Several of my friends have been sharing screenshots of this Twitter post:

Time for your annual reminder that, according to A Christmas Carol, Bob Cratchit makes 15 shillings a week. Adjusted for inflation, that's $530.27/wk, $27,574/yr, or $13.50/hr.

Most Americans on minimum wage earn less than a Dickensian allegory for destitution.

  —Chris Thompson

It's wrong, but it's wrong in an interesting way!

A Christmas Carol came out in 1843. The Bank of England Inflation Calculator, which incredibly goes back to 1209, estimates £1 in 1843 is equivalent to £130 in 2020. Since there were 20 shillings to the pound, 15s/wk in 1843 makes £98/wk in 2020. Cratchit would have been working six days a week at perhaps ten hours a day, so we get £1.63/hr or $2.15/hr. The post gave $13.50/hr; where's the factor of six coming from?

There are multiple ways to figure inflation. The most common one, and the one I use above, is purchasing power: what can you buy with your £? While I can't replicate Thompson's number exactly, I think he's instead using using labor value: how many hours of someone's time can you buy? 15s/wk in 1843 is, deflated by the cost of labor, £611/wk in 2020, or $807/wk. For a 60hr week that's $13.45/hr.

The problem is, no matter how much richer we get, using the cost of labor to adjust worker pay for inflation will always show that workers are paid the same amount. It's saying "Cratchit was paid X% of a typical wage for 1843, which you can think of as if he were paid X% of a typical 2020 wage, which is $13.45/hr." This sort of adjustment makes sense if you're trying to understand the social position of the Cratchit family, but since what we care about here is their absolute level of poverty, purchasing power is the right measure.

(Thanks to Marcus for pointing this out.)

Comment via: facebook

New Comment
18 comments, sorted by Click to highlight new comments since: Today at 5:02 AM

This feels like a nice crisp example of how Twitter is broken in ways that generate disinformation. A user with 531 followers, not someone who can reasonably be expected to treat their tweets as a journalistic product or employ a factchecker, made an understandable mistake. This produced a politically-potent but inaccurate soundbite.

A substantial fraction of the quote-tweets are refutations. These are only visible if you check for them explicitly. This creates a confirmation bias trap: if you think the economic left is pushing a false narrative, you'll be more likely to check, see that the tweet is wrong, and reinforce that belief. And if you think that inequality is a huge problem and the economic situation of the poor is dire, then you'll be less likely to check, again reinforcing that belief.

This is specifically a property of systems that don't have downvotes and of systems that don't have a good way of sorting replies, which is why this sort of thing is especially common and especially bad on Twitter..

The discourse of screenshots of tweets shared on Facebook (where I ran into this) is even worse, since they're not even attached to global replies

Downvotes don't seem to necessarily fix the problem?

These are only visible if you check for them explicitly.

How do you check for quote tweets?

Yes. And in addition purchasing power doesn't really capture the fact that £1 in 1843 can only be exchanged for the goods and services of 1843. For example

So is comparison at all sensible? Inflation calculations use a basket of goods, but when superior goods are substituted (e.g. light bulbs for candles) the comparison becomes less valid. Over 178 years, the accumulation of substitutions makes quantitative purchasing power comparisons much less meaningful.

I agree that purchasing power is a better way of measuring absolute poverty!

Interestingly, I could imagine the author of the Tweet being right about using labor value, if what they intended to point at is inequality.

Also very misleading to compare to a legal minimum, rather than to a median or comparable book-keeper's wage.  

I expect you'll also get different answers depending whether you convert GBP to USD in 1843 or in 2021. The exchange rates will be different across times and the inflation indexes will be different across currencies.

(Or will those cancel each other out? I wouldn't be shocked, and I haven't looked. But my intuitive guess is something like: "no, not reliably. Maybe if American and British societies look sufficiently similar at both endpoints. Maybe even, only if they follow sufficiently similar trajectories. I wouldn't expect either of those to hold.")

Which is more "correct"? I don't immediately see a reason to think either one is, or indeed that one or the other must be.

When we attempt to convert between 1843 GBP and modern USD, the overall process is essentially: convert from 1843 GBP to some quantity of goods and services, then figure out how much that quantity of goods and services would cost in modern USD.

Using GBP inflation tables gets us from 1843 GBP to modern GBP, and then international currency conversion does almost this. The problem is that we are missing a factor of but for 2021 GBP:USD it's 1.05x and so doesn't matter very much.

If you wanted to convert in the opposite order, you would need PPP for 1843, which sounds a lot harder to come by?

