Social Security as Viewed through Micro-data

Part 2: Characteristics of "Global Standard" Panel Survey

Consulting Fellow, RIETI

In discussing the social security system, it is crucial to incorporate explicitly the diversity of individuals and incentive factors. The only way to achieve this end is to collect continuously and analyze as many samples as possible in a manner as unbiased as possible, even though this process seems to be a roundabout way of achieving the goal. International efforts are now underway to construct globally comparable data on the middle-aged and elderly. In building up such a "global standard" dataset, it is vital to collect data from a sufficiently large sample at a high response rate, but the following three conditions are equally essential.

First, we must take a multidisciplinary approach to capture an individual's life in all aspects, ranging from the economic, health, work, family, and social participation. In surveys hitherto conducted in the sphere of economics, non-economic aspects such as health, family, and social participation have not been covered sufficiently. In contrast, surveys conducted for developing a global standard dataset ask respondents to answer as many as a few hundred questions about all aspects of their lives, which include grip strength measurement.

Second, we must employ a longitudinal approach, whereby we continue to follow the same respondents over time. The biggest advantage of this approach, called "panel survey," is that we can directly observe the genuine changes over time, netting out the influence of individual heterogeneity. This is important especially in the case of measuring the impact of a specific policy, where it is necessary to compare the situation before and after the implementation of the policy. Diversity across different samples makes it difficult to measure the genuine impact.

Suppose we want to know how the retirement rate changes over time (let's say in five years' time). Now, if we compare the retirement rate of 65-year-old men with that of 70-year-old men as found in the same survey, we cannot tell how much of the difference observed between the two age groups is attributable to the age factor. That is, when we compare two different groups of individuals, any observable difference is inevitably influenced by differences in individual characteristics that cannot be explicitly reflected in the data. Furthermore, people's retirement behavior may differ across generations. People aged 65 years now were born after World War II, whereas those aged 70 years were born during the war. Therefore, the environments in which these two groups of individuals were raised are quite different. However, there is no separating the impact attributable to such generation-to-generation differences from that attributable to other factors.

Then, what if we conduct the same survey twice on the same group of people, five years apart, first when they were at the age of 65 and second at the age of 70? This method is called a "pseudo panel," as it enables an inter-temporal comparison of the cohort of individuals. However, any observed change is aggregated at a group level and thus inevitably biased compared to the actual change at an individual level. For example, because people who live longer tend to earn greater lifetime income, estimation based on pseudo panel data tends to project a continuous upward trend in personal assets over an individual's lifetime, which is not always the case.

Third, international comparability must be ensured. Thus, surveys conducted in different countries are designed to allow this. At present, under the leadership of RAND Corporation in the United States, a harmonization project is underway to develop a database to facilitate easy access to internationally comparable datasets that allow for cross-national comparisons. The initiative involves comparing questionnaires for surveys in different countries and examining in detail their comparability to provide a standard set of questions. In conducting national surveys, these standard questions may be combined with country-specific ones depending on the circumstances of the respective countries. As such, the project aims to contribute to policy assessment and the discovery of new scientific knowledge in each country.

>> Original text in Japanese

* Translated by RIETI from the original Japanese "Yasashii Keizaigaku" column in the September 13, 2011 issue of Nihon Keizai Shimbun.

September 13, 2011