The Frontier of Economic Analysis of Aging Demographics and Social Security Reforms: An introduction to the state-of-the-art macro models

Date July 30, 2015
Speaker KITAO Sagiri (Visiting Fellow, RIETI / Associate Professor, Hunter College and Graduate Center, City University of New York)
Moderator TASHIRO Takeshi (Consulting Fellow, RIETI / Deputy Director, Macro Economic Affairs Division, Economic and Industrial Policy Bureau, METI)



KITAO Sagiri 's Photo

KITAO Sagiri

My objective today is to introduce the state-of-the-art method of studying the issues of aging demographics and social security reforms. Before getting to the framework and the main topic, I would like to ensure that we all understand the big picture. The issue of aging is not just a Japanese problem: all developed countries and even some developing countries are dealing with it.

There are two peaks in the age distribution in 2010 based on the National Institute of Population and Social Security Research (IPSS) data. People from age 60 to 65--the first baby boom generation--represent one peak. People around age 40, who are the children of the first baby boom generation, represent another peak. What's abnormal is that the population declines monotonically after age 40. The Japanese birthrate has been declining, as is well known.

This means that the number of workers in the Japanese economy is going to shrink. There are about 75 million workers now, but by 2050, it will be 45 million, and by the end of the century, it will be 30 million individuals--a decline of more than 50%.

The elderly dependent ratio is defined as the ratio of the number of people above age 65--those receiving pension benefits--to the number of workers, who are between the ages of 20 and 64. This ratio was less than 40% in 2010, but it will rise dramatically as the first and second baby boom generations retire in succession, exceeding 80% by 2050. In just 35 years, the dependency ratio will increase by that much. It will then stay close to 100% for the rest of the century. That's the magnitude of the issue with which we have to deal.

What that implies in terms of the fiscal situation is that we are going to spend a lot more on public pensions, medical insurance, healthcare, etc. How will that be financed? Raising taxes is an answer, but as the number of workers shrinks so does the tax base, so revenues are declining. That leaves borrowing, reducing expenditures, or raising taxes.

As a macroeconomist, I want to know just how big and difficult the problems associated with aging demographics will be. What exactly will be the cost? That is what we want to measure. We want to have a number to assess the seriousness of the problem.

We can ask different questions. If current policies remain unchanged (pension benefits, health insurance system, etc.), and if we use the consumption tax to finance the rising expenditures, then what should the tax rate be? Or, if we want to keep the tax rate as it is, then how much do pensions need to be reduced?

We also want to understand how the economy is going to respond to changing demographics. What's going to happen to the gross domestic product (GDP), capital, output, etc.? Ultimately, though, we care about our happiness. Even if the GDP is higher, people might not benefit. How are future generations going to like the changes in the economy? Which changes would we like best?

Micro-founded dynamic (stochastic) general equilibrium model of heterogeneous agents

To answer all of these questions, we need an analytical framework--an economic model. We need to include some key ingredients in the model to enable us to answer the key questions. We need a mechanism through which we can measure and quantify how individuals are going to respond to the changes in the economic environment for different policies. We also want to understand how firms are going to respond--if there is a change in the number of workers, what will happen with wages? Competition for good workers will increase, so wages may rise. If people will need to save much more for retirement, capital may greatly increase, causing interest rates to go down. Once we understand how individuals are going to change their saving behavior and how firms are going to respond, we can then look at and summarize the response of each individual and compute the macroeconomic variable. We can't fully comprehend the entire situation by only looking at the big numbers, such as output. We have to understand individuals' responses. The government must also be in the model, as it will change its policies and individuals will respond to the changes.

The framework discussed today is called a micro-founded dynamic (stochastic) general equilibrium model of heterogeneous agents. It includes all of the necessary ingredients, and is a standard and very well-established approach in macroeconomic studies of redistributive policies in the United States. To understand the causes of inequality--e.g., wealth distribution--this is one of the most frequently used approaches in the United States. Several recent papers have been published in Japan as well.

