To show how bad non-experts are at forecasting political decisions, a team of German economists asked 127 professional economists and thousands of German households to participate in an online experiment. The question at hand was whether Germany should spend more government money on its public school system.

Unfortunately, you cannot run a real-life experiment on the economic outcomes of higher education spending in a laboratory, so they had to do something different. They asked a representative sample of Germans if the government should spend more on public schools. Then they asked both the economists and the non-experts what percentage of Germans would favour higher spending on public schools. Note that this experiment was done in Germany, where support for public schools is high, but people are also very debt averse and will not be in favour of higher government spending if that means larger government deficits.

The chart below shows the key results for three different setups of the experiment. In the control group, people were simply asked if they supported higher education spending. In the information provision setup, both the general public and expert and non-expert forecasters were given the current government spending (€7,300 per pupil per year) as information about current costs. Finally, in the information acquisition setup, they were given a link where they could read up on the information and the current government spending.

The grey bars show the outcome of the survey and the percentage of Germans who favour higher education spending (77% in the case of the control setup). The blue bars show the prediction of experts about how popular such a policy would be among the public. The red bars show the predictions of non-experts based on how they said they made their predictions (using mostly their gut instinct, pre-existing knowledge, numerical reasoning or other means).

While all predictions underestimated true support for the policy, it is obvious that experts were much closer than the non-experts. In particular, those who relied on their gut feelings or ‘common sense’ were wildly off the mark, which once again goes to show that Voltaire was right when he said that the problem with common sense is that it is not so common.

Prediction accuracy of expert and non-expert forecasters

Source: Grewenig et al. (2026)

The charts below show the geopolitical fragmentation index and its components as developed by Jesus Fernández-Villaverde and his colleagues. While I am not surprised to see an increase in political fragmentation for more than a decade due to the rising distance between the US and China, I was surprised to see an increase in financial fragmentation and social mobility globally that seemingly has accelerated over the last ten years.

Global geopolitical fragmentation

Source: Fernández-Villaverde et al. (2024)

A team from Europe and South Korea has used this data to estimate how big the economic impact is across 121 countries if global fragmentation increases. As you might expect, it’s not good. A sudden 0.1 percentage point increase in global fragmentation reduces real GDP growth in affected countries by about 1.5%, driven largely by a 3% decline in business investments. The result is also an increase in debt-to-GDP by about 3% in the subsequent years because lower growth leads to lower tax revenues and thus higher deficits. What’s more, this economic shock typically lasts three years before it subsides.

If these numbers sound dramatic, they are. But if you live in a developed country, you can rest assured that you are not going to suffer such extreme outcomes. The numbers I have quoted above are for the total sample of 121 countries. In developed countries, the effect is much smaller, while it is larger in developing countries. The effect is also much larger in countries that have low political stability and higher corruption. A combination of geopolitical fragmentation and increasing corruption, together with democratic backsliding, really seems like the worst of all worlds.

But some things can dampen the negative economic effects, most notably strict fiscal rules and an independent central bank. The study shows that countries with independent monetary policy can reduce the duration of the economic downturn because they are able to stimulate the economy when it needs stimulus, but then take away the proverbial punch bowl and prevent inflation when that stimulus is no longer needed.

Similarly, strict fiscal rules as they exist in the EU and other developed nations also help mitigate the adverse effects because they reduce the ability of politicians to overspend on fiscal stimulus and keep interest expenses on government debt under control.

The economic impact of fragmentation shocks with and without fiscal rules

Source: Uddin et al. (2026). Note: FR = Fiscal Rules

It is a comparison that is, unfortunately, extremely relevant to UK government finances and bond investors, and it harks back to the dramatic episode where Liz Truss and Kwasi Kwarteng almost blew up the Gilt market in 2022.

Until that fateful mini-budget on 23 September 2022, UK Gilts were treated like the high-grade government bonds they are. Not as high a credit quality as Germany, but better quality than France, Italy, or Spain.

A team of Spanish researchers tried to understand the fallout from the UK mini-budget by trying to find similar crisis episodes in EU government bond markets. They did this by trying to identify ‘nearest neighbour’ bond market episodes that resemble the dynamics of the UK experience as closely as possible. The advantage of their methodology is that it allows for the identification of changing ‘nearest neighbours’ and nonlinear effects that often get filtered out in traditional econometric models.

The following three charts show how the behaviour of UK Gilts changed once the Bank of England had intervened to stabilise the bond market. The first episode is the immediate aftermath of the Bank of England intervention from 17 to 31 October 20222.

