2 - Tools for an Institutional Diagnostic

Summary

This chapter presents the methodology used in forming a comprehensive institutional diagnostic of the four case study countries included in the IDP project. An institutional diagnostic of development is defined as an exploration of how the institutions of a country affect the functioning of its economy, its dynamics, and the policymaking process, the ultimate goal being to detect the most serious flaws that hinder development. The methodology includes three steps. A first step reviews the economic, social, and political development of the country, examining institutional quality indicators, and soliciting from various types of decision makers their views on potential institutional obstacles to development. The second step consists of an in-depth analysis of selected critical economic areas, where the relationship between the institutional context of a country and its development is the most apparent. The third step synthesises what has been learned into a list of basic institutional problems, their economic consequences, and, most importantly, their causes, proximate or more distant, as well as the potential for reforms in view of the political economy context.

I Introduction

The case studies summarised in this volume, and which serve as raw material for our reflection on institutions and development, follow a particular analytical approach. Conceived in the same spirit as the ‘growth diagnostic’ introduced in the development literature by Hausmann, Rodrik, and Velasco (2005), the institutional diagnostic approach consists of identifying within a particular country at a given point in time which institutional dysfunctions or weaknesses may be responsible for hindering faster, more transformative, more inclusive, or more sustainable development. Based on such diagnostics in several case studies, our final objective will then be to examine whether common weaknesses or ‘generic’ institutional issues arise, which should help understand better the general relationship between institutions and development and provide a kind of analytical grid to explore development-costly institutional flaws in other countries.

The preceding reference to the growth diagnostic approach to the identification of development constraints must not be taken rigorously. A huge difference between that approach and the institutional diagnostic is that no formal general model linking institutional constraints or deficiencies is available in the latter case. Therefore, neither an a priori list of potential institutional constraints on growth and development, nor a specific variable unequivocally signalling the strength of these constraints is available. This is a major difference with the growth diagnostic methodology, which benefits from an a priori knowledge of the nature of the economic determinants of growth. Issues involving institutions are more complex. To illustrate, corruption does not necessarily imply slow growth, autocratic leaders are not systematically associated with development failures, and informal institutions may work better than formal institutions in overcoming key economic constraints.

Rather than designing and following a questionable, predetermined analytical path, we adopt for the IDP a methodology that can be characterised as heuristic. For a given country, it consists first of gathering all information available on the quality of institutions and their possible role in constraining development. This includes exploiting the international databases of institutional or governance indicators to see how the country differs from well-chosen comparator countries, and using formal and informal opinion surveys addressed to those local experts and people whose activity is likely to be directly affected by, or who are knowledgeable on the way institutions work. Such a step partakes of a sort of ‘mechanical’ approach to the identification of institutional weaknesses (mechanical in the sense that ways to process the information are well known). It is succeeded by a more inductive approach. Starting from an in-depth analysis of the historical economic development process of a country, including growth diagnostics when available, the idea is to first identify apparent economic weaknesses – or constraints – and then to ponder over the possible institutional causes behind them. This can be done at the aggregate level or by focusing on restricted thematic areas where the relationship between specific institutions and economic development constraints is likely to be easier to detect.

The final challenge is then to put all the collected pieces of evidence together, and then propose an institutional analysis based on them. This requires that we start by defining what seems to be the most binding institutional weaknesses of a country as well as their economic consequences, and that, thereafter, we diagnose their likely proximate causes and deep determinants. Such an endeavour inevitably leads to the question as to why reforms susceptible of attenuating or removing an institutional weakness were not implemented, which is often tantamount to investigating the political economy aspects of that particular institutional issue.

The above-described steps are presented and discussed in the rest of the chapter, approximately in the preceding order. The first two sections deal with mechanical approaches based on the use of institutional or governance indicators, on the one hand, and on opinion or expert surveys, on the other. The next two sections focus on more inductive approaches to the institution–development relationship in a country. A final section then presents the ‘diagnostic table, an instrument that was found particularly helpful to synthetise and summarise what was learned in all the preceding steps. Moreover, it has the advantage of bringing to the limelight the proximate causes and the deep factors at work behind the identified development-constraining institutional weaknesses.

Before getting to the crux of the matter, it is necessary to repeat that the present volume and the IDP case studies it relies upon do explicitly deal with low-income or lower-middle-income countries, that is to say, countries in their early stage of development. Therefore, several arguments developed in the rest of this chapter might not have the same relevance if we were dealing with emerging countries.

