What's more important to savers: nominal or real rates?

Unlike most factories throughout the globe recently, the rumour mill has been working at full capacity this week.  The recent cuts to the Saudi budget and ministerial salaries brought with them conjecture that the persistently low price of crude oil had breached an important pain threshold, and that as a consequence a supply response would be forthcoming from OPEC.   Sure enough, yesterday's confab in Algiers confirmed a 750k bpd production cut.  While it remains to be seen whether this move eases the supply/demand balance- given that non-OPEC producers are also feeling the pinch, perhaps some may choose to fill the supply gap- it is nevertheless reasonable to think that the first cut is the hardest, and having crossed that particular Rubicon OPEC may follow suit again should the price of crude slide sharply again.

Meanwhile, in Europe just about every man, woman, and child have denied that the German government is going to step in to support Deutsche Bank.   So naturally, the story has emerged that in fact a rescue plan is being put into place, which could in extremis call for a government purchase of a significant stake in DB.  One can only imagine the reaction of the Italians if this were to occur.  Just remember kids:   in Europe, all animals are equal, but some animals are more equal than others.

Speaking of German banks and Italians, ECB president Mario Draghi testified before the Bundestag yesterday, touching on the sensitive issue of ECB policy and German banks.   Unsurprisingly, Draghi deflected blame for the travails of the European banking system, claiming that "many banks have been able to offset declining interest revenues..."  As a reminder, here's what the SX7E has done since the ECB imposed negative deposit rates on the Eurozone banking system:

Not exactly a home run, is it?   By way of comparison, the BKX in the US is roughly flat over the equivalent period.

Draghi also addressed the problem of low/negative yields for savers, repeating the modern macro orthodoxy that "what counts for savers is not the nominal interest rate but the real interest rate."   This has been repeated so often that it seems like it must be true, though it completely ignores the well-known phenomenon of money illusion (the tendency to anchor on nominal rather than real interest rates, and behave accordingly.)

Macro Man decided to do a little empirical investigation into how savers actually respond to shifts in real and nominal savings rates.  Unfortunately, the data history in Europe is not what we might like, and Macro Man was only able to lay hands on quarterly unadjusted savings rate data going back to the beginning of the last decade.   Still, we can do some digging and see if anything is there.   He chose to measure changes in the savings rate in conjunction with changes in deposit rates, reasoning that a decline in real rates should enccourage consumption and reduce the savings rate, whereas an increase in real rates should incentivize savings at the expense of spending.  He ran the numbers plotting y/y changes in real rates and savings rates (sadly, only available quarterly), and this is what he came up with:

As you can see, there generally does seem to be a relationship, and sure enough the correlation is positive, but not massively so; the R-squared is 0.10.  Still, it looks like the theorists are right; on this basis, a decline in real deposit rates is associated with a decline in the savings rate, and a rise with a rise.  If the theory is correct, we should see a weaker relationship with nominal rates:

Oh dear.   Oh dear indeed.   There are two relevant points about this chart, one of which is obvious and one of which is not.  The first is that the relationship has flipped; namely, lower nominal rates are associated with high levels of savings, and vice versa.  The second, which is not terribly apparent from the presentation of the chart, is that the statistical relationship is quite a bit stronger, with an R-squared of 0.24.  In other words, this appears to suggest that nominal rates exert a stronger influence on savings rates than real rates.

Of course, correlation does not necessitate causation.   In this case, given the paucity of the sample size, what we might actually be capturing is two factors that are impacted in similar ways by a single driver.  For example, periods of weak or negative economic growth tend to be associated with heightened uncertainty and thus higher savings...and also with easier monetary policy that drives down deposit rates.   In this case, it is not necessarily shifts in the deposit rate that impact savings behaviour, but the economic weakness that drove the reduction in the deposit rate to begin with.

At the very least, however, we can say that the study is inconclusive that real rates are more important for impacting savings behaviour than nominal rates.  To dig deeper, Macro Man switched gears and shifted his gaze to the US, where there is a long history on monthly personal savings rate data.   There is not, unfortunately, the average household deposit rate series that exists in the Eurozone, so he used 3 month T bill rates as a proxy for bank depo rates.

To control for the economic cycle, he also looked at shifts in the unemployment rate and their relationship to shifts in the savings rate.  He ran a series of regressions over different time periods to get a sense of the relative importance of shifts in the unemployment rate and nominal/real T-bill yields in explaining shifts in the savings rate.  (Long time readers may recall that shifts in nominal spending can be explained largely by changes in nominal income, and also a bit by changes in nominal wealth.)