In that period, financial markets perceived UK Gilts the same way they did Italian and Spanish government bonds during the European debt crisis: Ready to drop into a death spiral and default at any moment. In the chart below, this shows up in the yellow line (Italy and Spain during the debt crisis), showing the smallest ‘tracking error’ to the performance of UK Gilts.

UK Gilts nearest neighbour 17-31 October 2022

Source: Andrada-Felix et al. (2026)

Once the immediate stress was digested and the government reversed course while the Bank of England continued to stabilise the Gilt market, Gilt yields dropped rapidly. UK Gilts behaved much more like core Europe around Germany, the Netherlands and other countries.

UK Gilts nearest neighbour 1 November to 31 December 2022

Source: Andrada-Felix et al. (2026)

However, once the situation calmed down in early 2023 (and presumably ever since), UK Gilts once again can be approximated by Italian and Spanish government bonds where investors feel confident holding them, but the moment there is trouble on the horizon in the form rising budget deficits or political instability, Gilt yields rise rapidly and much more than in core European markets because investors are quick to abandon UK Gilts.

What this analysis shows, then, is that years of government mismanagement (from both major parties, mind you) have cost the UK its status as a relatively safe bond market. The official credit rating of UK government bonds may still be AA- and thus higher than France (A+), Italy (BBB+), and Spain (A+), but bond investors have issued their verdict and treat it the same as Italy and Spain. And that costs all of us in the form of higher interest rates for the government, and thus also higher mortgage rates for UK homebuyers, for example.

UK Gilts’ nearest neighbour since January 2023

Source: Andrada-Felix et al. (2026)

The trick, if you want to call it that, is that in a capitalist economy, economic growth is far bigger than in a managed (i.e. socialist or communist) economy or in a feudal economy like the ones we have had practically everywhere in the world until the onset of the industrial revolution 250 years ago.

In an economy that is growing faster than inflation, living standards increase, and everybody gets better off over time.

So far, so good, but the story doesn’t end there. Because if that was all there was to say, then poverty in the US should have fallen much faster than in Europe, simply because for the last twenty to thirty years, the US had faster growth than Europe. And if you simply look at GDP per capita, it seems as if living standards in the US have indeed risen much faster than in Europe, even though that is largely an illusion, as I have explained in detail here.

The problem is that a rising tide doesn’t lift all boats equally, which is why a country that has strong economic growth but also rising inequality may actually see poverty increase. And that is the basic idea behind the work of Olivier Sterck at Oxford University.

Currently, most countries define poverty as a single level of household income, typically something like x% of the median income. And then you can debate until you are blue in the face to define the best value of x.

Instead, Sterck argues that there is a better, more practical measure of poverty: How long does it take a person to earn $1 of income, independent of the sources of income?

If you are poor, getting enough money to survive is the crucial measure. In different countries, this income is usually generated from a combination of work and welfare benefits from the government. Because the social safety net is much denser in Europe than it is in the US, poor people in the US have to rely mostly on work to get their income. In Europe, welfare benefits are a larger portion of income, thus reducing the need to work.

But be aware that this isn’t necessarily a bad thing. If you have to work less for every dollar you earn, it takes away a lot of stress and allows poor people to take care of their children and other dependents or spend time on education and training. And that has some value as well, because it makes it easier to escape poverty (by improving skills or finding a better-paid job) and it reduces social problems (by not being forced to leave children unsupervised at home, for example).

So, what’s the verdict? How long does it take to earn $1 in different countries? The chart below shows that in the US, the time to earn $1 has increased substantially from 43 minutes to 63 minutes between 1990 and 2025. In Germany and in France, meanwhile, the time to earn $1 has declined by about 25% and in the UK it has declined by a third from 51 minutes to 34 minutes. Hence, being poor has become much harder in the US over the last 35 years, while Europe has effectively managed to alleviate poverty and give people more chances in life.

Time to earn $1

Source: Sterck (2026). Note: Countries adjusted for PPP.

Of course, if you ask the people who live in host cities of World Cup matches, they may not be that happy about the influx of tourists from all over the world, but at least it is good for business. Or is it?

There have been numerous studies that show that the revenues generated from mega-events like the Olympic Games or the FIFA World Cup do not justify the costs of hosting them. But these studies have been done on a regional or country-level basis. And surely, the US economy will not feel any measurable benefit from the World Cup this year, nor will the Canadian economy or the Mexican economy. They are too large compared to the number of games played in these countries and the money spent on match days.