"Micro-founded" as opposed to aggregate means that the model will be based on individuals' decisions. With regard to the use of a dynamic rather than static model, the latter would only analyze the present. A dynamic model seeks to analyze such factors as how much individuals want to save and how many hours they want to work, etc. by taking into account their lives in 10 years, post-retirement, etc. General equilibrium means that, within the model, all prices are determined by supply and demand. If the number of workers is going to be 50% of what it is now, do we still want to assume that the wage rate is going to be the same? Wages will probably need to increase to attract workers. The model also makes use of "heterogeneous" agents rather than a representative agent. Use of a representative agent, as in the case of a simple economic framework, assumes the existence of an agent who represents the behavior of the economy. We need to take into account the participation of a variety of different types of individuals in the economy, making a heterogeneous agent model necessary. Lastly, "stochastic" means that unexpected shocks occur in the economy which can be important in terms of individuals' decisions and macroeconomic aggregates.

Why not just focus on the aggregates? To understand the big numbers, understanding the micro behavior is necessary. To understand the "big C"--total consumption--how much we are consuming must be known. To understand the small numbers, we need a behavioral model, but obviously this entails greater complication. Therefore, simpler models also have their place. One recent such paper found that a major increase in consumption taxes would be needed to stabilize the debt.

What do we actually do with the model? In macroeconomics, we can't really run experiments. At the micro level, experimental economics is very popular in building theories. This is difficult to do in macroeconomics, so we conduct the experiment within the model. People with different assets, ages, genders, etc. live in the model. The fiscal variables are taxes, public pensions, health insurance, etc. Demographic factors include longevity and population distribution. There is a lot of data work involved in this model. Other parameters include expected income and expenditures, macroeconomic factors, and various sources of uncertainty. Given all of the parameters, the people in the model make optimal life-cycle decisions. We do the same for firms as well. We can then look at the outcome for each individual, summarize, and see whether policy changes affect the economy. People in the model then re-optimize, and we repeat.

Recent papers using equilibrium models have found, for example, that consumption tax rates need to be increased by 30%-40% to stabilize the debt, and that a fiscal crisis is going to happen by 2040 if major reforms are not carried out, requiring a consumption tax rate hike to over 50%. I also wrote a paper using the overlapping generations model of heterogeneous agents but also incorporating the retirement decisions of individuals--an important economic decision--and details of the health insurance and long-term care systems, finding that the consumption tax rate will have to reach 48% by 2080 absent reforms.

I will use my above-described paper to walk through the kind of model that we used this time and the kinds of results there are to discuss. The question is: what is the fiscal cost of demographic transition? I want to answer that question for both the short- and long-run. Once we understand what will happen far in the future, we can talk about the transition to getting there. As others have done, I will estimate and evaluate the cost of demographic transition in terms of the consumption tax rate that we will need to finance the different programs. I will also break down the fiscal costs. Once a benchmark model is established, we can run some change experiments.

Consumption tax

One reason that people choose consumption tax is that if anything else is chosen, it would be very difficult to solve the problem. If it is assumed that everything will be financed by debt, the debt will explode and the model will collapse. I did an experiment using income tax. People will leave the labor market, and then eventually a point is reached where even if the tax rate is raised, revenue plateaus. Consumption tax may not be the best choice, but it is the only feasible one for the model.

As for average gross medical expenditures in Japan, after age 50, each person spends about 200,000 yen. After that it skyrockets, and by the time you get to 85, each person on average spends about one million yen. Long-term care is even more serious. It reaches one million yen annually by age 90 and then reaches two million yen for each individual. With demographics shifting and people aging, there will be a large increase in government expenditures. It is important to incorporate expenditures and how the government will cover the costs.

For demographics, I use the IPSS fertility and survival rate estimates up to 2060. I make some assumptions because no official projections exist after that: that fertility rates will recover very slowly after 2060 and will reach 2.0 by 2150, and that survival rates will remain at 2060 levels thereafter. In the baseline scenario, I assume no change in pension policy, health insurance policy, etc., and see how much additional revenue is needed and reflect that in the consumption tax rate.


