The impact of grandchildren on retirement timing: evidence from SHARE data

• Thesis title in Czech: Závislost načasování odchodu do důchodu na existenci vnoučat: evidence s využitím dat SHARE
• Thesis title in English: The impact of grandchildren on retirement timing: evidence from SHARE data
• Key words: vnouče, důchod, instrumentální proměnná, SHARE, IV, vícerozměrná data, 2SLS
• English key words: grandchild, retirement, Instrumental variable, SHARE, IV, wide-ranging data, 2SLS
• Academic year of topic announcement: 2019/2020
• Thesis type: diploma thesis
• Thesis language: angličtina
• Department: Institute of Economic Studies (23-IES)
• Supervisor: Mgr. Barbara Pertold-Gebicka, M.A., Ph.D.
• Author: hidden - assigned by the advisor
• Date of registration: 22.07.2020
• Date of assignment: 22.07.2020
• Date and time of defence: 16.06.2021 09:00
• Venue of defence: Výuka probíhá online
• Date of electronic submission: 04.05.2021
• Date of proceeded defence: 16.06.2021
• Opponents: PhDr. Mgr. Jana Votápková, Ph.D.

References

1. Lumsdaine, R. L.; Vermeer, S. J., Retirement timing of women and the role of care responsibilities for grandchildren. Demography 2015, 52 (2), 433-454.
2. Kridahl, L., Retirement timing and grandparenthood in Sweden: Evidence from population-based register data. Demographic Research 2017, 37, 957-994.
3. Feng, J.; Zhang, X., Retirement and grandchild care in urban China. Feminist Economics 2018, 24 (2), 240-264.
4. Bergmann, M., Scherpenzeel, A., & Börsch-Supan, A. (2019). SHARE Wave 7 Methodology: Panel innovations and life histories. Munich: MEA, Max Planck Institute for Social Law and Social Policy.
5. Börsch-Supan, A.; Axt, K.; Beck, P.; Bergmann, M.; Coscia, V.; Korbmacher, J.; Malter, F.; Oepen, A.; Scherpenzeel, A.; Schmidutz, D., SHARE-ERIC. 2017.
6. Coile, C.; Gruber, J., Future social security entitlements and the retirement decision. The review of Economics and Statistics 2007, 89 (2), 234-246.

Preliminary scope of work

Motivation
The “right” age for retirement has been an issue for policy-makers for a very long time. With better and safer working conditions as well as extended life expectancy people can work longer and often surpass the legal required age for retirement set in their field of work. This means many people tend to leave the decision entirely as their own personal choice. Many aspects become relevant in this decision-making process. One of which is the grandparenthood. In other words, some people see grandchildren as one of the reasons to retire and have time to help with care. On the other hand, the relationship can have an opposite sign for people who would decide to keep their jobs in order to earn money to support their grandchildren. This relation is vital to examine to properly comprehend the choice of people to retire at different ages. It does not explain the whole picture, but it is another important piece in the puzzle for lawmakers to design policies accordingly.

Hypotheses
1. Is there a positive relationship between arrival of the first grandchild and the willingness to retire in case of people around retirement age?
2. Is there evidence in the data that a single grandparent is more active in case of care-aid than a grandparent with a partner?
3. Is there a connection between family wealth and willingness of grandparents to provide care for grandchildren?

Methodology
I will use data provided by the project SHARE. It provides datasets with detailed information about people of age over 50 from 27 different European countries and Israel. My plan is to use as much of the available information as possible to set up a formula that will be estimated. Since it is certain that retirement timing is not only affected by grandchildren, there are many variables included in the dataset that should be present in the formula. From the more expected ones such as number of grandchildren, age, nationality, education, occupation etc. trough the more “advanced” characteristics such as financial situation, health conditions, living arrangements (elderly house or not) to extremely detailed information such as willingness to answer and/or clarify, memory status, physical activity or years of daily smoking.
The main variable of interest is a dummy which describes whether the respondent is retired or not. On the explanatory side I will use as many characteristics mentioned in the previous paragraph as possible. One of them is going to be a dummy for having a grandchild to be able to examine the relationship. The interesting aspect of this additional variable is not only its statistical significance itself (to show the importance of grandchildren in the retirement decision-making) but also the sign of the link (as explained in the Motivation part).
There is simply a wide spread of various characteristics about people that answered those questionnaires. When carefully estimated even simple OLS method should yield interesting results. However, OLS provides information on the correlation between the two examined variables but does not help to uncover potential causality. Three main forms of the relationship can arise. Firstly, there could be no actual link between retirement and a birth of a grandchild. In other words, children are born during the same time that their grandparents retire with no direct mutual effect between these two. That would be an example of a spurious correlation. Second option would be reversed causality which in this situation is a following scenario: future parents (the second generation) wait with their own children (the third generation) until their parents or at least mothers (the first generation) retire. The third option of the relationship is causality which is the expected relationship in my hypotheses. In this case a birth of a grandchild motivates its grandparents to retire.
There are a few ways how to examine the relationship more deeply to avoid potential confusing results described in the previous paragraph. It is possible to gather the data with a panel structure to follow people across several years to see if retirement came earlier than grandchildren or vice versa. Furthermore, estimation can be performed separately for men and women because women are believed to have children earlier than men. Another option is to use an instrumental variable to account for the birth of a child. Since women tend to have children earlier than men, people of age above 50 would more likely be grandparents sooner if their first child was a woman than if it was a man. Estimating this instrument would mean searching for a positive correlation between earlier retirements of the grandparents (the first generation) and having a girl as the first-born child (the second generation).
I plan to get additional inspiration about possible estimation techniques useful in this situation from the literature review.

Expected contribution
Relationship between retirement and grandchildren is very complicated to describe. Many papers have provided insides on the matter (usual only for a specific country) but using the advancement of technology nowadays more information can be gathered and then used in the estimation part to get more precise results relevant for larger areas. The purpose of the thesis should thus be to provide information for policy-makers to enable a development of a retirement system that reflects the outcomes of the paper.
My contribution to the field should be the examination of the relationships that can affect the decision-making for retirement using widely detailed data. This work is going to be relevant for European countries since the data origin from there and Israel. However, results can also be applicable as a “guide” for different countries (to some extend of course) that are on similar socio-economic levels such as countries in the EU.
In some cases, I can imagine, this work could benefit those who are not certain about their decision and need “justification” from the community. In other words, find comfort in the knowledge that there are others that deal with the same situation similarly.
Last but not least, results will be relevant even for the Czech Republic because it should reflect the general situation across the European countries. However, later in the process estimation of the restricted dataset with data only from the Czech Republic can be performed to possible indicate deviations of Czech people from the European average behavior in this matter.

Outline
My plan is to start with comprehension of the papers mentioned bellow in the Core Bibliography section. One of these is a paper that describes that process of data extraction and explains all variables included in the datasets provided by the SHARE program. Before studying other related papers, understanding the structure and capacity of the data is crucial for me. Then I can choose such papers that can inspire me throughout my work while dealing inside the boundaries of the provided datasets. My writing section will follow a similar path. I want to start with a brief summary of the important papers in the field and then move straight to the methodology and estimation section. That should be sufficient to exclude those papers that would later turn out to be irrelevant because of the limits of my estimation techniques or the lack of available data.