VoxEU Column Development Institutions and economics

Institutions matter for growth – but which ones and how much?

How much do institutions matter? This column provides a new insight into measuring their effects, suggesting that a survey of managers’ perceptions of the impact of institutions should be used as an estimate of the effect. It finds that the combined impact of improving public inputs in low-income countries to their level in high-income ones is equivalent to raising output by about 20%.

Most economists today would agree that “institutions matter” for economic growth. This marks a significant shift from twenty years ago and the “Washington Consensus”. Back then, the thinking was that growth required mainly sound policies: macroeconomic stabilisation, trade liberalisation, and privatisation. But since then attempts to quantify the impact of institutions have actually been too successful to be credible.

Empirical studies have isolated more potentially important institutions and impacts than there is growth to explain, and estimates of the scale of the impact of institutions range from miniscule to enormous. We suggest why this has happened, and what the data can really teach us.

There have been numerous cross-country econometric studies of the determinants of both the level and growth of per capita GDP. One approach has been to use proxy measures of the quality of institutions. For example, Hall & Jones (1999) regress output per worker on a proxy measure of “social infrastructure” constructed from indicators of government antidiversion policies and openness to international trade. Meanwhile Acemoglu, Johnson and Robinson (2001) regress GDP per capita on two alternative measures, one an index of protection against expropriation and the other an indicator of constraints on the executive. Both studies find apparently large effects of institutions – as proxied – on long-run growth outcomes.

But what do the proxies really stand for? There are too many possible underlying causes for the data to have any chance of choosing between them. The list of measures used in empirical studies compiled by Durlauf et al. (2005) begins with capitalism, capital account liberalisation, corruption, democracy, demography … and that only takes us to “D”. Separately identifying the effects of each is not possible with only 100 or so historical country growth experiences. Nor are there stable results across studies and estimates, in which a specific subset of institutions or policies always matters. Commenting on this state of affairs, Easterly (2009) has described the past two decades as a “Dark Age of Growth Econometrics”. This is possibly an overstatement, but probably not by much.

Unfortunately, micro-level data on firms do not solve the problem. Over the past 10 years, the World Bank and EBRD have collected a vast range of data on firms in their surveys of the “investment climate” and the “business environment”. The datasets now cover over 70,000 firms in about 200 surveys in 100 countries. Impressive as these numbers sound, there is less effective variation in the data than they imply. Institutions vary far more across countries than they do within countries. The legal environment faced by two different firms in the same country is essentially identical, and only the cross-country dimension provides any identifying variation that would enable researchers to uncover differential impacts of the legal environment on firm performance. The econometric challenge of trying to tease apart differences in the institutional environment faced by firms in a single country, while avoiding the problem of endogeneity, is too much for the data to bear. This is why a recent careful study that tried to do just this, Commander and Svejnar (2009), found largely null results.

New insight on measuring institutional effects

In our recent work (Carlin et al. 2010), we propose a different and much more informative way to interpret the micro-level data from firms. Instead of estimating regressions in which productivity and growth are the dependent variables, and institutions, somehow measured, are regressors, we use direct reports by managers of the impact of various institutions on their own firms' performance. These reports, we argue, are not observations of institutional quality to be plugged into a regression in order to generate a coefficient representing their impact on growth; they are direct reports of that regression coefficient as managers perceive it to apply to their own firm. We provide a formal framework in which these direct assessments by managers can be interpreted and used.

Our direct measures of the impact of institutional quality come from answers by managers to the following question: “Can you tell me how problematic are these different factors for the operation and growth of your business?” The factors covered include aspects of physical infrastructure (telecoms, electricity, transport), access to skilled labour, the stability of the macroeconomic environment, tax and customs administration, licensing, the legal system, labour regulation, and the impact of crime and corruption. Answers are on a scale from ‘1 - no obstacle’ to ‘4 - major obstacle’. These data have the great advantage that they employ a common scale to provide a direct assessment of the impact at firm level of a wide range of institutions. But before they can be used we need a framework in which they can be interpreted.

Early attempts to use these data interpreted the answers as objective measures of the business environment and put them on the right hand side in performance regressions. An example illustrates why this is wrong.

If we ask a firm in a rich country about the importance of internet access to their business, we may well get the result that access is problematic. By contrast, asking a firm in a poor country, where the internet has never been used, will get the response ‘not an obstacle’. The quality of the internet infrastructure is in fact better in the rich country, but its remaining deficiencies are more of an obstacle to growth, because the rich country has many more firms that depend upon it. Use of these data as objective measures of the quality of infrastructure rather than of the impact of the infrastructure is misguided.