By the way, in recent weeks, you may have come across studies by banks on how to invest to benefit from the World Cup. I really wouldn’t do that. Adidas or AB InBev (the owner of Budweiser) are too large to feel a meaningful impact on its revenues and profits even from a mega event like this.

But that doesn’t mean that the benefits are negligible for everyone. Florian Dorn and his colleagues examined hyper-localised credit card spending data in Germany during the Euro 2024. They found that yes, around match days, there is a significant increase in consumer spending, particularly on apparel (football shirts and fan scarves) as well as in bars and restaurants.

This increase is almost exclusively driven by foreign visitors and concentrated around the stadium and in the city centres of host cities. The local population doesn’t change its consumption habits, and if anything, reduces its spending on accommodation (presumably because they want to get out of town themselves rather than invite friends and family to visit).

So, there is some benefit to hosting the World Cup, but in economic terms, it is very small and hyper-localised. Meanwhile, the joy of watching the games will accrue to all of us, especially when we see our team win or – even better – when we see the team we despise lose.

Increase in consumer spending around match days

Source: Dorn et al. (2026)

A team around Anastasija Tetereva from Erasmus University in Rotterdam used the fund manager letters available in the SEC’s EDGAR database to analyse the tone fund managers strike in their letters to investors.

Specifically, they measured the following dimensions:

• Negativity: The ratio of negative words used by the fund manager.

• Confidence: The ratio of strong modal words conveying certainty and high confidence versus weak modal words conveying uncertainty.

• Personalness: The share of words referring to the fund manager and/or her team.

• Complexity: The share of complex words used and the length of the sentences.

• Similarity: How much overlap is there between the shareholder letter and the previous shareholder letter.

The chart below shows the correlation between an increase in each of these variables and the fund’s return in the subsequent six months. I have shown all regression coefficients, even though only the confidence variable is statistically significant, to give you an idea of the directionality.

Correlation between fund manager language and future fund returns (2006-2016)

Source: Tetereva et al. (2026). ** indicates statistical significance at the 5% level.

As you can see, letters to investors that are more confident are a signal for future strong fund performance. At the same time, a more negative tone indicates lower future returns (though this is not statistically significant).

The above results are pretty weak, but they become much stronger and more indicative of future returns in times of high market uncertainty. If the data is tested for the period of the financial crisis 2007 to 2008, the tone fund managers strike in their investor letters becomes highly correlated with future fund performance.

Correlation between fund manager language and future fund returns (2007-2008)

Source: Tetereva et al. (2026). *** indicates statistical significance at the 1% level.

During these times of high uncertainty, most fund managers will tone down the certainty and confidence used in their letters, because the world is objectively harder to predict. But when you encounter a fund manager who still expresses confidence and engagement (measured via personalness), that usually is a good sign for future strong fund performance.

Meanwhile, if letters to investors become more convoluted and complex, it tends to reflect confusion in the fund manager’s mind (I have long been an advocate of using ‘plain English’; confused and complex sentences or the use of jargon are an expression of not having thought through a problem properly). And if fund managers don’t update their views or resort to boilerplate statements repeated in every letter to investors, it tends to show that they don’t know what to do.

In the research, they quantify those variables and show that using them enables investors to rank funds by future expected returns and create superior performance via improved fund selection. Most of us will not do that systematically, but the research tells us what to look for in fund manager letters.

Unfortunately, many fund managers are also reading these posts, and I just gave them a way to ‘game the system’…

While housing affordability has become extremely low in major metropolitan areas like London, Paris or New York City, the situation remains more relaxed in Germany (though you wouldn’t know it if you spoke to Germans).

Yes, house prices have gone up significantly in Germany as well. Still, an apartment in Berlin will cost you only about $8,700 per square meter in the city centre and about $13,400 in Germany’s most expensive city, Munich. Compare this to the average price per square meter in Paris ($14,300), New York City ($18,500) or London ($21,000), not to mention Hong Kong ($26,000).

A team at the University of Würzburg analysed the cost to buy and rent residential space across Germany and found that bigger cities demand a premium. The charts below show the relationship between the number of inhabitants and house prices as well as rents across a large range of German cities.

Bigger cities have more expensive housing.

Source: Gehr et al. (2026)

The main driver for this size premium seems to be that housing in larger cities is perceived to be safer in the sense that vacancy rates tend to be lower. If you lose a renter, it is much easier to find another person who will rent your flat in a large city than in a smaller city. The market is simply more liquid. And the same goes for people who want to sell their homes.

