing evidence, this study found late response to be more prevalent
among people with a migration background. Contrary to theoretical expectations, excluding
late respondents did not significantly affect the estimated association between migration
background and the measured health behaviours in either the main or sub-analysis. The
findings were robust to variation in the late response threshold and group randomization.
Even though the findings of this study provide no reason to believe higher
nonresponse among people with a migration background significantly affects estimates of
health behavioural differences between people with and without a migration background,
such effects may exist in other settings and under different analytical conditions. Therefore,
future research should include a broader range of health behaviours and replicate this analysis
in different settings to assess the consistency and generalizability of the findings.
3
, Table of Contents
Acknowledgements..................................................................................................................... 2
Abstract ...................................................................................................................................... 3
1. Introduction ....................................................................................................................... 6
2. Theoretical framework....................................................................................................... 8
2.1 Health behaviour ....................................................................................................................8
2.2 Migration background, nonresponse, and health behaviour ..................................................8
2.3 Impact of nonresponse on estimates of health behavioural differences ............................... 11
2.4 Nonresponse and the continuum-of-resistance model ......................................................... 11
3. Methods ........................................................................................................................... 13
3.1 Sample ..................................................................................................................................13
3.2 Measures and data preparation .............................................................................................13
3.3 Analysis ................................................................................................................................16
3.4 Sensitivity analysis ...............................................................................................................17
4. Results .............................................................................................................................. 18
4.1 Descriptive statistics .............................................................................................................18
4.2 Baseline comparison of groups ............................................................................................21
4.3 Main analysis........................................................................................................................22
4.4 Sub-analysis with control variables......................................................................................23
4.5 Sensitivity analysis ...............................................................................................................24
5. Discussion ........................................................................................................................ 31
5.1 Key findings .........................................................................................................................31
5.2 Interpretation of findings ......................................................................................................31
5.3 Study strengths .....................................................................................................................32
5.4 Study limitations ..................................................................................................................32
5.5 Policy implications ...............................................................................................................33
5.6 Recommendations for future research..................................................................................34
4
