Cappuccinomics: What the Price of a Humble Coffee Reveals about Prices in the Global Economy

Measuring prices is fundamental to understanding any economy. This is exactly what Deutsche Bank does with Mapping the World's Prices. The 2026 edition, published in July, contains troves of data about the cost of various goods and services in different cities around the world [2]. Zürich and Geneva, both in Switzerland, are the world's priciest cities to live in. Hong Kong has the most expensive property, although prices in Budapest have tripled since 2012. London has the world's most expensive public transport, as measured by the cost of a monthly pass. One can only envy the Luxemburgers, who have had their public transport for free since 2020. More jovial price indices are also considered: the so-called "Cheap Date Index" measures the cost of a cab ride, dinner and two cinema tickets - and here too London ranks near the top of the list, coming in 6th place globally.

These are all fun statistics, and they can be consulted in much more detail in the report itself for those interested in an illuminating read. In the following few pages, however, I address a particular issue common to all measurements of "prices".

In general, converting nominal prices into figures that carry real economic meaning is a genuinely complex affair. Different countries have different currencies and different costs of living, so a wage of $3,000 a month means something quite different in Zurich than it does in Delhi [2]. Indexing wages to a good with a roughly constant global price is one way to overcome this issue. iPhones are generally sold at the same price everywhere across the world, with small differentials created by local taxes and VAT. If this were not the case, arbitrage would equalise prices (consumers could simply buy from a different country). Therefore, the question of "how many iPhones does your monthly salary buy" is an interesting (and crucially, currency-free) measure of effective real wages.

If you've ever queued for a coffee anywhere on campus, you'll know it doesn't come cheap in the UK. But why does a cappuccino at Warwick cost roughlywhat it does in London, or in Zurich, while being nearly three times the price of one in Mumbai? Getting to the answer necessitates a brief detour into economic theory.

In general, real-terms wages in the tradable sector are highly dependent on productivity. If firms in such industries paid too little, they would earn outsized profits that would be competed away. If they paid too much, they would be unable to retain workers. Rich countries end up with higher wages in non-tradable sectors not because workers in these specific industries are more productive, but because higher wages in the tradable sector mean that businesses selling non-tradables must pay comparably just to retain staff, who could otherwise take a job elsewhere in the local economy. This, in turn, leads to higher prices for nontradable goods like coffee or haircuts, since the business has no way to extract more productivity from a barista to cover the higher wage bill. This theoretical explanation is known as the Balassa-Samuelson effect [1, 3]. We can use data on the prices of several tradable and non-tradable goods in different cities from Mapping the World's Pricesto test whether the theory holds in practice [2].

Unsurprisingly, it does. Richer cities, measured in real, currency-free terms, do have relatively pricier coffee -- not because their baristas work harder, but because their wages get pulled up by the rest of a more productive local economy. The appendix provides a more detailed explanation of the empirical specification and exactly how the result is obtained.

But the story doesn't end there. Some cities are notable outliers, and the deviations are perhaps even more interesting than the headline relationship itself.

In Milan, a coffee is much cheaper than its iPhone price would suggest. Part of the explanation may be cultural -- Italy's espresso-bar culture runs on thin margins, high turnover, and an ingrained societal expectation of cheap prices at the counter. But there's a more detailed story than that: the Balassa-Samuelson mechanism only works if workers can freely move between a city's high-productivity tradable sector and its non-tradable service sector, via a single competitive local labor market. Café work in Italy is disproportionately informal, part-time, or family-run, and its wages may not track the economy-wide average used in this exercise at all -- meaning the mechanism this model relies on may simply be weaker in Milan than in, say, Zürich (Switzerland), where coffees are markedly more expensive.