What is interesting, though, is the bottom graph in the chart above. It shows the ratio of prices to rents per square meter, a proxy for the capitalisation premium across cities and an inverse proxy for the property yield.

In larger cities, the sale price per square meter rises faster than the rent per square meter because investors and homeowners are willing to pay a larger premium to buy a house than to rent one. This implies that the rising liquidity of the local property market in larger cities creates a systematically lower property yield driven by investor psychology rather than actual supply and demand imbalances in the underlying housing market.

A commonly held belief among many investors is that sell-side analysts don’t add value and that their recommendations and insights do not create outperformance. But as I have previously discussed, this seems to be the case in the US stock market, but not elsewhere.

One of the problems that every investor has is which analyst to trust when it comes to specific stocks or markets. Plus, if this analyst disagrees with the consensus, how much weight should you give to the analyst vs the consensus?

Naman Kumar from Carnegie Mellon University developed an interesting approach to tackle these questions. He was inspired by the Shapley Value used in machine learning and other applications. In essence, the Shapley Value measures how much a specific input improves the accuracy of the output. In machine learning, the job is to remove input variables with low Shapley Value to get to a halfway tractable model.

And in a way, the job is similar for investors. They must figure out which analysts are adding value and which ones can be cut out of the consensus or the inputs to the investment process of managing a portfolio.

Of course, this is in practice mostly a qualitative assessment (though many hedge funds run alpha capture models that do this quantitatively), but his research gives us some pointers on where to look for good analysts.

First, the good news is that the average value-added of equity analysts tends to increase over time and is particularly high during times of market turmoil like the 2008/09 financial crisis, the pandemic of 2020 and the inflation spike in the aftermath of the pandemic of 2022/23.

Average analyst value-added

Source: Kumar (2026).

That’s good news for people like me and my colleagues who are trying to keep our jobs in a market that is increasingly dominated by passive investors. On average, we are not useless, it seems.

But among us analysts, what kind of characteristics should you look for? Here are the key results (and remember, they apply on average, not to any specific individual):

• Analysts who deviate more from consensus tend to be more informative and tend to improve the accuracy of markets.

• More experienced analysts tend to be more informative, which is why the juniorisation that happened in large investment banks over the last two decades probably has decreased the value of their research.

• Indeed, all-star analysts and analysts at large brokerage firms tend to add less value than analysts at smaller firms (which I wholeheartedly agree with; until, of course, I get a call from Goldman Sachs to hire me as their new European equity strategist…).

• Analysts who update their forecasts and views very quickly after an earnings release or very late after an earnings release tend to be the ones who add the most value. The former because they are responsive and can shape investor reaction to news, the latter because they typically offer deeper insights that require some digging to get to. The analysts that are in the middle of the pack in terms of timing their recommendations are the ones that typically add little value.

• Forecast accuracy, interestingly, does not correlate with value-added. There are analysts with very accurate earnings forecasts that add little value to investors and some that add a lot. Similarly, if an analyst has huge forecast errors, it might not lead to low value added for price discovery and investors’ decisions.

Overall, what you should be looking for is an experienced analyst who works for a smaller firm and isn’t afraid to make contrarian calls based on well-developed analysis. And this person ideally should be very reactive when markets are in turmoil or take time to come to a conclusion. Ideally, this person should do both, provide guidance in the middle of a chaotic market, but also deliver well-considered arguments once the market has calmed down again.

Henrick Nilsson from the Stockholm School of Economics and his collaborators looked at the investment portfolios of auditors and their investments in Swedish stocks. Of course, an auditor cannot invest in the stocks of companies that she is actively auditing. Still, outside of blackout windows or restrictions due to corporate client relationships, they are free to invest wherever they want.

The research was conducted on the investment portfolios of auditors employed at the Big 4 audit firms between 1999 and 2015, and it found first and foremost that auditors – just like most people – like to invest in individual stocks. Half the auditors in their sample owned single stocks in Swedish companies despite the many restrictions on ownership.

And when they do invest in stocks, they are pretty good at it. The chart below shows the difference in actual investment returns for Swedish auditors compared to the Swedish stock market in the 12 months after they make an investment in a stock.

Average outperformance in the 12 months after buying stocks

Source: Nilsson et al. (2025)

On average, auditors outperform the Swedish stock market by a large 3.2%, with partners (who have more experience) outperforming by 4.0%. Crucially, industry experts tend to invest more in companies in the industries they cover and are able to generate higher returns there than in industries where they are not experts.