II Benchmarking A Country’S Institutions Using Global Institutional/Governance Indicators

Imagine that a set of indicators is available that describes the quality of countries’ institutions in their various dimensions and, as a result of a set of regressions across both countries and time periods, the impact of each indicator on economic growth and several other development outcomes is known. The institutional diagnostic of development in a country would then be greatly facilitated. The set of indicators would provide this right away. The issue would then be to go from the indicators to the institutions whose functioning they describe, and then to investigate how changes can be made to improve their performance.

Unfortunately, things are not that simple. First, the significance of the correlation between the quality of institutions and development outcomes varies according to what development outcome is chosen. Second, when the correlation is significant, the causality behind it is ambiguous: does it run from good institutions to favourable development outcomes or the opposite? Third, it is not clear whether available indicators describe the quality of specific institutions – like the accountability of the executive or the independence of the judiciary – or some joint observable outcome of the functioning of these institutions. Available indicators are often presented as ‘governance’ indicators, thus describing how the institutional framework of a country makes its governance more, or less, development efficient, rather than describing the quality of a specific institution. Fourth, the precision of indicators is limited so that there would be much fuzziness in using them to benchmark a country relative to others.

This section elaborates on whether the indicators available in various cross-country databases may reveal obstacles to institutional development in a country. Even though such a capacity may be limited, it nevertheless shows how indicators can be used to expose the idiosyncrasy of a country in the space of broad institutional domains, possibly paving the way to a deeper institutional diagnostic. In short, it shows that they can be a useful exploratory tool.

A To What Extent Do Governance Indicators Reveal Institutional Obstacles to Development?

The use of indicators meant to describe the quality of institutions to make a judgement about whether institutions in a country are more or less favourable to economic growth, and more generally to development, is justified by the theoretical arguments surveyed in the preceding chapter and, supposedly, by empirical evidence. The latter is statistically fragile, however, and not without ambiguity. Precautions should thus be taken in using those indicators as a tool to identify institutional strengths and weaknesses. This is even more necessary as the indicators themselves provide descriptions of the quality of institutions that do not exhibit the precision required by a diagnostic.

Empirical evidence points to a strong correlation between institutional indicators and the level of GDP per capita across countries. The problem is that this correlation is consistent with causality going both ways: from better institutions to faster growth and from growth to better institutions. Instrumental variables that are assumed to be correlated with institutional quality but not with the level of development or past growth are used to control for this problem.¹ This procedure tends to confirm that institutions affect economic growth, or the contemporaneous level of income per capita, among developing countries. However, the exogeneity of these instruments with respect to economic development is often debatable. On the other hand, the cross-country relationship between institutional indicators and the average rate of GDP per capita growth over ten- or twenty-year periods of time is weaker. Moreover, when other country characteristics are introduced to control for other exogenous factors that may condition growth it turns out that the effect of institutional indicators and their statistical significance tends to vary with the set of controls being used, which does not suggest a robust relationship.

These issues are thoroughly discussed in a recent survey by Stephen Durlauf (2020) of the imposing cross-country growth and institution literature of the past twenty years.² Its main conclusion is that, if there is no doubt about the influence of the quality of institutions in general on economic development, the exact channels for such an influence are essentially ambiguous. As an example, consider the following three well-known studies: Acemoglu et al. (2001) provided evidence on a cross-section of countries that the protection of property rights delivered by a country’s institutions matters for development; likewise, Mauro (1995) found with another instrument that corruption significantly and negatively affects growth; whereas Dawson (2003) applied Granger causation³ methodology on a panel of countries to show that the degree of economic freedom influences economic development. Those three studies show that, indeed, on average across countries, the quality of institutions matters for economic growth, but they do not say much about what institutions matter, and by what mechanisms these relationships are obtained. There are many different types of corruption – high-level politicians or civil servants siphoning away public money, taxpayers bribing tax authority personnel, the petty corruption of police officers – with a priori differentiated effects on economic efficiency and growth. A lack of protection of property rights may be due to corruption, to a weak judicial system, or to predatory rulers, while a lack of economic freedom may be due to over-regulation but also to excessive taxation or weak property rights. Surely the fact that significant relationships are found in those three studies, which are very representative of the empirical institution–development literature, means that the quality of ‘some’ institutions affects growth and development. Yet no one would accept an analyst making a diagnosis about what is wrong or right in a country’s institutions concerning economic growth based on those sole relationships.

