Disaggregating the economy: online prices

Posted on by

At the AEA session on measuring well-being, Austan Goolsbee and Pete Klenow write in an abstract,

We use transactions-level data from Adobe Analytics to analyze inflation online vs. offline. Online inflation from 2014-2017 period averaged about 100 basis points lower than inflation in the CPI for the same categories. Entry and exit of new product varieties is extremely important in most online categories (but less so in food and grocery). Data on quantities is particularly important because the entry and exit rates vary with product sales. The increased variety of products sold online implies an additional 100 basis points of lower inflation than in the matched model/CPI style indices.

In my view “the” inflation rate is an increasingly dubious concept. People can argue forever about whether true inflation is really much lower or much higher than what the CPI indicates.

Because we do not know the true inflation rate, we do not know the true real interest rate or the true growth rate of real wages. Not to mention the fact that there is no such thing as “the” wage rate, given the divergence in wage behavior across different categories of workers.

Consumers’ Surplus and well-being

Posted on by

Another paper from the AEA session on measuring well-being. The abstract of the paper by Erik Brynjolfsson, Felix Eggers, and Avinash Gannamaneni says,

In principle, changes in consumer surplus (compensating expenditure) provide a superior measure of changes in consumer well-being than GDP and metrics derived from it, like productivity, especially for digital goods. In practice, consumer surplus has been difficult to measure. We demonstrate the potential of massively scalable online Single Binary Discrete Choice experiments for addressing this issue. These experiments provide a measure of consumers’ willingness to accept compensation for losing access to various digital goods and thereby estimate the changes in consumer surplus from these goods. Drawing on several hundred thousand online experiments, our results indicate that digital goods have created substantial gains in well-being which are largely missed by conventional measure of GDP and productivity, and suggest that our approach can be scaled up to a broader set of goods and services. Two limitations of our methods are that they are much less precise than changes in GDP and they suffer from hypothetical bias. We show how much of an improvement in precision can be achieved with larger sample sizes and demographic controls and we document the direction and magnitude of hypothetical bias by conducting incentive compatible experiments with a smaller group of subjects. By periodically querying a large, representative sample of goods and services, including those which are not priced in existing markets, changes in consumer surplus and other new measures of well-being derived from these online choice experiments have the potential for providing cost-effective supplements to existing national income and product accounts.

A Social Progress Index

Posted on by

At this year’s AEA meetings, there apparently was an interesting session on measuring well-being.

One paper, by Daniel Fehder, Scott Stern, and Michael E. Porter, says

we describe the construction of a synthetic measure of non-economic performance, the Social Progress Index (SPI). Building on a wide range of prior literature, it incorporates more than 50 indicators into 12 components that are then aggregated into three primary dimensions of non-economic societal performance: Basic Human Needs, Foundations of Wellbeing, and Opportunity.

…
overall social progress is decomposed into three distinct dimensions, Basic Human Needs (“Does a country provide for its people’s most essential needs?), Foundations of Well-Being (“Are the building blocks in place for individuals and communities to enhance and sustain wellbeing?), and Opportunity (“Is there opportunity for all individuals to reach their full potential?”). Whereas Basic Human Needs centers on non-economic conditions that a society provides (e.g., achieving a low child mortality rate and a high level of sanitation, shelter, and personal safety), Foundations of Wellbeing focuses on whether a society offers individual an opportunity to invest in themselves and their communities to advance their wellbeing (e.g., allowing individuals to achieve a basic level of education, gain access to information, and maintain strong lifelong health and local environmental quality). Finally, Opportunity focuses on those components of social progress that concern the ability of individuals to achieve their own personal objectives, including their degree of personal rights and freedom in the context of an inclusive and educated society.

I like the idea of diversifying the portfolio of economic and social indicators. Recall my recent essay proposing to measure occupational satisfaction.

The back-sleep ideologues

Posted on by

This story says,

The American Academy of Pediatrics, or AAP, recommends that babies always be placed on their backs to sleep, even for just a nap.

I think this is worst advice ever. I don’t find convincing the evidence that this reduces death of infants. I am convinced instead that it leads to widespread sleep problems among babies at least until age four, it slows their motor development and probably their cognitive development, and leads to many children wearing helmets to reshape their heads.. A friend of mine quoted a pediatric physical therapist as saying, “The back-sleepers keep me in business.”

