Effect size matters

Mark Thoma:

People who believe consumption is driven, in part, by the consumption of neighbors and friends will like these results [open link].

What does it mean to believe consumption is driven in part by consumption of those people around you. This seems obviously true. I won’t even know certain products exist without observing the consumption habits of my friends and neighbors. I suppose, then, I believe consumption is driven in part by the consumption of others. I suspect, given the first link is to a discussion of Robert Frank’s book Falling Behind, Thoma is talking about something more. He means that people care about status and keeping up with the Joneses; they care about consumption not because of the utility it provides directly but because it makes them higher status. Status is a zero-sum game, so this is a bad, bad, bad thing.

Frank doesn’t believe, or at least his narrative suggests as much, that consumption is in part driven by status seeking. Frank’s book makes the argument that most consumption is status consumption. He isn’t claiming the existence of status seeking; he’s claiming it is a substantial economic issue. In econometrics, this distinction is the one between statistical significance and economic significance.

The paper Thoma points us to shows evidence for the existence of status seeking, but it doesn’t show a big effect size. Neighbors of people that win new cars are 5% more likely to buy a car themselves. This is a statistically significant result, but is it economically significant?

Well, 5% increase in likelihood to buy a car is significant in money terms. If a winner has 10 neighbors and new cars cost $30k, then we’d expect $15k worth of cars bought. One might be tempted to aggregate this up, but Thoma suggest the problem with doing this:

That is, when everyone wins the lottery, the social effect may have more impact on the type of spending than on the total amount that households spend.

If everyone’s winning cars, then nobody will need to keep up with the Joneses. Even still, this large money value seems like a large effect.

The question, though, is how much does status consumption matter relative to non-status consumption. Calculating the money value — the standard way for economists to calculate “economic significance” — doesn’t seem to get at this issue. The way I see it, a 5% effect suggests status consumption doesn’t matter that much. The study showed when a neighbor increases his status by a $30k car (a substantial increase in status I’d think) this only induces an increase of status spending by $1.5k. It seems if status mattered, that number would be closer to $30k.

Statistical significance always has us comparing the estimated effect against zero. If its very unlikely the effect is zero, then we say the effect is statistically significant. Testing economic significance, on the other hand, doesn’t require us to compare the effect against zero. For something to be economically significant it has to deviate from the proper counterfactual. This counterfactual can be given by gut instinct or by economic theory, but in either case the counterfactual may be far from zero.

My gut tells me “status consumption matters” means we should expect an effect size in this case close to 100%. Your gut may tell you something different, but it seems unlikely zero is what Frank’s gut tells him.

7 thoughts on “Effect size matters”

  1. Could a status purchase have a higher value for the first few days? What’s the delay between the gift and the increase in status spending?

    “Hey Bob, congrats on the Lexus you won last month, but check out my new big screen TV!”

  2. I’m just wildly speculating here: maybe the status value of a luxury item isn’t pegged to its market value, and maybe it drops precipitously with time, varying with the initial value and type of item. It seems plausible that, say, a brand new, impossible-to-get must-have $3000 Carbon Neutral Edition OLED big screen TV has a higher status value (for a little while) than a five year old luxury car worth 20x as much.

  3. The 5% likelihood is cumulative from the time of the win. If the urge to status consume is greatest when you’re exposed to a competitors status consumption then the biggest this effect can be is 5% (with 0 effect after).

  4. Ah – I read that as a 5% increase in status spending, not a 5% increase in likelihood of a big purchase. My bad.

Comments are closed.