Vast differences in living standards across countries remain a defining feature of the world economy. Despite decades of development research, policy, and progress, the world’s poorest countries still produce between 30 and 50 times less output per worker than the richest ones. What makes some countries so much poorer than others? One candidate explanation is that firms in poorer countries operate fewer machines, tools, and other capital equipment and are hence less productive. In agriculture, for example, they operate around 42 times less horsepower per acre of farmland compared to richer countries. Understanding income differences hence requires understanding differences in capital investments (Hsieh and Klenow 2010).

But why is there so little capital in poor countries? Standard economic theory suggests that low capital investment either reflects low benefits or high costs to installing capital in developing countries (or both). The focus of substantial previous work in macro and development economics has been the former of these two explanations: researchers have measured how much extra output and revenue firms in developing countries can generate from adding additional machines to their production process (e.g. de Mel et al. 2008) and modelled why these benefits might vary across the process of development (Banerjee and Duflo 2005, Caselli and Feyrer 2007). At the same time, much less is known about the costs that accrue to keeping capital in good repair. Businesses considering capital investments, however, very much do care about this side of the equation – the most productive technology might not be economical if it immediately breaks down and repair is expensive, delayed, or unattainable.

A new study in Uganda

In Graff (2025), I approach the question of how expensive it is to keep machines running in developing countries from both an empirical and theoretical angle. I measure the costs of capital breaking down using a large firm survey I conducted among firms across Uganda and find that they are substantially higher than comparable measures in the US. This stands in contrast to existing research which has often assumed that these costs are outside the realm of economics and rather represent a universal engineering constant, identical across the world. I then use a macroeconomic model to think through why I find such different costs and what it implies for our understanding of the sources of global income differences.

I first document that it is often quite expensive to keep machines in good repair in Uganda. Anecdotally, firms frequently report that shiny new equipment survives only a few years, as otherwise minor mechanical issues result in cumbersome searches for unavailable spare parts, interactions with unqualified or even fraudulent mechanics, or long delays to access adequate repair and maintenance. To transform these anecdotes into hard data, I conducted a large survey activity among 1,400 firms in two sectors central to Uganda’s economy: food processing (operating grain mills and coffee factories) and motorcycle transport (operating small motorcycle taxis). I interviewed firms about the frequency at which their machines experience breakdown, who repairs broken machines and at what price, how long it takes to get machines back online, as well as how often machines need to be replaced because no adequate repair could be accessed. I compile these answers into a unified ‘replacement investment rate’, capturing how much a firm needs to reinvest over the course of a year to keep its capital stock constant. For example, if a firm operating a $1,000 grain mill in my sample spends $300 over the course of twelve months on repairs, maintenance, or the occasional full replacement of its machine, this rate would be 30%. Akin to a depreciation rate in standard macroeconomic models, high rates of replacement investment imply that every piece of capital requires further continuous investments as it would otherwise become unproductive, rendering it less profitable to invest in in the first place.

Using this survey, I can isolate a series of new facts about the nature of machine repair and depreciation in Uganda. First, keeping machines in good condition is costly: the average firm in my setting faces replacement investment rates between 37% and 55%, about twice as much as comparable numbers from engineering estimates about replacement and repair standards in the US. Second, this pattern also replicates within Uganda: otherwise identical machines face much higher replacement investment rates in the least developed parts of Uganda, compared to cities and richer areas (Figure 1a). Intuitively, when a motorcycle breaks down in a remote village, it is often much more difficult to find adequate mechanics and spare parts compared to the same motorcycle breaking down in Kampala. In response, firms often wait longer to get their machines repaired, especially when they have machines that are not very common (Figure 1c). These trends are particularly severe for most microenterprises, which are not large enough to have their own in-house mechanic personnel to take care of their machines. Larger firms that have repair capacity within their boundaries (as is often the case in richer countries) pay less, leading to a strong gradient of capital depreciation rates with firm size (Figure 1b). Qualitatively, these dynamics are of great importance to firms – whether or not machines are easily repaired is a key concern when deciding which machines to invest in (Figure 1d).

