Malta: A Free-Market Success Story

By Dan Mitchell. 

"I’m in Malta for a bit of research before speeches in Amsterdam and Reykjavik as part of the Free Market Road Show. 

So today is a good opportunity for a column on Malta’s rather-successful economy (something I’ve done for other countries, such as Poland, Chile, Botswana, Singapore, and Estonia).

Let’s start by looking at Malta’s score from the latest edition of Economic Freedom of the World.

Malta is ranked #18, putting it easily in the top quartile.

It gets very good scores in every category other than fiscal policy (somewhat similar to Nordic nations).

What are some of the best features of Maltese economic policy? Let’s look at some excerpts from a column in The Business Picture by Nima Sanandaji.

"…while the big economies of Europe are stagnating, several of the smaller ones are outpacing the US. Malta is the best example, since it led the European growth league in 2024 with a five per cent growth. …Malta is succeeding thanks to competitive taxes and regulations, combined with talent supply, which make it a growing brain business jobs hub. …The share of adults employed in these jobs has increased substantially over time. Currently 9.5 per cent of adults in Malta are employed in highly knowledge intensive jobs. After Switzerland, Ireland and the Netherlands, this is the highest rate in Europe. …Large economies like Greece, Spain, Italy and France have due to regulatory and tax burdens stagnation in share of adults in knowledge intensive jobs. The same countries also struggle with economic growth and job creation."

I’m not surprised that Malta is growing faster than the United States. It’s a classic example of convergence.

What’s more interesting is to look at examples of divergence.

Here’s a chart, based on the Maddison database, showing Malta’s long-run performance (in red) compared to regional competitors, as well as a sampling of other nations.

 

"A few years after World War II ended, Malta was very poor. It ranked lower than Madagascar and its level of per-capita GDP was less than half of Greece.

Now it is has shot way past those two nations, as well as other countries that used to be richer.

Amazingly, Malta has almost caught up with Italy, which had nearly four times as much per-capita GDP back in 1950.

Does this mean Malta has perfect economic policy? Of course not. But it does have better economic policy than most other nations, especially its Mediterranean neighbors.

The moral of the story is that there’s a recipe for growth and Malta is doing a decent job of following the recipe. Assuming they want prosperity, other nations should do the same thing."

AI and the future of labor demand

See You are not a horse by Brian Albrecht. Excerpts:

"For simplicity, suppose human labor demand goes to zero. Not low. Zero. What does that require? It means no dollar you spend, anywhere in the economy, passes through a human hand at any point in its supply chain. Not the person who made the thing. Not the person who shipped it. Not the person who designed it, sold it, maintained it, or cleaned the building where it was assembled. Zero human labor embodied in final expenditure. That’s the target. That’s what I’m going to take “humans become horses” to mean, stated precisely. [tractors replaced horses and horses did not find employment elsewhere-the number of horses in the USA fell greatly after tractors came in]

This is the input-output idea Leontief built his career on. You can trace any final purchase back through its supply chain and add up all the human labor that went into it, direct and indirect. A cup of coffee has the barista, but also the roaster, the trucker, the farmer, the person who made the truck. “Embodied labor” means all of it. For labor demand to collapse, every one of those links has to go to zero, in every product anyone buys

The economy is not one production function. It is many activities. When AI makes some of them cheaper, people don’t just buy more of the same thing. They buy something else.

Every dollar you spend lands somewhere. Some dollars land in activities with lots of human labor inside them: a restaurant, a therapist, a roofer. Some land in activities with almost none: a streaming subscription, an automated checkout, cloud storage. So when we are tracing out what happens when AI gets cheaper, it’s not just “Can AI do my job?” It is “When everyone saves money because AI did my job cheaper, what do they buy next?”

Aggregate labor demand depends on three things: how much people spend in total, how much of that spending lands on activities with human labor inside them, and how much labor is embodied in each of those activities. For human labor demand to collapse, it’s not enough for AI to displace workers inside some activities. Every dollar of spending, wherever it lands, must lose all its embodied human labor. That’s three channels, and the horse argument needs all three to go wrong simultaneously.