What if the AI boom goes into reverse?

LONDON, May 29 (Reuters) - Investors should be asking not whether the AI boom will end, but what will happen to markets when it does and where safety may be found.

I have previously argued that the business models of OpenAI, Anthropic and other AI darlings look increasingly unsustainable and that hyperscalers such as Meta (META.O), opens new tab and Amazon (AMZN.O), opens new tab face ​daunting maths to make their data centre investments pay off. Yet, for now, the investment surge continues to accelerate.

Read the full piece here: https://www.reuters.com/commentary/reuters-open-interest/what-if-ai-boom-goes-into-reverse-2026-05-29/

Of course, this was always different in the US, where university is big business and colleges compete for fee-paying students. Plus, nowadays, there are a million university rankings that students can consult to apply to the most prestigious universities they can get into.

Yet, what still matters is the feel of the campus. I mean, if you get an application from a graduate of the University of Florida, they are going to have chosen that alma mater probably on different criteria than someone who graduated from the University of Illinois…

More generally, it is surprising to see that students choose their application to some degree based on the weather on the day they visit a campus. It’s no surprise that if you visit a college campus on a rainy, grey day, you are likely going to be in a worse mood than when the sun is shining and it is nice and warm. And this mood will likely influence your decision to apply in the first place.

Olivia Feldman and her colleagues tested exactly that. What was the weather on the day students visited a college campus, and how did that influence the decision to apply? If you are a university administrator, you definitely want a goldilocks scenario where the sun is shining, and it is neither too hot nor too cold. Campus visits on a rainy day reduce the chance of applying for the college by 8%, while a hot day reduces the chance of applying by 10%, and a cool day even more.

These kids today. No grit anymore.

The likelihood of applying to a college depends on the weather at the campus visit

Source: Feldman et al. (2026)

The survey was really an experiment in the psychology of business leaders with regard to both AI and robotics. The survey asked business executives about their beliefs on how many of their competitors had adopted robotics or AI. Interestingly, firms significantly underestimated how much their peers had adopted these technologies. About 93% of survey respondents underestimated peer adoption by 2.5 percentage points or more. On average, firms underestimated adoption rates by a quarter.

Then the researchers asked the business leaders if and how they planned to adopt robotics and AI in the next two years, until 2027 (the survey was done in 2025). But half the respondents were told the true adoption rates of their peers (which obviously was often higher than respondents expected), while the other half was asked without knowing the true adoption rate.

The interesting divergence here is that if business executives learned about the true adoption rate of robotics by their peers, they would become more likely to plan for investments in robotics themselves. A one percentage point underestimation of robotics adoption increased investment intentions in robotics by 0.7 percentage points to catch up.

For AI adoption, no such effect was visible. Business leaders did not change their plans to adopt AI if they knew their peers had already embraced the technology, which is surprising – at least to me.

We don’t know if this indicates that AI is a different kind of technology that creates less FOMO in business leaders than robotics, or whether AI is still so new that people simply haven’t had the time to generate enough evidence that AI is truly improving business; it is unclear so far.

In 2011, the Quarterly Journal of Economics published what I consider one of the most important economics papers in the 21^st century. Ulrike Malmendier and Stefan Nagel’s Depression Babies showed how experiences shape our attitude towards investments, the economy and politics, not just in the moment and the immediate aftermath, but for the rest of our lives. The experience of the Great Depression could be seen in attitudes some eight decades later.

A new paper in the same journal looks at the relationship between trust in government and lifetime experience of economic growth. The charts below show the experience of Americans born in the late 19^th and early 20^th century and those born in the second half of the 20^th century. The horizontal axis shows the average real GDP growth per year in the US during their lifetime (to the year shown on top of each chart) and compares it with their trust in the US federal government as measured in surveys. People who experienced higher average economic growth during their lifetime trusted the government more.

Trust in government and lifetime growth experience of US cohorts

Source: Besley et al. (2026)

What works across generations of Americans can also be seen in international comparisons. The chart below compares the average annual GDP growth from 2009 to 2019 across countries with the trust in government. Countries with low growth, like Greece, Italy or Japan, show very low levels of trust in government, while countries with high growth, like China, Bangladesh and Ethiopia, show high levels of trust.

Trust in government and growth from 2009 to 2019 across countries

Source: Besley et al. (2026)

Of course, there are many interlinkages between growth and trust in government. If corruption is rampant in a country, it will undermine trust in government and lower growth at the same time. Meanwhile, a country like Mongolia has seen high growth as it has benefited from the rise of China, but trust in the government remains incredibly low.

