#bryan caplan

He implicitly describes a calculation in his point 12:

“Why on Earth don’t we heavily restrict welfare benefits for migrants, so E(X)>0 for far more people?!”

which suggests that his expectation E(X) is calculated in terms of something like "net taxpayers". I don't know what calculation he actually has in mind, because - and this is a major issue - he doesn't specify the calculation that his argument leans on.

I will not criticise the specific "net taxpayers" possibility too harshly, for fear of strawmanning. I will instead mostly criticise him for the under-specification, and I think the general class of Expected Value Calculations consonant with "heavily restrict welfare benefits" are easily gamed while failing to capture the values and preferences involved in opposing mass immigration. The details Caplan has given are exploitable; the details Caplan has not given are a canvas on which each listener is invited to project his preference.

One can imagine a theoretical EV calculation for which it is tautologically true that one would want to import all E(X)>0 foreigners, but the hypothesis of a Platonic object tells one nothing about what that calculation is or how many foreigners are E(X)>0 to import.

I don't want to attack something Caplan didn't say, but he's skipped some important argument steps that I think would merit a great deal of attack if he did say them. For example the steps between the welfare-benefits-based EV calculation, and the "everyone with E(X)>0" assertion, because I can easily imagine people who are net negatives to have around even if they are not net welfare consumers.

Now you’re saying, roughly, that we should only welcome people that definitely have highly positive value. In your words:

If one is operating anywhere near a welfare-benefits-based kind of EV calculation, then it certainly behooves one to take only people who definitely have highly positive value on that calculation, because they may have negative other factors that bring the "true" EV calculation down and so one needs a buffer on the welfare-benefits side to ensure that "true" E(X) > 0.

Which brings me to:

2. The Black Swan of Rotherham.

In 2005, if you had suggested that mass migration would reintroduce slavery to Britain, you would probably be laughed out of the room for absurd paranoid fearmongering. Around 2010-2015 the Rotherham rape ring scandals came to light, first in Rotherham itself and then in Telford and Rochdale and elsewhere, and it gradually turned out that mass migration had reintroduced slavery to Britain.

I speak here of "slavery" in an institutional sense. A single man who is coerced to work and cheated out of his rightful pay may be called a slave, and statistically speaking that probably happened numerous times after Britain abolished slavery in 1833, but an incident does not make an institution. Whereas circa 1980, Britain imported Pakistani rape-gangs numbering in the hundreds of slavers, who took thousands of British girls as sex slaves across dozens of cities over the next few decades.

That was institutional by scale; it was also institutional by policy because when the fathers of enslaved children attempted to recover their children, the police repeatedly ran interference for the slavers, sometimes arresting the slaves, other times arresting the fathers for disturbing the peace, other times saying nothing could be done because of 'racial tensions'. Politicians helped hush up the slaver rings because they were concerned that people noticing the Pakistani slaver rings raping thousands of British girls might lead to the native Brits being racist against Pakistanis.

Problem 2a is the object level: importing slavers to enslave the existing population has very large negative EV for an intuitive EV calculation, but it hardly shows up in the ratio of taxes paid to welfare benefits received by the slavers.

Problem 2b is the future unknowns: considering how unpredicted and unnoticed this was until after the fact, how many other problems of the general type of "Ooops, we reintroduced slavery" may be lurking? India has for example Hindutva vigilantes who murder people for violating the sacredness of cows (at least 9 dead in 2024); imagine importing those from a place where beef is banned to a place whose national dish is the hamburger.

Problem 2c is the lies and coverups: many powerful people thought it was more important to preserve the reputation of immigrants in general and Pakistanis in specific than to stop slave-rape-gangs. This creates a credibility problem when looking for sources to calculate the impact of migration. Caplan seems more honest than average, but still shows some sympathy for coverups in his point 14:

My friend and colleague Tyler Cowen recently advised you to stop publicly defending high-skilled immigration, and “just work behind the scenes.” Maybe he’s right, but I think he underestimates your powers of persuasion.

Problem 2d was the weak response: Britain jailed only a small fraction of the slavers, jail sentences were short, the policemen got a slap on the wrist with one police commissioner stepping down, there should have been a targeted re-education program to break the Pakistani-British culture and enforce assimilation or deportation, there was not, instead there was a stricter control of people saying racist things on the internet, while importing more Pakistanis. This has generated substantial ethnic resentment among the native British population.

