Hacker News Reader

Comment by ZeroGravitas | Hacker News Reader

I want to agree with this conclusion, but it feels like the GDP being accurate, and the GDP not being a good reflection of citizen welfare is both more likely and more telling.

I used to love the Economist but they appear to have gone downhill more recently and those graphs seem designed to make a specific point rather than present data objectively.

edit: after fighting my way past the paywall, this was poorer than I suspected:

> Assuming that the most democratic countries reported growth figures accurately, he then used the satellite data to estimate if other countries under- or over-stated theirs.

So they just graphed the assumption as if it was data, rather than an assumption.

Though now I'm wondering why Australia isn't considered democratic enough to have their GDP taken as correct.

View on HN

Assume (I don't have an Economist sub) it's based off this paper https://bfi.uchicago.edu/wp-content/uploads/2021/07/BFI_WP_2... Others include this from 2015 https://sci-hub.se/10.1111/isqu.12143

What the data explicitly shows is that night light intensity increased faster than GDP increased in countries classed as "free" than countries classed as "non-free". (the model itself makes no assumptions that the most democratic countries report growth figures accurately. An alternative interpretation equally compatible with the data is that democratic countries deliberately underreport their GDP growth, but that's a lot sillier than assuming autocracies lie, or exaggerate more)

The obvious perturbing factor is that countries classed as "non-free" also tend to start off poorer and more inegalitarian (the link between the two is generally assumed causal, but both ways causal) so the oilfields raking in cash doesn't necessarily translate to more electricity for the average citizen, and the countries classed as non-free are on average growing faster (even after taking into account the estimated inflation of the GDP figure) and since it's not estimated on a per capita basis the GDP figure grows through population growth. Economists can try to control for this, but controls aren't perfect. The "free" countries are also a relatively homogenous group of countries which started relatively wealthy and experienced modest GDP growth with relatively little disruption or demographic change.

Satellite light data also isn't a particularly great proxy for citizen welfare. It's an indicator as it's broadly proportional to increased access to electricity (and its regional distribution) and light intensity growth correlates well with GDP growth. But those "free" countries tended to have affordable household lighting everywhere anyway; the extra light being pumped into mall car parks and advertising hoardings and bigger traffic jams are pretty loosely linked to the welfare improvements we've experienced since 1991.

Yes, I agree, you can get a more nuanced take on what I'm getting at by looking at papers that just ask the question: "Can we measure GDP from space?".

And the answer is, yeah kind of, but with various caveats.

And looking at those that ask "can we measure human welfare with GDP?"

And the answer is, yeah kind of, but with various caveats.

This paper builds on both of those and somewhat sweeps the caveats under the rug to make a point about autocracies.

It feels like a PR pitch for "free markets", which I broadly agree with, but still don't want to see badly drawn graphs of.

I think it was the naming of only a few countries that really set of my "how to lie with graphs" alarm.

> So they just graphed the assumption as if it was data, rather than an assumption.

Many studies make an assumption and then explore the implications of that assumption. They are not claiming that democratic countries have accurate official GDP figures. They are simply saying: if democratic countries' GDP figures are accurate then the correlation factor between GDP and light output is X. If the correlation factor between GDP and light output is X for all countries (another assumption) then the discrepancy between expected GDP and official GDP is Y. That's all they are saying, they are not presenting made up data as facts.

Not only that the GDPR must be reported in the same way but also that the system be comparable/generalisable. The free country correlation pretty much is the 'traing error' here. If there is e.g. a systematic bias with light and freedom one also might have a problem. Also the pretty much assume a linear model as the GDPR is much higher in the 'free' world. Still it is actually good to have such a proxy measures. Wonder if authoritarian regimes will now have dedicated lighting development strategies...

It might be useful to note that the Economist did not do this research. This is a summary of a study by an academic economist:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3882062

It's likely your questions are answered therein.

EDIT: Skimming it briefly it seems like section 4, equation 4 is relevant to your criticism. It seems to be allowing for some level of independence in the internal relationship between a country's gdp growth and light, and trying to capture the parameter of interest in part fromm the within-country variation over time.

The chart (currently; it might have been updated) does show both reported GDP and light-at-night inferred GDP for "free" economies. The article doesn't say explicitly, but we must assume they're using a linear mapping between the two, since there are minor discrepancies in the "free" economies (as one expects with necessarily noisy data).

And also the question is what data from satellite?

Electricity usage may not be all encompassing indicator; Ex. increase in cultivated area, or more production from same land parcel (by using fertilizers, pesticides, modern techniques) will not show up as corresponding increase in electricity usage at night.

I believe the state of the art in these things do mix multiple different types of satellite data, night time lights is just the longest dataset we have since even crappy old satelittes could capture it.

Several parallels with the various reconstructions of earth's atmosphere via proxies like tree rings and ice cores and its later recalibration with satellite measures.

I was expecting a "light intensity vs GDP" graph to see if they really correlate.

> it feels like the GDP being accurate

This has been my personal gripe for a long time now! Imagine that I needed to borrow £1000 from my mum, she lends me that money, and then I return that much back to her shortly after; therefore generating £2000 in GDP. It excludes the reason why I needed to borrow that money, ie it could be that I am in dire economic situation, and could not pay for essentials, therefore this GDP growth is due to poverty rather than growth!

Additionally, in the US, there is a thing called "imputed GDP", which is essentially GDP that is not realised due to an owned property being lived in and not being rented out. This generally accounts for around 15% of US's GDP https://www.bea.gov/help/faq/488 I guess companies such as Black Rock seem to be doing good for the GDP by buying people's properties and then renting it back to them.

tl;dr I am also not a fan of purely looking at GDP as an accurate metric for the economy.

> I'm wondering why Australia isn't considered democratic enough to have their GDP taken as correct.

Australia is clearly shown in the "Free Countries" graph.

I agree with your criticism. I want to add one more: GDP is a measure of activity. Measuring lights on satellite images is a state. So somehow this study is correlating stock (# of lights) to a flow (economic activity). Seems like a pretty basic comparison error.

There have to be a gazillion other caveats - something stinks about this study. This content sings to many of their readers who cant moss an opportunity to tell others that china’s books are cooked.

This doesn't measure lights (stock). It measures light (the emission). This is an active process of the stock of lights, emitting light, and might correlate closely with other economic activity.

Reply on news.ycombinator.com