The dynamics are complex, but I the results confirm my belief that all politics is based on economics. If people don’t feel like things are getting better, no government can survive. Meanwhile, if people feel like things are getting better for them, they will accept a whole lot of misery in other domains of life (look at the illiberal democracies or authoritarian countries with high growth). “It’s the economy, stupid”, helped Bill Clinton into the White House, and it remains as true today as it was back then.

But here is the bad news for most of us and for aspiring politicians in particular: Growth is not going to get better in the future. Trend growth is declining across the developed world and in much of the developing world as well.

Expected trend growth rates per decade

Source: OECD

An ageing population and slowing productivity growth (despite AI and all the other technological advances) conspire to a dismal future where growth declines decade after decade. And that means political stability is unlikely to return for long unless we finally come to our collective senses and accept that the promises that governments made to us in the past (pension systems, social welfare systems) are simply not sustainable and that we have to make hard choices. But I doubt that any political party is willing to confront these hard choices. And in democracies, I remain convinced that no politician who tells the truth to voters about these hard choices will ever be elected.

Happy days are coming….

Owen Kay and Reid Taylor from the Dallas Fed, together with Robert Raeser from UC Davis, tried to estimate how much electricity prices in the US have increased due to the rapid buildout in data centres. The chart below shows that, depending on the utilisation rate of data centres, electricity production costs are now between 5% (low capacity utilisation) to 15% (high utilisation) higher than without any data centre buildout in the last five years.

Increase in power generation costs relative to no data centre buildout

Source: Kay et al. (2026)

That is not too much, if you ask me. I can live with that since any new technology will increase electricity demand and, as a result, production costs, which in turn end up raising our bills. If we weren’t prepared to pay that kind of money, we might as well shut down all factories and lay everyone off.

But of course, it doesn’t end there. The authors of the study also simulate how much electricity bills could rise if the data centre boom continues as is currently projected by the hyperscalers and others. In that case, additional power demand from data centres increases generation costs between 20% and 60% by the end of 2028.

Increase in power generation costs relative to no data centre buildout: 2026 to 2028 projections

Source: Kay et al. (2026)

Is that going to happen? If you believe in the AI and data centre hype, then yes, and you’d better prepare for much higher electricity bills. But if you are like me and think that the hype is overdone and many of these data centres that are planned today will never be built, then you can relax. Your electricity bills are going to be reasonable.

Sometimes being a sceptic about a hype has unintended benefits.

Wolfgang Breuer and Sophia Marie von Banhans (née Schmidt) wanted to know whether one can induce prosocial emotions in people with the help of music and, if so, if that had any effect on their investment decisions. Why? Because reasons.

So, they went along and recruited 204 volunteers for an experiment that the volunteers could do at home on their own computers. Half the volunteers were exposed to music (three English-language songs and one German-language song) with prosocial lyrics. The other half were exposed to four songs with neutral lyrics.

In case you are wondering, the four prosocial songs were ‘Heal the World’ by Michael Jackson, ‘We are the World’ by USA for Africa, ‘Ein bisschen Frieden’ by Nicole, and ‘Love Generation’ by Bob Sinclair.

Now, I don’t know about your taste in music, but the most prosocial thing you can do with any of these songs is put them on mute so as not to torture the people around you any longer.

Compare this to the ‘neutral songs’ which were ‘On the Line’ by Michael Jackson, ‘Englishman in New York’ by Sting, ‘Spiel um Deine Seele’ by Peter Maffay, and ‘Rock this Party’ by Bob Sinclair.

At least it’s got one good song in there from Sting, but I apologise for having exposed my readers to the other three.

In any case, the results were clear. People who listened to prosocial songs invested more money into a hypothetical green stock investment than people who listened to neutral songs. Or maybe they just paid more in order to make the experiment stop sooner.

You may laugh at my suggestion, but when measuring empathy, the research found that people who have more empathy invest more in green stocks. But people who have more empathy and then listen to the prosocial music invest less in green stocks.

Which brings me to the obvious musical coda to this post:

Ok, one detour before we get into that (I just can’t let go of that…). The price-to-sales-ratio of 93.7x is quite ambitious. For comparison, Elon Musk’s other major company, Tesla, trades at 15.3x 2025 sales, while AI darling Nvidia trades at 21.2x 2025 sales. The only major company I could find with a similar price-to-sales ratio was Palantir at 94.1x.