How do these figure into the E(X) of mass migration, or the determination of whether it's above 0? God only knows. The error bar on the value of Pakistani immigration looks larger than the value itself to me.

Perhaps Caplan intends to filter out such people from mass migration as part of EV determination. If so, he's handwaving over both a calculation problem and a practical implementation problem.

3. Rights, Privileges, Serfs, and Riots

Some time ago, a fellow on Tumblr bemoaned how difficult it was to move to another country.

I responded that it was trivial to "move to" as in transport myself to another country, which I had recently done that year for my summer vacation, but it was difficult to "move to" as in acquire political power and entitlements in another country for arguably good reason, and challenged the fellow to clarify which he meant. I never got a response.

Bryan Caplan trades on similar ambiguity when he posts cartoon panels such as this:

As written, I deny the claimed "right". I do not think he believes it himself; he would deny my "right" to live and work in his house.

But even interpreted charitably, Caplan is playing games, he is pulling a bait and switch maneuver, he is doing a motte-and-bailey between move as in transport and move as in acquire entitlements, and he is skipping important steps. Once again it's hard to give specific criticisms because I don't know which end of the ambiguity he really intends, so I will attempt to suggest some problems at either end:

If Caplan asserts a right for people to live and work and acquire political power and entitlements in foreign countries, he's arguing for a self-contradictory 'right to privileges', and he's arguing for the 'right' to destroy every small country in the world, in particular the ROC (population 23 million) which is susceptible to the PRC (population 1.4 billion) finding the 0.2% most patriotic loyalists (28 million), sending them into ROC and holding a majority vote to integrate the ROC into the PRC. This strikes me as an obviously wrong conclusion, reductio ad absurdum.

On the other hand, if Caplan asserts a right for people to live and work where they like but only as long as they're a powerless underclass banned from the ER, existing to pay taxes to the native population, and getting deported if they become welfare cases, then a moral problem is that he's advocating something like the return of serfdom.

Spare me the medieval nitpicking, I know the word is not exactly accurate, serfs had more rights than that. Helots is more accurate, but I think most English-speakers have an intuition for "serfs" that they don't have for "helots".

A practical problem following from that is that a large helot class in America would be very hard to keep as helots, when the country has a tradition of democracy, a history of expanding the franchise over time, and riots. The helots would be political tinder waiting to burn.

I ask Caplan: Suppose you get your helot class, American GDP goes up, and then a photogenic helot dies in a way that might have been prevented by welfare, leaving behind a pair of sad orphans. The Democrats spring into action to demand helot welfare and enfranchisement, organizing a helot riot. What do you do?

Of note here is that a helot riot doesn't have to win to wipe out the tax gains from helots. BLM's fiery rioting in 2020 caused at least a billion dollars in damages (as measured by insurance payouts) without abolishing the police.

4. Wage Suppression and Automation

In his proposal to optimize net taxpayers or something like it, Caplan would optimize at the expense of a great many Americans, particularly low-skill Americans who would suffer from extreme wage suppression. America is a nice place to live partly because of the twofold effects of labor scarcity: labor had more bargaining power against capital, and was able to demand better working conditions, and labor scarcity incentivized automation, which freed people up to do other jobs.

Regarding bargaining power: Caplan analogizes America to a corporation, I would respond by analogizing America to a union, and Caplan's proposal to colossal amounts of scab labor intended to break the union to save the CEO some money. Why should the union put up with this?

Regarding automation: If one goes back a few millennia, almost all of humanity worked one of two jobs: producing food or producing clothes. By inventions such as the plow, the loom, the horse collar, the spinning wheel, and the tractor, machines* took almost all these jobs and humans were freed up to do other work like steelmaking and glassblowing. Then machines mostly took those jobs too, the process of automation repeated, and now the average American benefits from machine-power equaling the manual labor of hundreds of humans.

*horses are machines in this context.

Caplan proposes to import large amounts of unskilled migrant labor to do scut work, and doesn't say how this interacts with automation. Concretely:

But look at your own companies. You don’t just hire top engineers and programmers. You hire receptionists, assembly-line workers, janitors, gardeners, and construction workers. With good reason: Otherwise, your top engineers and programmers would have to waste their precious time answering Tesla’s phones, assembling its cars, cleaning its toilets, mowing its lawns, and pouring its concrete.

Many of these look automatable, particularly the assembly line, which is already well into the process.

Once again I'm frowning at an ambiguous gap in Caplan's proposal, where I can imagine several possible views but criticising any particular one is something of a strawman because Caplan hasn't committed to it.

