E-piphany

I have three items to discuss in this week’s post.

The first item is an announcement made by the BLS on Tuesday regarding upcoming changes to how the CPI for Health Insurance will be computed.

The backdrop for this change is that the CPI for Health Insurance is an imputed cost for the CPI. When a consumer buys health insurance, he/she is actually buying medical care, plus a suite of insurance products related to the actuarial benefits of pooling risks (that is, it’s much cheaper for people to buy a share of an option on the tail experience of a group of people, than it is for each person to buy a tail on their own experience – which is the main benefit/function of insurance). If all of the cost of health insurance was actually for health insurance, the weight of medical care itself (doctors’ services, e.g.) would be quite low because most of us pay for that care through the insurance company.

So the BLS needs to disentangle the cost of the medical care that we are buying indirectly from the cost of the embedded insurance products. The link above goes into more detail on all of this, but the bottom line is that once per year the BLS figures out what consumers paid for health insurance, how much of that was actually used by the insurance company to purchase health care, and therefore how much is attributable to the cost of the insurance product. Because they do this only once per year, and smear the answer over 12 months, you get step-wise discontinuities in the monthly figures. For many years this was not a big problem, but since 2018 there have been several fairly significant swings. The chart below shows the m/m percent change in health insurance CPI. You can see it went from stable, to +1.5% per month or so in 2018-2020, to -1% for 2020-2021, to +2% for 2021-2022, to -4% in the most-recent year.

That latest period has been a significant and measurable drag on the overall and core CPIs, and it was due to reverse starting with the October 2023 CPI released in November. Estimates were that it was going to be something like 2% per month, roughly. The change announced above introduces some smoothing so that these swings should be significantly dampened. The basic method doesn’t change, but it should be smoother and more-timely since the corrections will be every 6 months instead of every year. In order to make the new calculation method match endpoints, though, this means that starting in October, the +2%ish impact will bedoubled because the BLS will make the ‘normal’ adjustment but smear it over 6 months instead of 12, then transition to the new method.

The implication is that Health Insurance, which will have decreased y/y core CPI by about 0.5% once we get to October, will add 0.25% back over the 6 months ending April. So, we already know about a significant swing higher in core inflation that is coming soon. Take note.

The second item I want to note is M2. It’s a minor thing at this point, but after three months it is worth noticing that M2 is no longer declining. It isn’t a lot, as the chart below shows, but the three months ended April showed a contraction at a 9.6% annualized pace and the most-recent three months saw an increase at a 3.7% pace.

In the long run, 3.7% would certainly be acceptable but remember we still have some M2 velocity rebound to complete. What is interesting is that this is happening despite the fact that the Fed is continuing to reduce its balance sheet and loan officers are saying that lending standards are tightening. It may simply be a return to normal lending behaviors, with a gradual increase in loans that naturally accompany the rising working capital needs of a growing economy. Remember, banks are not reserve-constrained at this point, so they’ll keep lending. Anyway, I don’t want to make too much of 3-month change in the M2 trend, just as I was reluctant to make too much of those early M2 contractions…but this is what I expected to happen. I just expected it earlier. We will see if it continues. If it does, then that in concert with the natural rebound in M2 velocity means that further declines in inflation are going to be difficult, and we might even see some reacceleration.

Finally, the third item for today. In my podcast on Tuesday, I asked the question whether China’s recent sluggish growth, caused partly by its property bubble and overextended banks, meant that we should be looking at recession and disinflation in the US – which is the current meme being promulgated by many economists. I discussed the 1997-1998 “Asian Contagion” episode, and explained that a recession in a “producer” (net exporting) country hits the rest of the world very differently from a recession in a “consumer” (net importing) country like the US. A recession in consumer countries causes recession in producer economies, because the consumer economies are ‘downstream.’ On the other hand, a recession in producer countries can have the opposite effect on its customers – because, when an economy like China is in recession, that means it is providing less competition in the commodity markets that we also use. In turn, that means we can actually grow faster, all else equal.

This is what happened in the Asian Contagion episode, and I wanted to put some charts around that. The Thai baht was the first domino, and it collapsed in August 1997. It wasn’t until fears that the Hong Kong Dollar would de-peg from the USD, in October of that year – precipitating a 7% one-day drop in the Dow – that people in the West started getting very concerned and the Fed started citing troubles in the former Asian Tigers as a downside risk. Here are charts of the period. The first one shows quarterly GDP, which never increased less than 3.5% annualized; the second is median CPI, which was continuing a long period of deceleration from the 1980s prior to the crisis…but which began to accelerate in mid-1998.

The bottom line is that as long as our export sector is relatively small and as long we remain a developed consumer economy, weakness in producing economies is not a dampening effect for us but rather, if anything, a stimulating effect.