Unpleasant Facts and Other Musings

I thought that it would be worth looking into the demographics of China.  It would be nice to actually quantify its effect a bit more closely.  Luckily, I found a paper that did just the type of analysis that I was looking. David Bloom and Jeffrey G. Williamson wrote “Demographic Transitions and Economic Miracles in Emerging Asia”, published in April 1998. They used demographic and economic data from 1965 through 1990 to generate estimates on the impact of a demographic dividend.  They measured the impact of the growth of the population, the growth of the working age population (15 through 65) and other common variables on per capita GDP growth.  I took the average of the coefficients in table 4 to do a country specific analysis of China and India.  On average, the growth of a working age population helps per capita GDP growth by about 1.7% for every additional percent growth in working population, while every percent growth in the total population hurts per capita growth by 1.4%.

Without a demographic Dividend, China’s GDP growth would be marginally lower:

The demographic dividend’s effect (blue line) is disappearing, and will go negative after 2015. The orange line measures the total impact of population on GDP, combining the dividend’s effect on per capita income with the overall population change.

For India, their demographic dividend is mainly in the future.

India’s overall GDP growth will slow down with their general population growth, but their per capita GDP growth will be benefiting from a demographic dividend for quite some time.

This analysis relies on coefficients from 1965 through 1990, so the impact of the elderly is not going to be properly quantified.  There were no countries facing an abnormally large elderly population that expected benefits from the government. When the Bloom & Williamson paper separated out the youth dependency and elderly dependency ratios, they found that an elderly population actually added marginally to growth (Which makes some sense since unlike with youth the elderly population is marginally productive).  However, after economies have had to deal with the potential fiscal crisis due to their aging population the effect might be found to be quite a bit more negative.

I suspect that some measure of this nature might be helpful in separating out convergence growth from demographic driven growth. Unfortunately, the nature of the demographic dividend is that the dividend occurs during convergence, so a simplistic analysis can only explain so much.  

Data: April 2010 IMF WEO database was used for PPP GDP data. The 2008 UN World Population Prospects database (medium variant projections) was used to get the age structure of China and India.

The Consumer Expenditure Survey by the BLS just released an interesting article highlighting their 30 years as a continuous survey.  I decided that it could be interesting to see what their data combined with Census demographic projections imply about the future trend in savings. One way to do this is to break down the savings rate by age and project it forward.

As the savings rate obviously wasn’t 17.5% over the past few years, it should be noted that the CES implied measure of savings is rather different than the personal savings rate data calculated by the BEA.

In 2004, the CES changed the way it calculated income for incomplete survey responders, so the savings bias is calculated as an average of 2004 through 2008, which is about 15%.

The analysis assumes that the savings rate of each group will remain constant, and combines CES data with the US Census projections with constant net immigration to predict the future path of savings rate. The census cohorts are adjusted by the relative amount of consumer units because they suffer the same discrepancies as household data, with fewer young consumer units per capita.  This savings rate is adjusted downward by about 15% to be comparable to the BEA’s personal savings rate data.

The results are interesting in their mildness. The savings rate is only expected to drop a little over 1% due to demographic factors.  Of course, this study is biased for many reasons.

1. The accuracy of the using the aggregate consumer expenditure data without adjustment is questionable.
2. Social security and other retirement benefits are assumed to exist at levels comparable to the past few years. This biases the savings rate of the old cohorts higher compared to a likely scenario in which government and pensions have solvency issues due to the increasing dependency ratio. 
3. The educational and racial composition of the projections is held constant when the savings behavior of these cohorts is actually rather variable.
4. Wealth effects are completely ignored.

The Census also has voter participation data broken down by age or race, but not both.

This chart shows us what everyone knows: Old people are more likely to vote. This chart also explains why social security reform is not likely to happen, as old people won’t vote to give away their benefits and they may even vote against other reforms that only affect the young out of paranoia. The same could be said about programs that might have knock-on effects to Medicare down the road.

A little bit more surprising is the contrast between turnout in presidential election years and congressional election years:

I repeated the study using age based voter turn out numbers from the last two years as a guide, also adjusting the Hispanic voting population down by one third to adjust for the illegal immigrants who are unable to vote.  One surprising effect was that the slightly lower age of the growing minorities was not enough to affect the general thrust of the voting pattern.

Next I looked at the race based voter turnout.

And again, I compared it presidential election turnout with turnout from the previous year:

The results were rather extreme, showing how off year elections had a much bigger impact when voting by race was taken into account than when voting by age was taken into account.  This is because more whites vote during off years, with their off year voting at 75% of presidential election voters while only 66% of minority voters in presidential elections decide to vote during off years.

No model is ever truly complete, and this model is still lacking in numerous ways.

1. The predilection for the youth to vote for democrats has not been taken into account. An older population might be more conservative in nature, which would be a boon to the republicans.
2. Secondly, the recent data on black voter turnout was driven upwards by Obama’s election and it remains to be seen whether or not this recent shift will revert to the mean. 
3. Gerrymandering and state based Electoral College make national projections of the total popular vote less important than close analysis of swing states.
4. Short term economic fluctuations and personality politics means that the above charts are at best vague projections to show the direction of things to come where the emergence of a larger minority population makes republicans fall behind by an average of 0.1% or 0.2% of the population each year.

Furthermore, this analysis assumes no large shifts in party politics that might change racial voting preferences.  It does suggest that any big move towards the center is likely to be made by the republicans and not the democrats, as the democrats are riding a trend towards more power and while they may occasionally get ahead of themselves as they have with the healthcare legislation, the voting population is drifting further left every year.

One aspect of demographics that is often overlooked is the change in the working age population, the population between the ages of 20 and 65.

Source: UN Medium Variant Projections, World Population Prospects: The 2008 Revision

In the above graph, the US looks like it is pretty well situated compared to Japan and Western Europe. Japan’s economy, in addition to its other problems, needs to overcome one percent of shrinkage in its work force each year. Incidentally, this means that the nominal 3% GDP target recently announced by the government is in reality more of a 4% nominal GDP per worker target.  Australia, New Zealand and Canada (Not pictured) are in pretty good positions, with their steady immigration policies preventing their working age population from shrinking.

The United States looks like Canada and New Zealand, but the composition of workers is quite different. Australia and Canada’s immigration laws encourage educated immigrants. In the United States, the most fecund populations and the largest share of immigrants are not educated, but are instead composed largely of illegal immigrants who are periodically granted amnesty or their children who are natural born citizens.  Data from the US Census shows that the earnings gap per person is currently relatively significant:

The Census’s constant migration model of the US population shows that white non-Hispanic working age group is actually shrinking:

Assuming that the wage differential between ethnic groups does not shrink to a negligible size, this adds up to a working age demographic picture that isn’t quite as rosy for the United States when compared with Canada and Australia. Even so, it is still in better position than Europe and Japan who will have have to run the red queen's race to keep their economy the same size.

The goal of this blog is to focus on long term problems while avoiding the pessimistic bias by recognizing the positive impact of human ingenuity compounding over long time periods