Does he imagine that automation will decline naturally as a result of the cheap labor? Because that sounds like trading long-term benefit for short-term net taxpayer count.

Does he imagine that automation should be held back? Same but worse.

Does he imagine that automation can't take these jobs any time soon? Sounds fake.

Does he imagine that automation will take these jobs soon but then America can just send all the migrants back once they're out of work, and wash its hands of them? Sounds unlikely and impractical.

Does he imagine that automation will take these jobs soon but low-skill migrants will simply retrain and develop skills for new jobs? Sounds wrong by construction.

And once more I ask: E(X) for whom? Caplan says America "needs" mass unskilled immigration, but large numbers of Americans would suffer from this.

5. Social Contracts

I am generally skeptical of social contract theorists as trying to claim too many specifics from too little evidence. Hobbes in particular was lying about the war of all against all. Even so I am sympathetic to a minimal account which goes something like this:

I (we in general) will give up my natural right to take amateur vengeance on and extract compensation from anyone who has wronged me. In exchange, the State promises to perform vengeance by a professional enforcer class in a way that's predictable and reliable and won't lead to blood feuds, and to pay me compensation from the collective compensation fund.

The modern American state has enforced a broadly similar new social contract which I might describe thus:

I (we in general) will give up my natural right of freedom of association, to decide which countrymen I will admit or exclude to my business, will hire or not. In exchange, the State promises to perform exclusion at the country border by a professional enforcer class, lowering friction internally and lowering costs of maintaining an exclusion around the collective American identity.

I have complaints and nitpicks about this, but I can see a meaningful value proposition in it.

When Caplan argues for open borders, he is arguing to take away what Americans received in that second contract. Again there's a gap where I don't know what Caplan believes, so I will comment on the two likely interpretations I can think of:

Is Caplan trying to tear up the new social contract in its entirety? Then I want him to bite the bullet and say out loud that he wants to restore freedom of association and overturn the Civil Rights Act.

Is Caplan trying to take away the benefit of the new social contract and give nothing in return? Then many Americans might reasonably want him jailed for conspiracy against rights or similar offense.

6. Policy Change Friction Around Humans

Bryan Caplan argues against "safetyism" and makes the analogy that just as an investor should want to make every investment with E(X)>0, a country should want to admit every migrant with E(X)>0. He admits "While there are obviously major differences between running a corporation and running a government", then ignores this obvious point as though it made no difference.

I want to highlight a particular point of dis-analogy: it is much easier, practically and morally and legally, for an investor to ditch a million-dollar investment at the first sign of it turning bad than it is for a country to ditch a million migrants at the first sign of them turning bad.

Migration is difficult and costly to reverse, and most countries have significant political constituencies opposed to that reversal, backed by international organizations such as Amnesty. You'd be hard pressed to find a single elected official with a strong opinion that Jane Doe must stay invested in Acme Corp, for most values of Jane and Acme.

Under these circumstances, some form of safetyism is correct: the threshold should not be E(X) > 0, but E(X) > Cost(Deportation), estimated around eighty thousand dollars per person by the American Immigration Council. Which is probably biased, but the sources I can find for estimating this number seem to amount to either AIC knockoffs or else Trump fanatics insisting "deportation will pay for itself".

7. Assume a spherical cow in a vacuum...

The "spherical cow" is originally a physics joke about greatly simplified modeling that discards many features of the object under consideration to simplify calculations. In physics, this is often good enough because the features under consideration, e.g. "mass of an object", range between the 10^-21 grams of an atom and the 10^33 grams of the Sun, so one can afford to round off (ha) a great deal and still be close enough on an exponential scale. The Earth is approximately spherical even if a mountain rises a few miles above sea level, that's very little compared to the circumference of thousands of miles.

Outside of physics, the spherical cow approach is less applicable.

I would like to see Bryan Caplan distinguish more sharply between the realistic policy changes he's pushing for on the margin, and the spherical cow policies where he imagines a friendly Supreme Dictator who can copy UAE policies to the US. I would also like to see more awareness from Caplan of when he is assuming a spherical immigrant who can be rolled across a frictionless border in a political vacuum. Caplan's talk of E(X)>0 is spherical talk, assuming a simple calculation. Caplan's neglect of ethnic resentment in the implied EV calculation is discarding important features. Caplan's implication of ditching migrants if their recalculated E(X)<0 is handwaving over a great many issues and costs.