This is an interesting analysis and can only really be accurate if we assume the level of mapping and number of mappers in an area is the same from area to area. I don't think this is true.
However, its possible statistics to find the number of mappers and the level of activity in an area from OSM, so a better analysis should normalise for this too.
But! This only considers OSM from the point of view of a "hobby mapper" - the old school view of individuals going out with GPS units. Today a growing proportion mapping is contributed by governments, open data sets, batch imports, businesses and mapping teams, all of which vary by geography too.
The fact that something is mapped at all is something of an economic indicator, since OpenStreetMap mappers do the heavy lifting, are likely to be big nerds with free time / retired, and are going to map their local areas preferentially.
To a degree this is true, but they do find the abundance of some "amenities" negatively correlated with national income, some of which, like "marketplace" make more sense than others (like "hospital", at least until you consider not all hospitals are equal and the state of the art ones tend to be big...). When it comes to ATMs being more commonly depicted in lower income countries (within the sample) it's an interesting question whether that's because they actually are more common or whether they're more commonly marked on OSM because locating them is more important to people in a cash economy (cf most of the ATMs near me are unmarked)
>like "hospital", at least until you consider not all hospitals are equal and the state of the art ones tend to be big...
I'm willing to bet this and possibly some of the other amenity correlations are driven either by a few outliers, some uncontrolled confounding variables, or by measurement error.
E.g. if one of the poorer nations has a strange policy or linguistic quirk around the definition of hospital, that could be driving up the total number of locations marked as hospital and thus introducing some hidden bias.
In terms of confounding variables, I'm just taking a guess in the dark, but hospitals per capita is probably a strong proxy measure of how rural a country is. The more rural the population, the more hospitals are required to serve the same number of people, because hospitals need to be close for emergency situations.
Also, there's probably a selection bias. Hospitals are likely one of the first things to get put on OSM for a given area. Because there aren't many of them and they tend to be one of the most important items people are looking for on a map. So poorer countries, with fewer OSM users, are more likely to have hospitals marked relative to the other amenities.
Meaning, you could have a negative relationship between all amenities and GDP/GNI/HDI, but only hospitals and other items consistently marked across all countries is being measured well enough to demonstrate this. Even though naively one might assume more amenities/person is better, it's possible for many amenities that centralization/consolidation actually correlates with better economic performance. Obviously that's probably not true of park benches, but it might be for schools, hospitals, and other public infrastructure.
The idea of using (mappable/recordable) public amenities per capita as a measure of prosperity is good.
However, even discounting for the massive regional biases in OSM mapping, having 10 universities per capita doesn't mean anything. The qualitative part is extremely important on all aspects. Universities, hospitals, parks, even benches..
Using aerial photography you can probably infer things about the quality of the roadwork and efficiency of transportation, parking space availability, recreational park "quality", rooftop utilization etc. But with incomplete/heavily biased mapping and no quality index this is pretty much useless.
As it stands, all metrics include a hidden “quality of OpenStreetMap data” feature that may explain significant fractions of the significance of features.
So, I think they should try to compensate for that. How? I wouldn’t know.
It may be more interesting to work in the other direction, where the comparison with the economic data is used to estimate the quality of the OSM data.
However, its possible statistics to find the number of mappers and the level of activity in an area from OSM, so a better analysis should normalise for this too.
But! This only considers OSM from the point of view of a "hobby mapper" - the old school view of individuals going out with GPS units. Today a growing proportion mapping is contributed by governments, open data sets, batch imports, businesses and mapping teams, all of which vary by geography too.