The simulation runs from 2010 to 2200. The consumption tax rate is determined by the government, but hypothetically I use the consumption tax to clear the budget. In 2010, with the dependence ratio at 39.8%, the consumption tax rate is 5%. After that, it changes endogenously. In the long run, by the end of the next century, the consumption tax rate would have to be 19%. At that point, the dependence ratio will be 56%--much higher than the 2010 level but the consumption tax rate is not so bad, especially compared to some European countries. In this economy, there will be more old people and fewer workers. People also will live longer. Pension relative to GDP will increase by three percentage points. Medical expenditures covered by the government will rise from near 6% to 8.2%. The population is going to shrink, from about 98 million to about 22 million, until the fertility rate recovers. Capital will not decline as much as population, labor, or output. Output and capital will be smaller. Output per capita and capital per capita will both increase: people in this economy will live longer so they will save more and the banks will lend that money, thereby increasing capital. The labor supply will be smaller so that labor per capita will be smaller. People will be working 1.5 years longer, but the labor force will be precious, so wages will also increase.

If all policies remain unchanged, the consumption tax rate would reach 40% by 2050, and 48% by the end of the century. Then, as the demographics normalize, the rate would decline. What explains the rise?

As a percentage of GDP, government pension expenditures will increase if the system remains the same. Right now, Japan spends around 10% of its GDP. It will reach 20% of GDP by 2080. This already accounts for a large increase in the consumption tax rate. Total gross medical expenditures (healthcare and long-term care) are currently 7% of GDP and will rise to 13% by 2060.


What if we reduce pension benefits by 20% or 40% and increase the pension age from 65 to 70 over a 50-year period? Experimenting with the model, we find that each of these would reduce the consumption tax rate by about 10%. Would such measures be feasible? No one wants to increase the consumption tax rate by 40% or cut benefits by 40%. However, we should all realize that these are the kinds of costs that we have to split and share across generations.

If reducing the pension or raising tax is not an option, another would be shifting from the current pay-as-you-go pension system to a mandatory individual retirement account (IRA) system, in which each individual saves for his or her own retirement. Keeping the basic pension intact, the money taken from a person's paycheck and given to retirees will instead be put into his or her own account, and the money will continue to be invested until retirement age is reached. The contribution rate is exactly the same percentage of the person's earnings. The money is invested over 45 years at maximum, making a big impact economically. Capital increases and labor become scarcer, raising wages. Introducing an IRA would significantly reduce the required consumption tax rate, especially because government expenditures on pensions would be reduced and tax revenues are higher because of the increased economic activities. In the long run, this may be very helpful, but it would not work well in the short run due to the high transitional cost.

Some combination of different policies will need to be formulated so that the issues can be resolved over the short, intermediate, and long run.

Foreign workers are another possibility. In a working paper I co-wrote on introducing a guest worker program, where foreign workers are allowed to work in Japan for a certain number of years, I determined that such a program would not increase revenues unless implemented at a very large scale. 200,000 workers staying for 10 years contributing taxes might reduce the consumption tax rate by 3% in the medium-run and 5% in the long-run.


What I would like to do now is think more about health insurance and medical expenditures. Macro-health is very important. 20% of GDP in the United States is spent on health. Ignoring that sector would be ignoring one-fifth of the entire economy. In Japan, it is about 10%, but it will soon rise. A great deal can be learned from the special field of health economics as well. Also, when thinking about how much people want to save or how much they want to work, medical insurance is really important. Hospitalization entails huge expenditures. That's something I would like to explore. Another promising field is family structures. The model we discussed today was fairly complicated yet also fairly simple in that we assume individuals as a unit of the model, but what is the family structure? Thinking about family structure in terms of an individual's decision-making is quite important, but it further complicates the model, too. It is a promising and growing field. Thank you very much.

*This summary was compiled by RIETI Editorial staff.