← All articles
Manufacturing

Lower wages don't mean lower production costs

When companies look for a cheaper place to make something, the first number on the table is usually the hourly wage. It is easy to compare and it looks convincing. But a factory does not sell hours. It sells good parts. What a part really costs depends on how many parts one worker makes, how many of them end up as scrap, and how big a share of the cost is labour at all.

I went through several studies to see when a low wage really saves money and when it does not. The answer depends mostly on the product.

Shirts and shoes: low wages win

The World Bank compared factories in Ethiopia and China making the same products. In polo shirt factories, an unskilled worker in Ethiopia earned $26–48 a month. In China it was $237–296.

Ethiopian workers made fewer shirts per day, 7–19 compared with 18–35 in China. But the wage gap is so large that labour per shirt still comes out lower in Ethiopia. By my estimate from the study's numbers it is about $0.31 per shirt, against $0.44 in China.

There is a catch. Buyers paid $2.95–3.41 for an Ethiopian shirt and $5.38–5.80 for a Chinese one. Quality, finish and reliable delivery are worth money too.

Leather shoes look similar. Labour in Ethiopia is about $5 cheaper per pair. That is more than enough to pay for imported leather, which adds about $1 per pair. Quality is again lower and the shoes sell for less.

So far the low wage works. These are simple products, and a worker in Ethiopia is slower, but not dramatically slower.

Wooden chairs: low wages lose

The same study, the same country, a different product. A worker in a Chinese furniture factory made 4.5 wooden chairs a day. In Vietnam it was 1.9. In Ethiopia 0.3.

Ethiopian wages were about five times lower than Chinese wages. Output per worker was fifteen times lower. Labour per chair came out at $3 in China and $10 in Ethiopia. Ethiopian workshops also used about 500 g of varnish per chair, Chinese ones about 50 g. In total, a chair cost more than twice as much to make in Ethiopia.

A chair needs more skill than a shirt. Cutting, joining, sanding and finishing all have to be right. When they are not, the low wage disappears into slow work and wasted material.

Metal parts: scrap cancels the wage advantage

In metal products the World Bank found 32% scrap in Ethiopian plants, against about 5% in other countries. In bottle cap production, Ethiopian workers made about half as many pieces per day as Chinese workers. Production cost ended up about 20% higher than in China, mostly because steel was more expensive, but the scrap rate shows where the wage advantage goes.

A third of the material thrown away is not a wage problem. It is a process problem.

Coating and machining: the same pattern in Europe

This is not only about Africa and Asia. Eurostat publishes data on labour costs and value added for individual sectors in EU countries.

In metal coating, an hour of work in Romania costs a company €9.45. In Czechia it costs €15.37, so Romania is almost 40% cheaper per hour. Yet in Romanian coating shops labour takes 62% of the value they create. In Czech shops it is 40%.

Machining shows the same direction. An hour costs €8.74 in Romania and €15.61 in Czechia. Labour takes 51% of the value created in Romania and 43% in Czechia.

Put simply, a Czech shop pays more per hour but gets more out of each hour. These are averages for whole sectors, so product mix and prices play a role too. But the direction is clear.

Cars: low wages win, but not automatically

Cars are a complex product, so one might expect the same result as with chairs. On average, that is not what happens.

Oliver Wyman estimates labour cost per vehicle at about $305 in Mexico, $597 in China and $1,341 in the United States. An assembly worker in Mexico earns about $5 an hour, in the US about $25. On average, Mexico comes out well ahead.

But that is an average across many plants. Car plants differ a lot, even inside one company and even when they are highly automated. In the 1980s, GM's most automated plant, Hamtramck in Michigan, needed 33.7 hours to build a car. NUMMI in California, run by Toyota and GM with less automation, needed 19 hours. A study of 62 assembly plants by John Paul MacDuffie found the same pattern: well-organised plants needed about 21 hours per car, the others about 37. Automation did not explain the gap. The way the plant was run did.

The same is true today. Internal Volkswagen figures reported in 2026 show 29 cars per employee a year in Emden, 40 in Palmela in Portugal and 51 in Tianjin in China. One company, very different results.

A car plant also depends on what is around it: suppliers, toolmakers, maintenance firms, logistics, and technicians who can get a stopped line running again. Siemens estimates that one hour of unplanned downtime costs a large car plant $2.3 million. Mexico still imports about 90% of its moulds and dies, and about three quarters of automotive companies there say they struggle to fill jobs, maintenance technicians among them.

So even car assembly, which can look simple from the outside, is not simple. Mexico wins on labour per car because carmakers and suppliers have spent decades building plants, suppliers and skills there. A new plant without that base should not expect the average from day one.

Chips: wages barely matter

At the top end are semiconductors. TechInsights calculated that labour costs in the US are roughly three times higher than in Taiwan. Yet a wafer from TSMC's plant in Arizona costs less than 10% more than the same wafer from Taiwan. Labour makes up less than 2% of the cost. More than two thirds is equipment.

Here the question of wages hardly comes up. What decides is the process, the yield and the equipment.

What this means

Low wages help when output per worker stays close. That happens with simple products like shirts and shoes, and in places like the Mexican car industry, where the plants, suppliers and skills have been built up over decades. When the product needs more skill, or when the process is not under control, the lower wage is quickly eaten by slower work, scrap and rework.

So when someone shows me a quote from a cheaper country, the hourly rate is not the first thing I look at. I ask how many good parts one person makes in an hour, how much ends up as scrap, and how long it takes before a new plant runs as well as the old one.

Where do low wages stop paying off for your products?

Sources: World Bank, Eurostat, Oliver Wyman, INA, TechInsights, Siemens, MIT, Bild. The labour cost per shirt is my estimate from World Bank data.

Thinking about moving production, or bringing it back? Let's talk.

Get in touch LinkedIn