As can be seen from the previous examples, the difficulty is as much in providing evidence of a causal relationship as in identifying what the institutional indicators used in cross-country analyses of the institution–development relationship stand for. To a large extent, this ambiguity results from the fact that these indicators most often describe the consequences of something being wrong in the way in which institutions function but not what is wrong. In other words, they point to symptoms rather than dysfunctions. Corruption is a case in point. It can always be described as the consequence of a judicial system that is unable to enforce the law – for instance, due to a lack of resources or (honest) personnel – but it may also be the consequence of loopholes in the law or in the regulatory framework that create rent-seeking situations, or of a lack of transparency of operations in the public sector. Yet the information gathered from experts relates to their perceptions of corruption, rather than the relative importance of the institutional causes behind it. In other cases, indicators rely on a set of very precise questions that are then aggregated into a single index. Yet the field covered by these questions is often incomplete. For instance, the ‘Rule of Law’ indicator in the Global Integrity Index relies exclusively on questions about the independence of the judiciary from political influence and the transparency of judgements, but no information is gathered on the time it takes to clear a case, the degree of corruption of judges and judicial officers, or their level of competence. By contrast, other indicators rely on long lists of questions covering various, often heterogeneous subfields.

Where does all this leave us concerning the institutional diagnostic of a specific country? Mostly to the fact that institutional indicators only provide a measure of the overall quality of institutions and, at best, some more detailed information on the strengths of various types of symptoms that may point to specific institutional flaws. It must be clear, on the other hand, that the measurement precision of these indicators is limited. In the Corruption Perceptions Index published by Transparency International, for instance, accounting for a reasonable degree of measurement error, it is not possible to say whether Kenya, ranked 124, significantly differs from Madagascar, ranked 149, or Egypt, ranked 117. When benchmarking a country relative to others, the lesson is that not too much meaning must be attributed to a country ranking 10 or 15 slots ahead or behind another. Attention should focus on those institutional domains where indicators show substantial differences.

A last remark is of importance when using global rankings of institutional indicators as an input into the institutional diagnostic of a country. It is that the correlation observed across countries between indicators and development outcomes applies to the ‘average country’, not to all single countries, far from it. In other words, it is not because the Risk of Expropriation of Foreign Investment published by Political Risk Services is shown to affect negatively the level of GDP in a cross-section of countries⁴ that a specific country where this risk is perceived to be high will necessarily underperform. There is some strong idiosyncrasy behind statistically significant cross-country relationship, which cannot be ignored when analysing a single country.

B Benchmarking Low-Income and Lower-Middle-Income Countries According to Institutional Indicators

The first question to ask when using institutional indicators to benchmark a country against others is what indicator to use. As mentioned earlier, many indicators are available, and even when they are supposed to cover the same institutional domain it turns out that they may substantially differ in some cases. Aggregating indicators covering close domains is a way of extracting from their diversity more robust differences across countries. This is the approach taken by the authors of the widely used Worldwide Governance Indicators (WGI).⁵ An alternative approach, based on the extensive set of institutional and governance indicators stored in the Quality of Government (QoG) database⁶ and focused on developing countries only, was also used in the IDP case studies. They are briefly described in turn.

The WGI indicators cover six broad domains: (i) rule of law; (ii) voice and accountability; (iii) control of corruption; (iv) government effectiveness; (v) political stability; and (vi) regulatory quality. Each aggregate indicator results from a statistical procedure that involves extracting from the large number of individual indicators which seem to fit the domain under analysis a ‘common factor’ in the way these various indicators rank countries. Practically speaking, this is done through looking for a linear combination of individual indicators whose average correlation, so to speak, with each indicator is the closest.⁷ This common factor is then taken as the aggregate indicator which best describes the quality of institutions in the domain being considered.

