hell is other robots
by John at 7/28/2004 04:14:00 PM
So, this morning Boing Boing (natch!) pointed me to a Reuters article that said that the more a nation's people believed in Hell, the lower corruption was and the better the standard of living. "Crap! Hard data to prove that atheists are asses," I thought, "but, at least I can take comfort in the old adage that correlation does not imply causation." Then I went on with my day.
Later, Boing Boing pointed to the original St. Louis Fed study that the Reuters article referred to. I looked it over, and curiously, they never plotted belief in Hell vs. per capita GDP. Sure, there was a correlation between corruption and belief in Hell (a weak one) and a strong correlation between corruption and GDP. So, the study says, belief in Hell must be correlated with GDP. But I was suspicious. Luckily, the study provided an Excel file with the data. I figured I'd take a look at it later.
Then, checking Boing Boing once again, someone else (Morgan Foust) had done the work for me. He said the the belief in Hell / GDP correlation was actually negative (but small)!
So I checked it. Foust is right. Here's my chart.
The data is a little confusing, and this may be where those crack economists made their mistake. GDP rank is 1 for highest and 35 for lowest, and Belief in Hell rank is 1 for lowest and 35 for highest. Clearly, there is a weak correlation for strong belief in Hell and low GDP. Just the opposite of what the study says.
I didn't like the ranking data (that's no way to fit a line!), so I went to the base data of % belief in Hell and adjusted per capita GDP, but I got the same result. High belief in Hell corresponds to low GDP.
(Actually, I'd say there might be two, if not three, groups of data on this chart. A significant positive correlation only appears to exist for one of the groups. Maybe I'll check up on that later.)
Pretty sloppy work, Federal Reserve. But, I guess that we can all take comfort, at least, in the adage that correlation does not imply causation. There's hope for us yet!
Update: Looking back at the St. Louis Fed study, the ranking thing is where they screwed up. From their charts, if high corruption corresponds to low GDP (this is the common wisdom), then Corruption rank goes from 1=most to 35=least (similar to the GDP rank), and low belief in Hell weakly corresponds to less corrupt. They forgot that they ranked belief in hell from least to most, not most to least.
On top of that, fitting a line to a ranking is often all but meaningless, because it removes the structure of the data.
In addition to being in the opposite direction to what they study says, the GDP correlations with belief in Hell are also so weak as to be meaningless (if I recall my stats, they fail a hypothesis test).
Update 2: There really are three groups, and they have some geopolitical sense. Group 1 is "W. Europe" (except Ireland, plus Japan) where GDP doesn't correlate with Hell. Group 2 is "Third World" (sorry Chile, nothing personal) where GDP doesn't correlate with Hell. Group 3 is "E. Europe, Mediterranean, and N. America" (except Mexico, plus Ireland) where GDP does correlate to Hell. Except for Ireland and the US, Group 1 has as high or higher GDP than Group 3. Turkey is an outlier on the list, and is the only predominantly Islamic nation represented (I think Nigeria is half-and-half). I've broken down the groups here.
Later, Boing Boing pointed to the original St. Louis Fed study that the Reuters article referred to. I looked it over, and curiously, they never plotted belief in Hell vs. per capita GDP. Sure, there was a correlation between corruption and belief in Hell (a weak one) and a strong correlation between corruption and GDP. So, the study says, belief in Hell must be correlated with GDP. But I was suspicious. Luckily, the study provided an Excel file with the data. I figured I'd take a look at it later.
Then, checking Boing Boing once again, someone else (Morgan Foust) had done the work for me. He said the the belief in Hell / GDP correlation was actually negative (but small)!
So I checked it. Foust is right. Here's my chart.
The data is a little confusing, and this may be where those crack economists made their mistake. GDP rank is 1 for highest and 35 for lowest, and Belief in Hell rank is 1 for lowest and 35 for highest. Clearly, there is a weak correlation for strong belief in Hell and low GDP. Just the opposite of what the study says.
I didn't like the ranking data (that's no way to fit a line!), so I went to the base data of % belief in Hell and adjusted per capita GDP, but I got the same result. High belief in Hell corresponds to low GDP.
(Actually, I'd say there might be two, if not three, groups of data on this chart. A significant positive correlation only appears to exist for one of the groups. Maybe I'll check up on that later.)
Pretty sloppy work, Federal Reserve. But, I guess that we can all take comfort, at least, in the adage that correlation does not imply causation. There's hope for us yet!
Update: Looking back at the St. Louis Fed study, the ranking thing is where they screwed up. From their charts, if high corruption corresponds to low GDP (this is the common wisdom), then Corruption rank goes from 1=most to 35=least (similar to the GDP rank), and low belief in Hell weakly corresponds to less corrupt. They forgot that they ranked belief in hell from least to most, not most to least.
On top of that, fitting a line to a ranking is often all but meaningless, because it removes the structure of the data.
In addition to being in the opposite direction to what they study says, the GDP correlations with belief in Hell are also so weak as to be meaningless (if I recall my stats, they fail a hypothesis test).
Update 2: There really are three groups, and they have some geopolitical sense. Group 1 is "W. Europe" (except Ireland, plus Japan) where GDP doesn't correlate with Hell. Group 2 is "Third World" (sorry Chile, nothing personal) where GDP doesn't correlate with Hell. Group 3 is "E. Europe, Mediterranean, and N. America" (except Mexico, plus Ireland) where GDP does correlate to Hell. Except for Ireland and the US, Group 1 has as high or higher GDP than Group 3. Turkey is an outlier on the list, and is the only predominantly Islamic nation represented (I think Nigeria is half-and-half). I've broken down the groups here.
» Post a Comment