Figure 1 Stylised facts about replacement investment in Uganda

Reasons for high depreciation rates

Why is it so hard to keep machines in good repair in less developed parts of Uganda? One hypothesis I explore in my work is that the very nature of how machines break down makes it tricky to repair them quickly in more remote areas. the key idea is that demand for machine repair is both unpredictable (i.e. it is hard to know when a machine will next break down) and timely (i.e. once it has broken down, you would ideally want it repaired as quickly as possible). These two ingredients intuitively make it very hard to repair machines quickly if there are not many of them around. The owner of the very first grain mill in a remote village will find it very expensive to have mechanic knowhow or the necessary spare parts to repair a machine just in case it ever breaks down. A busy industrial area, however, has enough machines that might plausibly break on any given day, so that quick and able repair is more easily supplied.

A related form of these dynamics is present in the common wisdom that it is much harder to repair vehicles of rare brands or makes. In Uganda, for example, the market for motorcycles is dominated by the leading brand Bajaj, which means it is much easier to repair a Bajaj motorcycle if it breaks relative to a more obscure brand, which in turn makes Bajaj even more attractive, and so on. My hypothesis is that a similar effect plays out across the path of development at large: if there is hardly any capital (of any kind), repairing the little equipment that does exist is expensive, which is partly responsible for why there is so little investment in the first place.

Implications and outlook

How large are these effects and can they plausibly account for differences in living standards between economies? To answer this question, I enrich a canonical Solow (1956) style growth model with a repair sector and capital depreciation that responds to the level of local economic development. Calibrating the model to the Ugandan economy, I can then assess how important this channel really is. I find that accurately accounting for depreciation plausibly being higher in poorer areas explains around 7%-9% of income differences across local regions within Uganda. In other words, a little less than one tenth of why the richest parts of Uganda are richer than the poorest ones is due to this previously overlooked fact that depreciation is higher in more remote and less capital-intensive areas.

Since my detailed firm survey only covers Ugandan firms, the question remains of how similar these dynamics play out in other contexts and across the globe. Collecting similar data on the nature of capital repair and depreciation in more countries could help our understanding of just how important this previously ignored dimension of investment really is. Given that the magnitude of income differences across Ugandan regions is similar to the magnitude across countries in the world, it is at least plausible that heterogeneity in depreciation rates across the world could have a similar impact as they do within Uganda.

Explaining what drives capital investments is crucial to understanding differences in living standards across the world. My research in Uganda rests on a straightforward insight: firms will not make many investments if their purchased machines break down immediately or are very expensive to repair, regardless of how productive they otherwise would be while running. The key innovation of my approach is to understand these costs as an economic object, shaped by supply and demand on the market for repair. Both empirical measurement and a model suggest that repair is more expensive, and hence depreciation more severe, in areas that lack thick capital markets, widening the gap in income per capita across places.

References

Banerjee, A and E Duflo (2005), “Growth Theory through the Lens of Development Economics”, in Handbook of Economic Growth, Vol. 1, Elsevier.

Caselli, F and J Feyrer (2007), “The Marginal Product of Capital”, The Quarterly Journal of Economics 122(2): 535–568

de Mel, S, D Mckenzie, and C Woodruff (2008), “Returns to Capital in Microenterprises: Evidence from a Field Experiment”, The Quarterly Journal of Economics 123(4).

Graff, T (2025), “Depreciation and Growth: Evidence from Machine Repair in Uganda”, Working Paper

Hsieh, C-T and P J. Klenow (2010), “Development Accounting”, American Economic Journal: Macroeconomics 2(1): 207–223

Solow, R M (1956), “A Contribution to the Theory of Economic Growth”, The Quarterly Journal of Economics 70(1): 65–94.