By contrast, cities like Bangkok, Manila, Dubai, and Mexico City have coffee that is more expensive than real wages predict. The simplest reason is that the venues sampled skew toward international chains rather than smaller establishments serving locals -- a chain latte in Bangkok and a counter espresso in Rome are different products wearing the same label, which the model has no way to distinguish. A few more substantive economic mechanisms could also be at play. Commercial rent in tourist- and expat-heavy districts may be bid up by international capital rather than local wages, decoupling coffee prices from the local labor market the theory assumes. Cafés serving a visibly split customer base may simply be price-discriminating towards wealthier tourists who are prepared to pay "western" prices. And in cities with significant income inequality, an economy-wide average salary can overstate what a typical local resident actually earns, making local prices look artificially high relative to an inflated benchmark. Testing these mechanisms directly is beyond what the data at hand can support: it is ultimately only a small snapshot of a very complex and interconnected labour market.

None of this means the price on a humble coffee is arbitrary. The headline finding -- that real wages and real coffee prices move together across dozens of very different economies - is a reasonably clean confirmation of a fifty-year-old idea in international economics. The outliers, though, are the more important lesson: the theory describes an average tendency, not a law. Culture, market structure, and who actually buys the coffee all leave their own mark on the price, on top of anything the local labour market is doing.

Deutsche Bank's report prices an identical good (the iPhone 17 Pro), since Apple sets a single global product and prices differ mainly by tax, tariff, and distribution margin rather than local production cost. Because the product is arbitraged internationally, its local price acts as a reasonable numeraire: dividing any other local price by the local iPhone price cancels the exchange rate entirely, since both prices were converted to USD using the same rate. I construct two such ratios, matching each of the 40 iPhone economies to a representative city from Deutsche Bank's salary and cappuccino tables:

$w_i$ is a currency-free measure of local purchasing power ("how many iPhones does a month's salary buy"); $p_i$ is a currency-free measure of local service prices ("how many iPhones does a cappuccino cost"). Note that $w_i$ uses an economy-wide average salary rather than the wage of the specific workers selling coffee, so it proxies for a city's general income level rather than testing wage diffusion into the café sector directly.

The slope implies that a doubling of a city's iPhone-adjusted real wage is associated with local coffee prices (relative to the iPhone) that are roughly 32%higher ($2^{0.397} \approx 1.32$). The change is positive, and less than proportional, consistent with the Balassa-Samuelson prediction that nontradable service prices rise with local income. The pass-through is less than 1 because wage diffusion from the tradable to the non-tradable sector is imperfect: non-tradable wages are sticky, non-tradable productivity grows somewhat independently across countries, and how labour-intensive a "cappuccino" is to produce varies by market. The residual for each economy indicates whether its coffee is cheaper or pricier than its wage level alone would predict.

Since the numeraire relies on the iPhone being priced consistently across economies, a natural concern is that local VAT and import tariffs distort this assumption. Deutsche Bank's own report points out that Türkiye, Brazil, and Hungary as the economies where taxes most inflate the local iPhone price. Re-estimating the specification excluding these three economies gives a slope of 0.364 (SE 0.041, $n = 37$, $R^2 = 0.689$). This is a modest shift and well within one standard error of the baseline estimate. Crucially, the sign of the result is unchanged. This provides a simple "back of the envelope" confirmation that the headline finding is not being driven by the most tax-distorted observations in the sample.

The sample consists of 40 economies, with one representative city chosen per economy from Deutsche Bank's salary and cappuccino tables (largest financial center or capital used as the matching rule). The price of the iPhone 17 Pro is used as the relative numeraire. This is imperfect because local VAT and import tariffs inflate the local iPhone price beyond pure currency effects, adding noise to both ratios without biasing the direction of the relationship (see A.3).

1. Balassa, B. (1964). The Purchasing-Power Parity Doctrine: A Reappraisal. Journal of Political Economy, 72(6), 584--596.
2. Deutsche Bank Research Institute (2026). Mapping the World's Prices -- 2026. Authors: Jim Reid and Galina Pozdnyakova. 13 July 2026.
3. Samuelson, P. A. (1964). Theoretical Notes on Trade Problems. Review of Economics and Statistics, 46(2), 145--154.