The important starting point for thinking about labor is te idea that nobody wants labor. A restaurant doesn’t want waiters; it wants orders taken, customers reassured, mistakes corrected. So labor demand is derived demand. How does AI change how much firms demand?

When AI can do the things firms are actually buying, cheaper AI does two things at once. Firms substitute AI for workers, which reduces labor demand per unit of output. But cheaper AI also lowers output prices, output expands, and the expansion pulls labor demand back up. Whether labor demand rises or falls depends on which effect is larger. This is the Hicks-Marshall decomposition of derived demand into substitution and scale effects.

This is going to be the organizing principle for everything. When a dollar is saved, where is it redirected? To new tasks? To new jobs? To new sectors? It must go somewhere

This is obviously true for many things. Early models had this even. For example, the early GPT exposure paper by Eloundou, Manning, Mishkin, and Rock estimated that roughly 80% of the U.S. workforce could have at least 10% of tasks affected by LLMs. With complementary software, 86% of occupations cross the 10% exposure threshold

And lots has been done on this. The task-level evidence backs this up. In a large customer-support setting, access to generative AI raised issues resolved per hour by about 15%. In a professional writing experiment, ChatGPT reduced average task time by 40% and raised measured output quality by 18%. In a controlled GitHub Copilot experiment, developers completed a coding task 55.8% faster. These aren’t tiny effects.

But they’re effects on tasks. The saved dollar doesn’t vanish when a task gets automated. It creates new tasks within the same job, such as more review, more client management, more judgment calls. Just as there’s not some fixed amount of demand so the scale effects matter, there is not some fixed job.

There’s a ritual in AI discourse where someone posts a demo, the demo does a task associated with a job, and people conclude the job is doomed. Sometimes they’re right. But the inference skips about fifteen steps. What does it actually cost to deploy, errors included? Do customers trust it? Does management know how to reorganize around it? A chatbot demo can appear overnight. A hospital reorganizing clinical liability around AI cannot.

We need to think not just about jobs but organizations. Often the result is a team, not a replacement. A human-AI pair produces output. But complementarity is not free. A pair that produces only slightly more than the AI alone doesn’t justify the human wage. The human has to add something the AI can’t replicate cheaply.

Surgery, aviation, structural engineering, fiduciary advice, for legal reasons alone are areas where we can expect the damage from an error dwarfs the savings from cheaper production. Again, that can always change one day but not soon. When failure on one component destroys the value of all others, you don’t care about the sticker price. That’s the O-Ring logic. You care about cost per unit that actually works. When damage stakes are high enough, human-supervised production wins regardless of how cheap AI becomes.

Suppose substitution wins inside most jobs. The saved dollar escapes the workplace entirely. Where does it go? Most standard models aggregate into a single final good, so this question plays no role. The real economy has many sectors, and the dollar has to land somewhere.

Start with software as a microcosm. This is a sector that has already been heavily automated by digital inputs for decades. If substitution were going to drive labor out of a sector, this is where you’d see it first."

"The most software-intensive industries don’t just retain human labor;they have a higher labor share (67%) than the least software-intensive ones (55%). Heavy digital inputs didn’t drive out human labor. If anything, the industries that automated the most are the ones that spend the most on workers. BLS projects U.S. employment to increase by 5.2 million from 2024 to 2034. Software-developer employment? Up 17.9%, despite direct AI exposure.

The scale effect won within the sector most exposed to digital automation. The BLS could be completely off but the evidence so far points strongly toward the scale effect dominating in software-intense industries.

Software is one extreme but we basically have the same pattern holds across the whole economy, over a much longer period.

For another angle on the problem, let’s go bigger and look across the biggest sectors in the economy: services vs. goods. In 1929, most consumer spending went to physical goods. Today, roughly two-thirds goes to services. As manufacturing got cheaper, people didn’t just buy more stuff. They shifted spending toward healthcare, education, restaurants, personal services. That’s the saved dollar in action at a more not-quite macro but close level — the savings from cheaper goods flowed toward services.