The results are presented below, leading off with a time series chart of the four series since 1960 Unsurprisingly, it's largely a mishmash that's difficult to discern any sense out of.

However, the regressions yielded some quite interesting results.    As noted above, Macro Man regressed 1y changes in the personal savings rate with changes in nominal and real T bill yields, and also with  changes in the unemployment rate.  He ran the numbers over three time periods:  1960 to present, 1990 to present, and 2000 to present.   A summary of the results is set out in the table below.

As you can see, over the entire sample the unemployment rate has by far the most explanatory power over changes in the savings rate, with the largest coefficient and t-statistic and a high degree of significance.   Interestingly, changes in nominal yields are also statistically significant, but with a positive coefficient.   In other words, higher rates have been associated with higher savings, and lower rates with lower savings.  Economic orthodoxy wins!   Well, maybe.   Observe that there is essentially zero relationship between changes in real T bill yields and the savings rate over the last 56 years.   Remind me again why the economics profession has the horn for R*?

If we shorten the time scale, a fascinating change occurs.   Looking at data since 1990, we find that changes in nominal rates bear the strongest relationship to changes in the savings rate, but with the same negative coefficient that we saw in Europe!  Real yields also carry some explanatory power, with the expected coefficient.  The unemployment rate, meanwhile, falls just outside the realm of statistical significance.

Since the 2000, the negative relationship between nominal yields and savings has strengthened further, with the highest coefficient of the entire study.  Interestingly, the significance of real yields, while still important, fades when the 1990's are stripped out, and the unemployment rate is deemed utterly irrelevant.

So over the last two and a half decades, data from the US corroborates what we saw in Europe; shifts in nominal yields are more closely related to those in the savings rate than are real yields, and in the opposite direction expected by mainstream economic thought.  To highlight the difference in another way, Macro Man ran rolling 5 year correlations of the one year changes in nominal and real yields with those in the savings rate.

You can see the shift that seemed to occur in the relationship between nominal yields and savings in the early 1990's and has prevailed for most of the period since.   It has endured sufficiently long that it is difficult to call it a temporary aberration.   Ironically, that 5 year correlation has recently flipped positive (suggesting higher T bill yields should increase the savings rate), though Macro Man finds it difficult to place much faith in that development, given that nominal yields flat-lined for five years while the savings rate whipped around, as you can see below.  It is interesting to note, however, that the savings rate has declined 0.4% since the Fed hiked rates last December, consistent with the findings since 1990.

So is there an explanation for the apparently bizarre finding that nominal rates have a stronger (negative) relationship with savings rates than real rates do, contrary to theory?   Macro Man has an idea, which can be summed up in one word and one chart:  demographics.

As you can see, the median age in the United States was largely unchanged from 1960 to 1980, increased slightly by 1990, and rose steadily in the 90's and 2000's.  The median age is now roughly eight years older than it was in 1980.  Perhaps not coincidentally, Europe has also seen aging, and indeed is generally a much older society than the United States.

The causality between age and sensitivity to interest rates may not be particularly difficult to fathom.  As society ages, citizens pursue an asset allocation geared more towards fixed income, thus increasing the sensitivity to interest and deposit rates.   Older citizens heavily geared towards fixed income start with the premise of a desired level of interest income and work their way backwards.  The higher the nominal rate, the less principal must be allocated to fixed income investments to generate that income, and the more can be allocated to other things, including consumption.  The lower the nominal rate, meanwhile, the greater the principal must be allocated to fixed income...and the more risk averse the person will be, knowing that there is little cushion from such low interest rates.

While inflation and real rates of return impact the level of desired income, it seems reasonable to expect that those expectations and budgets are much slower to adjust than reacting to nominal interest income and its prospects, which are available with every monthly statement.

Interestingly, the findings presented here appear to corroborate those of an earlier IMF study performed on China, which is summarized here.  China, of course, faces its own demographic challenges which may explain the result there.

While this study was relatively limited in scope, the result was pretty clear.   In each iteration of the analysis performed in Europe and the US, changes in nominal interest rates exhibited a stronger statistical relationship with changes in savings behaviour than did changes in real interest rates.   More to the point, in a number of countries the data appear to suggest that lower rates actually encourage consumers to hoard savings rather than spend it.  The implication is that from a consumer's perspective, at least, the ongoing policy fetish with zero or negative interest rates is counter-productive, in line with a neo-Fisherian model of the world.

Sadly, prospects do not look good for getting the adherents of policy orthodoxy to challenge their accepted view of the world.   After all, as an exercise in science economics is pretty dismal indeed.