But back to the total addressable market, which at a headline number of $28.5 trillion is the largest one ever identified and almost as large as the GDP of the entire United States ($32.4tn). According to SpaceX, this market opportunity breaks down to traditional space tech opportunities of $370bn in rocket launches, etc., plus another $1,610bn in Starlink solutions, which they call connectivity solutions but are generally classified as space tech. That makes for a total addressable market for space tech of $1,980bn.

SpaceX’s projections of its total addressable market

Source: SpaceX S-1 filing

Now, let’s ignore the AI part and compare these estimates for the space tech opportunity with the work of Luisa Corrado and Christos Makridis. The interesting twist about their assessment is that they don’t use company projections working in the Space tech industry, but rather use US input-output tables to get a top-down estimate for the size of the space tech industry in the US. Obviously, the US is just one country among many, but it is by far the biggest player in space tech, so I think these numbers are valuable as an independent check on the projections that businesses like SpaceX make.

Using not just the direct production and revenues generated by space tech companies, but also the upstream and downstream value generated, the chart below shows the growth of the US space tech industry from 2007 to 2017. Direct industry output grew from about $60 billion in 2007 to $75bn in 2017 (all numbers are in 2017 dollars adjusted for inflation). The bigger gains have been in indirect activity, i.e. the use of space tech by the government and a variety of private industries from aviation to real estate, etc. The economic output from these indirect uses has grown from $80bn to about $120bn in 2017.

Growth of the US space industry 2007 to 2017

Source: Corrado and Makridis (2026)

But 2017 was a long time ago. How is the space economy going to grow in the next couple of years? Thankfully, the authors ran Monte Carlo simulations to 2030 to estimate the future growth of the industry. They come up with a median estimate of the direct output by space tech firms like SpaceX of c.$115bn to $120bn by 2030. That is a c.50% to 60% growth in real terms in 13 years, or about 3.5% per year in real terms. Add back 2.5% inflation, and you come up with an estimated annual growth rate of c.6% per year between 2017 and 2030.

The indirect benefit of the space economy is projected to grow to $230bn to $240bn, for an annual real growth rate of c.5%. Add back 2.5% inflation, and you come to an annual nominal growth rate of 7.5% between 2017 and 2030 for the indirect beneficiaries of the space economy.

Growth of the US space industry 2017 to 2030

Source: Corrado and Makridis (2026)

The bottom line is that this study, based on official US data, claims the total addressable market for space tech by 2030 is going to be about $350bn, compared to SpaceX, which claims a total addressable market of $1,980bn. Add to that the ‘AI and data centres in space ’ opportunity, and SpaceX claims a total addressable market of $28,500bn.

This one is in the Financial Times today, so you need a subscription to read it: https://www.ft.com/content/32bf8935-8d21-4689-ae34-8b4d3d5f6d93

PS: In a third piece due out next week at Reuters, I will ponder what may happen to stock markets if the AI boom ends, so look out for that.

Remember when former Warsaw Pact countries tried to rebuild their economies after fifty years of communist rule? Factories were in a terrible state, and companies were mismanaged. Economists back then argued for shock therapy and radical liberalisation of the local economies. That didn’t work out too well in the end. But at least it showed how difficult it is to invigorate an ailing economy dominated by unproductive businesses.

Unfortunately, Ukraine shows some parallels to these countries. Here is the annual productivity growth in Ukraine until the pandemic. There is a clear decline in productivity growth.

Ukraine’s declining productivity growth

Source: Akcigit et al. (2025)

This decline in productivity growth, according to the study, is driven, among other things, by a declining responsiveness to productivity shocks. Ukrainian businesses adapt less and less to advances in technology and hence lose competitiveness in an international comparison and experience lower growth domestically.

Declining responsiveness to productivity shocks

Source: Akcigit et al. (2025)

This, in turn, seems to be driven by a lack of creative destruction. There are fewer bankruptcies, and fewer start-ups can challenge the status quo. It seems that a lot of it has to do with state-owned enterprises, which are notoriously unproductive but are not allowed to go under.

State-owned enterprises are overrepresented among unproductive businesses

Source: Akcigit et al. (2025)

The key to growth in Ukraine and a successful Ukrainian reconstruction will thus be not just how much money is invested, but how to tame the state-owned enterprises and funnel investments into start-ups and innovative companies and allow them to grow in the face of unproductive incumbents.

Easier said than done, I know.

In 2004, South Korea introduced an Employee Permit System (EPS) for low-skilled workers to attract people in the manufacturing industry who could get entry-level jobs as machine operators, etc. This visa system was introduced because the country faced a shortage of manual labourers and was facing a shrinking workforce due to an ageing population.