The framework we propose instead interprets the answers as measures of the cost to the firm, in terms of lost production, of poor quality or unreliable public inputs. (Our formal framework incorporates unreliable public inputs into Kremer’s (1993) O-ring production function.) With answers from a representative sample of firms across the range of elements of the external environment, we can observe, for example, how judgements of the costs imposed by unreliable public inputs vary with firm characteristics. Is it the case that, for instance, small firms or exporting firms or foreign-owned firms tend to be differentially burdened by poor quality institutions? In addition, with data from a large number of countries across the full range of GDP per capita, we can see how the reported impact of different institutions and types of infrastructure varies by the level of development.

We find, for example, that it is indeed the case that larger and higher productivity firms generally report higher costs of institutional and infrastructure constraints. These results support our interpretation of the answers – larger firms are better placed to take advantage of public inputs yet they report them as more of a burden on their activities than do smaller firms. Another finding comes from the stylised fact of rural/urban dualism in economic development. Given that in the early stages of development, urban areas are the pole of growth attracting labour from the hinterland, the prediction is that in poor but not rich countries, there should be higher reported costs of public input constraints in urban than in rural areas. In fact, we find this pattern in the data across a wide range of institutional constraints.

We also use the data in these surveys to quantify the aggregate output losses of poor quality public inputs. Managers report output losses due to electricity outages as well as rating electricity as an obstacle on the 1-4 scale. Using this calibration, we can calculate an index of output relative to the counterfactual in which public inputs are fully reliable. We find that the combined impact of improving the reliability of public inputs in low income countries to their level in high income countries would be to raise output in the former by about 20%. The output effect is higher for improvements in institutions as compared with improvements in physical infrastructure and access to skills.

Table 1. Index of output relative to the counterfactual in which public inputs are fully reliable1

Physical infrastructure (including land
& skills
Institutions (macro,
tax administration, labour, customs, licensing, legal, corruption, crime)
High income
Upper-mid income
Lower-mid income
Low income
Full sample

These are static estimates and take no account of the dynamic effects that better institutions might produce via the choice of different factor inputs and technology by existing firms, and via new entry. Nevertheless, this exercise suggests that we should be modest about the share of GDP differences that can be attributed to institutional differences of this kind. To take an example from Acemoglu, Johnson and Robinson, even if institutions broadly defined raise Chile’s income by a factor of 7 as compared with Nigeria’s, improving the reliability of specific identifiable institutions cannot be expected to have anything like the same scale of effect.

Finally, we can use this framework to rank constraints at the country level. In our paper we look in detail at six countries: Brazil, Chile, Bangladesh, Pakistan, Mozambique and Senegal. We find, for example, firms in Brazil and Chile are very concerned about macro stability and policy uncertainty, but it turns out this is standard for countries of that level of development. What Brazilian and Chilean firms are unusually concerned about – compared to other middle income countries – is labour regulation. Firms in all four low-income countries, by contrast, rate electricity as a major constraint, but again, and not surprisingly, this is typical for low income countries. But once we benchmark the countries based on their income level, we see that Bangladeshi, Pakistani and Senegalese – but not Mozambican – firms are indeed unusually concerned about electricity supply constraints. The benchmarks allow the policymaker to ask whether the monetary and political costs of remedying the public input identified as a priority are especially high in their country or whether there are unexploited social investment opportunities.

Can our approach be employed by other researchers for specific purposes? We think so, but we also believe the scope of potential applications would be increased if the survey instrument were modified. In particular, quantitative conclusions using these surveys would benefit from revisions that would increase the reliability and range of options for calibration of reported constraints to real or financial variables.


1 In column 1, the sample is 28,164 firms from 57 countries, in col. 2, 46,869 firms from 69 countries; and in col. 3, 34,424 firms from 57 countries.


Acemoglu, Daron, Simon Johnson, James A. Robinson (2001), “The Colonial Origins of Comparative Development: An Empirical Investigation”, American Economic Review, 91(5).

Carlin, Wendy, Mark Schaffer, and Paul Seabright (2010), “A framework for cross-country comparisons of public infrastructure constraints on firm growth”, CEPR Discussion Paper 7662.

Commander, Simon and Jan Svejnar (2010), “Do Institutions, Ownership, Exporting and Competition Explain Firm Performance?Review of Economics and Statistics, forthcoming.

Durlauf, Steven, Paul Johnson and Jonathan Temple (2005), “Growth Econometrics” in Philippe Aghion and Steven N. Durlauf, eds, Handbook of Economic Growth, North-Holland.

Easterly, William (2009), “Why there’s no “GrowthGate:” Frustration vs. Chicanery in Explaining Growth”, Aidwatch, 10 December.

Hall, Robert E and Charles I Jones (1999), “Why Do Some Countries Produce So Much More Output Per Worker Than Others?”, Quarterly Journal of Economics 114(1).

3,255 Reads