Today’s ‘Amazon Effect’ Is Not As Large As Yesteryear’s ‘Walmart Effect’ – Yet

Obviously, one of the most contentious debates in the econ sphere of late has been the extent to which the Fed and other DM central banks are relying on outdated models and assumptions when it comes to measuring and/or forecasting inflation.

This debate has taken center stage in 2017 for one very simple reason. Policymakers have realized that persisting in accommodation carries very real risks in terms of inflating asset bubbles and so, they’d like to normalize – or at least they’d like move in that direction if for no other reason than to free up some countercyclical breathing room so they’re not completely out of ammunition when the next crisis rolls around.

The problem is that inflation is still stubbornly subdued (or at least by their measures). That makes it difficult to use “data-dependency” to justify normalization. Of course you don’t want to admit that you’re more “market dependent” than you are data-dependent because not only does that open you up to criticism (i.e. finger pointing when the asset bubbles inflated by QE, ZIRP, and NIRP finally do burst), it also means that implicit in any normalization push is the idea that policymakers are worried about asset prices. Merely conveying that worry could be enough to tip the first domino on the way to a panic.

So what have policymakers done? Well, they’ve described lackluster inflation as “transitory.” Indeed, Yellen’s contention this month that she and her colleagues do not fully “understand” the causes of weak inflation was the source of more than a few jokes and memes, like this one:

Our contention is that the this “lack of understanding” is to a certain extent feigned. Policymakers don’t want to come out and say “we’re worried about asset bubbles, so we’re going to go ahead and normalize, weak inflation be damned.” So what they say instead is “well, the factors weighing down inflation must be transitory.”

But when you say that, you send everyone else off on a hunt to explain why deflationary forces are not in fact transitory, but rather endemic and structural. And if there were a dictionary entry for “structural deflationary force” this would be the picture next to it:

Clearly, Bezos and Amazon have ushered in an era that will be defined by price wars. That’s readily apparent. But not everyone is convinced that the impact in terms of disinflation is as yet something to be particularly concerned about. One interesting question is whether the epochal shift to E-Commerce has had as big an impact on inflation as the shift to Big Box did.

Focusing on pure E-Commerce players, the share of total retail sales is growing by something like 0.3-0.4pp per year (again, for internet-only retailers):

As Goldman writes, in a note out Saturday, this shift can reduce consumer prices in two ways.

1. if online retailers sell similar goods at lower prices than brick-and-mortar stores, the average price paid by consumers declines as online penetration increases
2. if increased online competition prompts traditional brick-and-mortar stores to lower their prices, this holds back overall inflation.

How big are these effects? Well, as to the first effect, Goldman says this:

It’s essentially equal to the change in the online share multiplied by the average price gap between the online and offline sector. According to a study by the Billion Prices Project, this gap averages around 6% in the case of goods sold by Amazon and offline retailers. Combined with the 0.3-0.4pp per year increase in the online share shown in Exhibit 1, this implies a drag on (true) goods price inflation of only about 0.02pp, a fairly negligible impact.

And as to the second:

The second effect is potentially larger, though quite uncertain. To estimate it, we recently conducted an analysis relating inflation rates across detailed PCE categories to changes in the online share in each category. In this analysis, we included both category fixed effects (which control for differences in the long-term price trends in each category) and time fixed effects (which control for the ups and downs of overall inflation). In this analysis, we also included PCE service categories, allowing the coefficients to differ between goods and services. As shown in Exhibit 2, our point estimate is that a 1pp increase in the online share holds down inflation by 0.6pp in the goods sector and 0.04pp in the service sector. However, neither estimate was quite statistically significant, i.e. the uncertainty is large.

When you add all of that up, you get a back-of-envelope calculation that suggests “the shift to online shopping is subtracting 0.25pp from core goods inflation and 0.1pp from overall core PCE inflation”:

So that’s “not nothing” (so to speak), but guess what? It’s not on par with the Walmart effect.”Today’s Amazon effect looks, if anything, somewhat smaller than yesterday’s Walmart effect,” Goldman goes to write, before showing you the following chart which depicts the overall retail market share of selected big box retailers:

As the bank notes, “the headline number shows annual gains of up to 1pp for the big box stores in the 1995-2007 period, a number more than double the increase in the internet-only share in the 2015-2017 period.”

Goldman completes their analysis by estimating that the Walmart effect “subtracted around 0.5pp from core goods inflation and 0.15-0.2pp from overall core PCE inflation.”

The takeaway: when it comes to structural disinflationary dynamics, today’s “Amazon effect” is not in fact as large as the “Walmart effect” of yesteryear.

Now let’s revisit this analysis five years down the road and see what the picture looks like.