Then the pandemic hit, and virtually no new EPS visas were issued because of travel bans. So, from one year to the next, South Korea experienced a large drop in low-skilled immigration. Jongkwan Lee, Giovannni Peri, and Hee-Seung Yang used this shock to the Korean immigration system as a testing ground to measure the impact on businesses if low-skilled immigration is reduced.

The headline result is that companies that relied more on these low-skilled immigrants were more likely to go bankrupt than the average Korean firm, even when controlled for the size of the firm. They only looked at Korean manufacturing companies and found that in 2020 and 2021, about 4.4% of all manufacturing companies in Korea closed shop. Among those that relied the most on low-skilled immigrant workers, the bankruptcy rate was 17.7 percentage points higher than among firms that used no immigrant workers.

This difference in bankruptcy rates could have all kinds of causes, not just the shock to labour supply, so the researchers looked at the surviving firms and how they reacted to the loss of immigrant labour. Of course, if you lose an important driver of your workforce, you face either labour shortages or higher costs to hire other workers. And indeed, a one percentage point larger drop in these EPS immigrant workers was associated with a 1.7 percentage point increase in the likelihood of a company facing revenue drops and a 1.36 percentage point increase in the probability of experiencing production disruption.

Crucially, though, Korean firms did not respond by hiring more workers. Instead, they asked existing workers to do lower-skill jobs and replaced the missing immigrants with the existing workforce. Alternatively, they retained underperforming Korean workers who, in similar circumstances before the pandemic, would have been let go. In a sense, the reduction in low-skilled immigration acted like a job retention programme for underperforming workers.

If that couldn’t be done, the companies invested in automation to replace missing immigrants with machines, but that wasn’t easily possible in most cases, so once the pandemic ended, these companies increased their efforts to hire more low-skilled immigrants. They simply couldn’t find enough Koreans to do the jobs.

Finally, one result that struck me was that, on average, Korean workers experienced a decline in wages after low-skilled immigration dropped. The mechanism here seems to be that once a Korean worker lost his or her job, they could find a new job, but one that was doing lower-skilled tasks and thus paid less than their old job.

As the authors say in their note: “Overall, the evidence suggests that immigration restrictions did not improve outcomes for native workers within surviving firms. Firms were largely unable to offset the loss of EPS workers through domestic hiring and instead adjusted by retaining incumbent workers, reallocating tasks internally, and reducing wages.”

My personal experience in that area comes from my childhood. Both my parents are from Hungary, and when I was small, we went back and forth between Germany and then-communist Hungary once or twice a year. Inevitably, we would get stopped at the border or by a police car on the road because – well – we were driving a German car with German registration plates. Inevitably, these checks could last for an hour or more. Or they could be ‘expedited’ by conveniently forgetting a couple of Deutschmarks in the documentation you handed to the officer.

Facing these realities, some countries allow their businesses to pay bribes in jurisdictions where this is considered a common or necessary practice. In Switzerland, it was for many years legal for businesses to deduct foreign bribes as an expense and lower their tax liability – though this rule has since been changed.

In the UK and the US, paying bribes is illegal, and the Foreign Corrupt Practices Act (FCPA) allows the SEC to prosecute any company listed on a US Stock exchange (even European ones with ADRs trading in the US) if they paid bribes anywhere in the world. On average, FCPA fines were $150 million per case that the SEC enforced.

Then came Donald Trump, and in February 2025, he suspended any prosecution under the FCPA and signed an executive order that stopped the enforcement of that law.

Lorenzo Crippa from the University of Strathclyde and his collaborators examined the market reaction to this non-enforcement order. He found that on the day of the executive order, the share price of US companies that had previously been investigated under the FCPA rose by 0.69% more than the average company in the US. The average gain in market cap per company was $160m, eerily close to the average fine imposed by the SEC on these companies in the past.

However, this may not be the result of the market cheering the ability to pay bribes. Instead, it may just be a reflection of the market assessing the impact of lower compliance costs. But if that were true, companies that have never been targeted by the FCPA should have seen their share price rise as well. Yet, on average, the companies that have never been targeted by the SEC under FCPA saw no increase in their share price and a small negative abnormal return in the days following the executive order. And this is hard to reconcile with the assumption that lower compliance costs pushed share prices higher for affected companies.

Abnormal share price reaction to suspension of FCPA

Source: Crippa et al. (2025)