The WGI methodology for defining aggregate indicators regroups individual indicators available in various sources according to the six institutional domains listed above on an a priori basis. An alternative approach consists of being agnostic about these domains and regrouping individual indicators according to their proximity in the way they rank countries. The number of groupings is decided a priori, and a ‘cluster analysis’ procedure determines which indicators enter each group. In other words, each group comprises a set of indicators that are highly correlated to each other in the way they rank countries, while this common ranking is made to differ as much as possible across groups. Then a common factor is identified in each group that summarises the way indicators in that group ranks countries. As the clustering is the result of a purely statistical procedure operated on the whole set of individual indicators, it is not clear whether they should be conceptually close to each other. Yet it turns out to be the case that they are close, suggesting that available individual indicators from a host of different sources tend to describe the functioning of institutions within a country according to a small number of key dimensions.

In the application of this methodology to some of the IDP case studies, the QoG database was restricted to developing and emerging countries so as to avoid aggregate indicators being mostly based on differences between advanced and less advanced economies, which may be a problem of the WGI indicators. Even though the dataset comprises more than 2,000 indicators, coming from practically all sources of institutional/governance indicators available, only those that were available for all countries in the sample were kept. They numbered 350, which were then clustered in six groups or ‘categories’.⁸ Upon inspection of the indicators they comprised, the six categories were identified as:

1. a. democratisation;

2. b. human rights;

3. c. administrative capacity;

4. d. control of corruption and rule of law;

5. e. conflict and violence; and

6. f. competitiveness.

It is interesting that some of these categories very much overlap with the WGI domains – that is, ‘democratisation’ and ‘voice and accountability’; ‘administrative capacity’ and ‘government effectiveness’; and ‘control of corruption and rule of law’, corresponding to the two domains with the same name in WGI. Yet the overlap is far from perfect since the ‘human rights’ and ‘competitiveness’ categories have no clear counterparts among the WGI domains, even though the primary sources used to define the latter include datasets oriented towards competitiveness or human rights issues.

With these institutional indicators at hand, a second issue is which comparator countries are to be included in the benchmarking. Clearly, it does not make sense to compare the institutional quality of low- or lower-middle-income countries to countries that are much more advanced in the economic development process, have the resources to maintain better institutions, and whose population demands better-performing institutions. The comparison must allow for income differences, but within a reasonable range of variation.

Two sets of comparator countries were used in the IDP case studies. The first one comprises neighbour countries, with the justification that these may share with the country under analysis a common geo-physical context and, depending on the region, some common cultural or historical references – such as, for instance, the same past coloniser. Lack of significant differences within this set of countries may then reflect the strong influence of this context, as well as some homogeneity in terms of living standards. By contrast, variations across countries could mean either that the geo-physical and cultural context are not major determinants of the institutional features of countries in that region, or that the region is rather heterogeneous with respect to these geographical and historical characteristics. The significance of a country differing from comparators may be stronger in the second case.

The second set of comparator countries consists of countries which were at the same income level as the country under analysis two or three decades ago but that have managed to grow substantially faster since then. The question then is whether some institutional domains were of a better quality in the latter when growth accelerated, compared to the country being studied. The difficulty is that institutional indicators rarely go back as far as two decades or more. The comparison can only be performed on relatively recent years, and, in some cases, there is no possibility of going back in time.⁹

Examples of the kind of benchmarking based on the WGI and QoG-based indicators are given in Figures 2.1 and 2.2 for Tanzania. The radar chart in Figure 2.1 compares that country with its neighbours in 2019. The WGI indicators range from −2.5 (worst institutional quality) to 2.5 (best), and the standard deviation of the six indicators – within the sample of developing and emerging countries – is around 0.5. The radar chart is thus constructed in such a way that the difference in graduations along all axes is precisely around one standard deviation, which allows us to pass a judgement on the significance of differences between countries. On the other hand, country scores tend to concentrate on the negative part of their interval of definition, which means, unsurprisingly, that governance and institutional quality in this set of low- or lower-middle-income countries are below the world median¹⁰.

Figure 2.1 Comparing Tanzania and neighbour countries according to the WGI indicators, 2018

Figure 2.2 Comparing Tanzania and neighbour countries according to the QoG-based indicators, 2018¹¹

Looking at the figure from the point of view of Tanzania, the conclusion is undoubtedly that there is no difference when compared to the bulk of its neighbour countries, except Rwanda and Burundi, since differences with other countries never exceed one standard deviation. When restricting the comparison to these countries, one would tend to conclude, as suggested earlier, that Tanzania shares with them common geographical, demographic, and historical factors that lead to comparable institutional quality features. On the other hand, except for the rule of law in Mozambique, scores tend to be similar across the six institutional domains, so that no particular domain can be singled out for special attention later in the diagnostic exercise.