In terms of our guiding decomposition, coods got cheaper."

"Demand for physical stuff didn’t explode. Instead those freed-up dollars migrated to services, and the scale effect showed up there. The substitution effect won inside goods-producing industries. The scale effect won across sectors. Output overall expanded. So if you’re thinking as a macroeconomist, the scale effect dominated."

"But migration alone doesn’t help workers unless the destination still has human labor inside it."

"Services consistently pay a higher share to labor than goods-producing industries. Spending didn’t just migrate. It migrated toward sectors where more of each dollar ends up in someone’s paycheck."

"there is a margin of adjustment, there is an escape hatch when you are looking at an economy as diverse as the modern U.S. economy."

"comparative advantage always pops up fighting against this. When automation makes some things cheap, the things that remain expensive tend to be the things that are hard to automate. And the things that are hard to automate are, almost by definition, the things where humans still have comparative advantage. The saved dollar drifts toward where humans are still worth paying. That’s not optimism. That’s what comparative advantage means.

"In early textiles, power looms cut labor per yard of cloth. But cloth got so cheap that demand exploded, and total employment in textiles rose for decades. Same in early steel, early autos. Eventually demand saturated, prices stopped falling fast enough, and automation reduced employment in each sector. The question for AI isn’t “does automation destroy jobs?” It’s “which phase are we in, for which sectors?”

Where might the AI-saved dollar land today? Healthcare is already 18% of GDP and rising. Elder care will grow as populations age."

"this time isn’t different: new tasks appeared, comparative advantage held, products we couldn't imagine created new work." 

"If AI keeps inventing new varieties of goods that compete with human-produced ones, even a strong initial preference for human labor gets diluted by expanding choice.

I take this seriously. It’s a possible scenario.But notice what it requires. Not just that AI-produced variety expands (which it will) but that it expands fast enough and broadly enough to pull spending away from every human-intensive category at once. The question isn’t whether AI competes with some human goods. It’s whether any human-intensive island survives. Does anyone still spend money on something with a person inside it?

The numbers still have to be extreme. Suppose AI eats 85% of the economy. Software, accounting, law, medicine, logistics, most management, most media. All gone or nearly gone as human labor categories. Suppose the remaining 15% of spending goes to things with at least 30% human labor inside them. Elder care, in-person education, surgery, live performance, skilled trades, therapy, status goods. Then the aggregate human labor share is at least

S ≥ 0.15 × 0.30 = 0.045

That may not sound great but I’m literatlly just putting a bound. Knowing nothing else, we can sustain this. Not large. Not utopia. But not zero, and that’s the absolute lowest possible bound. And remember, labor share declining is not the same thing if the pie is growing much larger."

"As AI makes commodities cheap, real incomes rise, and richer people systematically shift spending toward what he [Alex Imas] calls “relational” goods 

 There's a huge literature in economics on structural change, the long-run pattern where spending shifts from agriculture to manufacturing to services as countries get richer. The big question is why. Is it because prices change and people buy more of whatever got cheaper? Or is it because incomes rise and people just want different stuff? Comin, Lashkari, and Mestieri, for example, decompose the two and find that income effects account for over 75% of the shift. That matters here. If spending migration were mostly about chasing cheap goods, AI making things cheaper would pull dollars toward AI-produced stuff. But it's mostly about what richer people want. And richer people have consistently wanted more services with humans in them."

"Human-created artwork gains 44% in value from exclusivity, versus 21% for AI-generated artwork. AI-made goods feel copyable. Human-made goods feel scarce even when they aren’t. People want what other people can’t have. That wanting doesn’t run out, and it sticks to things a person made."

"income effects dominate price effects by three to one. When basic needs get cheaper, humans don’t say “good, I’m done wanting.” They invent new ways to compare themselves with neighbors. Whether the new wants land on human-made goods or AI-made goods is the open question, and the experimental evidence so far favors humans.