I forgot about PPP, thanks.

To clarify, you're saying it would be hard to do this in the other direction, not that doing it in the other direction would give the same results?

When we attempt to convert between 1843 GBP and modern USD, the overall process is essentially: convert from 1843 GBP to some quantity of goods and services, then figure out how much that quantity of goods and services would cost in modern USD.

I like how this makes it more concrete what we're trying to ask. And it kind of suggests we should get the same answer whichever way we do it.

But we haven't made it explicit what basket of goods and services we're looking at. I guess that's implicitly chosen by the inflation index? And those indexes are frequently swapping things in and out, so if we use something like CPI, what we're actually measuring is: "buy a bundle of goods and services in 1843 for £1. Let's say it includes candles. At some point we swap out the candles for some number of incandescant lightbulbs, and then we swap out those for LED lightbulbs. Maybe other swaps on the way. Also, probably things just sometimes get straight added to the bundle or removed from it. In 2021 we sell the bundle for USD. How much USD do we get?" And the UK and US indexes will start with different bundles of goods and services, and make different updates on the way. I guess partly based on the whims of economists having to choose between a bunch of reasonable options, and partly based on the UK and US having different situations causing different options to be reasonable?

(Also, even if we specify the goods and services, "what does it cost" doesn't have a single answer; things will cost different amounts in London versus York, or in New York, New York versus New London, Alabama. When we "buy" the basket we're taking some weighted average and the same when we "sell" it. But this might be the same no matter which direction we go?)

Hm, so if I'm right about this... if we did have an inflation index for both currencies that used the same bundle of goods and services, we'd be able to calculate PPP in 1843 as "whatever value means we get the same answer both ways"? (But it would be PPP for that specific bundle of goods and services.)

In theory, you should be able to get the same answer both ways, because they represent the same concept. In practice, I agree that you would get different answers because of the kinds of issues you're bringing up around what you choose to include in the basket and how you adjust for quality/substitution

Not sure how to gain insight by making such a comparison in the first place, twitter is... I agree that using the cost of labor is not the best way to calculate; however, there is an issue with the calculation of poverty for the period prior to 1981.

This paper finds that the $1.90/day (PPP) line is lower than the level of consumption of enslaved people in the United States in the 19th century.  

Likely, the issue here is that inflation calculations use a basket of goods that are substituted over time. Comparisons over time are not the best because a basket of goods from 2021 might include an iPhone, while a basket of goods from 1850 might include a pound of hay... 


That all being said, possibly using a harmonious pricing index might be more fruitful for this discussion. I believe that the UK and Canada use the HPI over CPI, mainly due to the issue of substitutes and compliments over time.

I don't think you are correct. Just using raw absolute poverty isn't a measure that a supporter of government social investment and economic equality would accept.

When you say "what we care about" the we might work in the context of libertarians/classical liberals who may or may not read Lesswrong . com but that we wouldn't include the original tweeter or the people amplifying the tweet.

Why don't you think "what can you buy with your wages" is what the original poster and re-sharers were thinking about? Alternatively, what sort of metric do you think they would have preferred?

A lot of it might be hedonic treadmill stuff but some of it is basically a complexity tax. And of course some people look at relative poverty to some degree.

Basically the average person has access to cheaper and safer, if not always healthier, food than a Dickens character could ever dream, even cheap apartments are superior in almost all ways.

But the requirements to function effectively in modern society are quite a pain complexity wise. You probably need at least a cheap cell phone, you probably need to own or have access to a computer with capabilities around the $500 laptop range. Transit is a bit geographically varied as a problem but is similar to other such stuff.

Basically what you can buy is way up but so is what you "need" to buy. Humans are intensely social and while you don't have to go full "keep up with the Jones family" there is strong pressure to be somewhat close to comparable people. Also for kids they are far more susceptible to comparison psychologically.

You could always dispute the validity of the view of the people sharing the tweet regarding "needs" vs "wants". But that is distinct from them being incoherent/inconsistent in their claims vs the facts internally.

I think it's pretty unlikely that this is what the person who started all of this off was getting at? They shared an update saying it was just sloppy research: And I haven't seen anyone sharing it adding commentary along these lines.

There are other examples but this was a pretty prominent reply. Plus many tweets don't give their opinion directly on the question as this one does.

The tweet you are linking seems to be pointing out that deflating by wages and deflating by prices are different things, but I don't see it giving an opinion? Perhaps this is me not really understanding Twitter?