Of course, it is not typically rational for a non-expert to challenge orthodoxy. So call me crazy.

Null Hypothesis Watch

Posted on by

Two papers that claim to reject it.

1. Michael Lovenheim and Alexander Willen write,

We see consistent evidence that 12 years of exposure to a collective bargaining law negatively impacts both cognitive and noncognitive scores among men. AFQT percentile declines by 10.2, a 20.9% effect relative to the mean.

See also the abstract, quoted by Tyler Cowen.

2. Michael Gilraine, Hugh Macartney, and Robert McMillan write,

California’s statewide class size reduction program of the late-1990s. . .caused marked reductions in local private school shares, consequent changes in public school demographics, and significant increases in local house prices — the latter indicative of the reform’s full impact. Second, using a generalization of the differencing approach, we provide credible estimates of the direct and indirect impacts of the reform on a common scale. These reveal a large pure class size effect of 0.11 SD (in terms of mathematics scores), and an even larger indirect effect of 0.16 SD via induced changes in school demographics. Further, we show that both effects persist positively, giving rise to an overall policy impact estimated to be 0.4 SD higher after four years of treatment (relative to none).

I am skeptical of both papers. I am not convinced that the methods used truly eliminate possible confounding factors. But I have not read either paper closely.

Moderate voters?

Posted on by

James Taranto (WSJ) writes,

Those old enough to remember the decades before the ’90s, then, may tend to see permanent majorities around the corner because they expect a return to normalcy. Mr. Fiorina, by contrast, argues that frequent shifts in political control are now the norm because of the way the parties have changed. He rejects the common view that American voters are “polarized.” Instead, he says, the parties have become polarized, in a process he calls the “sorting” of the electorate.

So we have parties captured by extremists, and voters trying to find the moderates. Possibly related: Nassim Taleb’s forthcoming book.

The Tyranny of Metrics

Posted on by

The book by Jerry Muller will be out shortly. It makes a strong case against the over-use of quantitative measures to fix compensation. Education is one example.
If you believe the null hypothesis, then compensating teachers based on outcomes only introduces randomness into their pay.

Meanwhile, without referring to the book, in talking about health care and education, Megan McArdle writes,

So when we measure outputs, we are getting at best a very distorted picture of the value of the services provided. Modern industrial management is simply not designed for this sort of situation. If you feed human inputs into a machine system, you are quite likely to grind up the humans in the process.

Read the whole thing.

Resistance Watch

Posted on by

Charlie Stross writes,

However, Facebook is trying to get eyeballs on ads, as is Twitter, as is Google. To do this, they fine-tune the content they show you to make it more attractive to your eyes—and by ‘attractive’ I do not mean pleasant. We humans have an evolved automatic reflex to pay attention to threats and horrors as well as pleasurable stimuli: consider the way highway traffic always slows to a crawl as it is funnelled past an accident site. The algorithms that determine what to show us when we look at Facebook or Twitter take this bias into account. You might react more strongly to a public hanging in Iran than to a couple kissing: the algorithm knows, and will show you whatever makes you pay attention.

Pointer from Tyler Cowen.

Trigger warning: lots of smug rhetoric presuming that the left is correct on climate change, net neutrality, financial regulation, etc.

Looking ahead, Stross writes,

Your phone will be aware of precisely what you like to look at on its screen. With addiction-seeking deep learning and neural-network generated images, it is in principle possible to feed you an endlessly escallating [sic] payload of arousal-maximizing inputs. It might be Facebook or Twitter messages optimized to produce outrage, or it could be porn generated by AI to appeal to kinks you aren’t even consciously aware of. But either way, the app now owns your central nervous system—and you will be monetized.

One key point on which I agree with Stross is that I am surprised and disappointed that of all of the possible ways to pay for content on the Internet, the advertising model dominates. I understand why micropayments did not take off–Clay Shirky diagnosed the “mental transactions costs” involved. But if the subscription model (what I called “clubs” in my essays from twenty years ago) were dominant, then the interests of consumers and content providers would be better aligned. With the advertising model, the relationship is necessarily adversarial. The content provider needs to grab and hold your attention, whether that works to your benefit or not. Bad consequences follow.