A falling labor share is not falling labor demand. There is a range where labor’s share of income is declining but total labor demand is still rising, because the pie is growing faster than labor’s slice is shrinking. That range may be where we are right now. It would look like “AI is taking over” in share terms while employment keeps growing. The popular argument runs these together and they are not the same claim.

We already see that. Higher income people consume more services. Services tend to be high labor share. Again, that can always flip in the future but this is the evidence we have."

Affordable manufactured housing versus unaffordable climate regulations

By Ben Lieberman of CEI.

"The Biden administration had a field day piling on one costly climate-related regulation after another, not knowing – or caring – that affordability would emerge as a much more pressing concern for Americans than climate change ever was. But now, the Trump administration and Congress have the opportunity to undo these ill-advised rules that are driving up costs for everything from utility bills to cars and light bulbs. We have already seen some progress, but there is much more to do. Next on the list should be Department of Energy (DOE) regulations targeting manufactured housing.

The housing affordability challenges are real, and government is a big part of the problem.  According to the National Association of Home Builders, regulations at all levels of government account for almost 25 percent of the cost of a new single-family home. This includes a growing contribution from federal climate measures, such as those raising the price of major home appliances like air conditioners and furnaces. Worst of all are rules that make the most affordable homes less affordable, thus threatening the dream of homeownership for low-income and younger households. That is why the 2022 DOE energy efficiency rule for manufactured housing warrants a second look. 

The DOE sets energy efficiency standards for manufactured housing. And, as with appliance standards, the agency has a knack for rules that raise up-front costs beyond what is likely to be recouped through energy savings. In this case, the agency admitted that the 2022 rule raised home prices up to $4,500, though manufacturers fear higher costs will outweigh any energy savings.

For perspective, estimates suggest that every $1,000 increase in a median-priced home disqualifies about 156,000 prospective homebuyers. And the effects may be more severe at the lower end of the home spectrum, including manufactured homes, which are the choice of the most price-sensitive buyers. Indeed, it is quite possible that the DOE rule alone is enough to place the dream of homeownership out of reach for hundreds of thousands of lower-income Americans.

As was often the case for the Biden DOE, climate change was a finger on the scale favoring its draconian energy limits on manufactured housing. In fact, the final rule mentions the social cost of carbon dioxide and other greenhouse gases a whopping 50 times. By the agency’s own estimates, the rule’s climate benefits fell short of the claimed consumer savings. Even so, they undoubtedly played a role in the agency’s decision to adopt such stringent standards, despite their effect on prices.   

Fortunately, the president and Congress have not ignored the regulatory plight facing manufactured homes and their prospective purchasers. President Trump’s March executive order titled Removing Regulatory Barriers to Affordable Home Construction, specifically mentions manufactured housing in its section urging regulatory reforms.

Both the House and Senate have passed bills addressing housing affordability, and both contain provisions specific to manufactured housing. Importantly, both bills eliminate the costly and unnecessary requirement that manufactured homes have a steel chassis, however they also fell short of repealing the DOE rule.

A separate House-passed bill, H.R. 5184, the Affordable HOMES Act, would have completely repealed the DOE rule, but it has not been taken up by the Senate. Total repeal deserves consideration if Congress is serious about addressing housing affordability."

How Measurement Choices Shape the Housing Debate—and the Charts in the President’s Economic Report

By Norbert J. Michel and Jerome Famularo of Cato. Excerpts:

"The Council of Economic Advisers’ 2026 Economic Report of the President tells a familiar story: The American dream of homeownership is slipping away. Chapter 6, in particular, leans heavily on a series of charts meant to show that housing has become less affordable, less attainable, and more distorted by regulation.

While it is true that regulation adds unnecessary costs and distortions to the housing market, much of this “unaffordability” narrative depends on how the data are presented. Change the framing, even slightly, and the story starts to look very different.

Take the report’s central claim that housing has become dramatically less affordable because home prices have outpaced income. That conclusion rests on a simple comparison of real house prices to real median income. Among other problems, this comparison ignores the fact that homes being built in recent years are not the same as those built in years past: They have more standard features and, most notably, are larger in size on average."

"A similar framing issue shows up in the report’s treatment of homeownership rates. By comparing 2000 to 2023, the report suggests a worrying decline, especially for younger Americans. But 2000 was not just any year—it had higher than average homeownership rates relative to other years.

Zoom out, and the trend looks much less alarming. For instance, the homeownership rate for Americans under age 35 is roughly in line with where it was throughout much of the 1990s. The overall rate shows a similar pattern. You can make the numbers look bad by picking convenient endpoints, but that doesn’t tell you much about the underlying trend (Figure 2)." 

 

"More broadly, federal policymakers are overly concerned with so-called housing shortages. 

In reality, as communities grow and people earn higher incomes, higher demand for housing can put upward pressure on prices. (Even if that demand comes mainly from “rich” people, it can put upward pressure on average housing prices.) Viewing this kind of price increase as a shortage or market failure is counterproductive. Over time, this demand tends to be met, keeping up with the needs of a growing population.

Either way, the supply of housing is not the sole determinant of house prices. Ignoring this lesson and implementing policies that merely focus on boosting supply (especially through federal subsidies, grants, tax credits, etc.) can lead to depressed home values and oversupply, just as implementing demand-boosting policies can distort markets. The best thing for the federal government to do is to stop interfering with the market.

Of course, certain policies make it more difficult and expensive for builders to meet new demand, and state and local officials should implement the best policies for their local growth conditions. To be clear, supply constraints and regulations matter. Zoning rules, permitting delays, and other restrictions make housing more costly. However, if rising housing costs partly reflect larger homes and higher incomes, then policies aimed at forcing down prices (such as mass deportations and bans on institutional investors) could have unintended consequences. 

Sure enough, recent research from the San Francisco Fed suggests that faster income growth, not supply constraints, explains much of the differences in house price trajectories across metro areas. This finding makes sense because, for the past few decades, Americans have been earning higher incomes. All else equal, this fact should help explain higher housing prices.

Ultimately, policymakers should be wary of solutions built on a misleading diagnosis. In housing, as in economics more broadly, how you measure the problem often determines how poorly you solve it."

ICE has not improved U.S. labor markets

From Tyler Cowen.

"We provide the first causal, national empirical analysis of the labor market impacts of heightened immigration enforcement during the second Trump administration. Enforcement increased everywhere, but, we take advantage of the fact that the increases have been uneven across geographic areas to classify areas as treated or control and then implement an event study and difference-in-differences design. Areas that experienced particularly large increases in the number of arrests also experienced a decrease in work among likely undocumented immigrants who remain in the U.S., compared to areas with smaller increases in arrests. We find no evidence of positive spillover effects to U.S.-born workers and U.S.-born workers who work in immigrant-heavy sectors are harmed.

That is from a new NBER working paper by Elizabeth Cox & Chloe N. East."

The "Great Awokening" Had a Strong Upper-Class Accent

From Marco M. Aviña.

Abstract 

"Recent scholarship hails rising racial liberalism among white liberals as a racial reckoning and even a “Great Awokening.” I find that affluent white liberals led these changes. I develop a status-signaling account in which members of this group, embedded in dense, politically homogeneous social environments, face competing reputational and gatekeeping incentives to express alignment with racial equality in principle but not in policy. I leverage the murder of George Floyd and the ensuing Black Lives Matter protests as a focusing event when anti-racist norms surged. Using regression-discontinuity-in-time and event-study designs on national public opinion surveys and local public meeting transcripts, I show that post-Floyd movement was concentrated among high-income white liberals, centered on symbolic engagement rather than material policy commitments—whether universalistic or group-targeted—and short-lived. Finally, I provide evidence consistent with this account: post-Floyd, an income gap emerges in implicit bias testing, a form of self-monitoring, but not in implicit bias scores, which are less amenable to impression management. These findings complicate narratives of racial progress and recast the “Great Awokening” as recognitional politics without commensurate redistribution, consistent with concerns about elite capture in identity politics."

Capitalism does a far better job than socialism of feeding people.

A tweet from Chris Freiman.