The chart may also be looked at from the point of view of other countries. If a diagnostic had to be established for Rwanda, for instance, this benchmarking exercise would have led to the conclusion that Rwanda tends to perform better than other countries in the region, except for a very low score in ‘voice and accountability’: that is, the democratic functioning and transparency of the government’s action. This is a valuable clue for an institutional diagnostic. Likewise, Burundi is shown to perform worse than other countries in the region – and as a matter of fact very badly in absolute terms, but a little less badly for ‘regulatory quality’ – whereas Mozambique would be comparable to other countries if it were not for the ‘rule of law’ domain.

To evaluate the consistency of using the WGI indicators, Figure 2.2 shows the same benchmarking exercise but now based on the aggregate developing countries’ indicators built based on the QoG database using cluster analysis. Roughly speaking, the same proximity among the comparator countries, except for Rwanda and Burundi, is observed and most country profiles exhibit the same regularity, with scores comparable across institutional domains, although somewhat less so than with the WGI indicators. The salient features are: (i) the superiority of Rwanda in all domains except ‘civil society and voice’, and an impressive advantage in ‘competitiveness’; and (ii) the inferior performance of Burundi, in all areas but competitiveness and democracy and accountability, where it is similar to all other countries. Mozambique’s chart also departs from the mean shape in the ‘competitiveness’ dimension.

A comparison is now made between Tanzania and countries which, although at a roughly comparable level of GDP per capita in the late 1980s, grew so much faster since then that they have reached an income level double or more that of Tanzania, on average. These are essentially Asian countries: Bangladesh, Cambodia, Lao People’s Democratic Republic, and Vietnam. The comparison is made using both 2019 and 2005 WGI indicators, with 2005 being the year when the income gap was roughly half what it is today (see Figure 2.3). Unfortunately, it is not possible to go back much before 2005 because indicators lack precision, due to the fact that fewer observations are available.

Figure 2.3 Benchmarking of Tanzania with respect to fast-growing Asian countries: WGI indicators, 2005 (top figure) and 2018 (bottom figure)

Two lessons may be drawn from this new benchmarking. First, back in 2005, Tanzania’s institutions did not seem to be worse than those of Bangladesh, Cambodia, and Lao People’s Democratic Republic. On the contrary, Tanzania was surpassing these three countries in almost all institutional domains. It also compared well with Vietnam, except for ‘voice and accountability’, where Tanzania prevailed, and for ‘political stability’, where the situation was the opposite. The view that faster-growing developing countries are endowed with institutions of better quality is thus unwarranted when looking at this particular case. The second lesson stems from the rather strong improvement observed in several dimensions among some of Tanzania’s outperformers. This is clearly the case of Bangladesh, Cambodia, and Lao People’s Democratic Republic, and to a lesser extent Vietnam, but not of Tanzania. It is interesting that Tanzania’s development outperformers saw the quality of their institutions improving while growing substantially faster. An obvious hypothesis in the case of Tanzania would thus be that the lack of progress on the institutional front may have delayed progress on economic development.

Other examples taken from the IDP case studies could further illustrate the use that can be made from the comparison of institutional indicators over time and across countries. For instance, a striking example of worsening institutions is Mozambique, whose WGI indicators scored close to the average of the sample countries at a comparable level of income per capita in 2005, and then drastically worsened in practically all domains after 2010 (see Figure 2.4).

Figure 2.4 The worsening of institutions/governance in Mozambique, 2005–2019

This example confirms an important fact to be taken into account when establishing an institutional diagnostic: the quality of institutions, as gauged by institutional indicators like the WGI or the QoG-based indicators, may vary considerably over time, most often following political changes. In other words, it would be wrong to consider that the institutional framework of an economy or, more exactly, the way it is used, has some degree of permanence. Observing an institutional weakness at a point of time may thus result from a real flaw in the institutional framework being temporarily ignored by the power in place. In other words, the law may be flawed, or it may be temporarily disobeyed. The distinction is clearly important.

Overall, aggregate institutional or governance indicators like the WGI indicators, or those indicators obtained by aggregating in a different way those individual, more focused, indicators available in the QoG database, are useful instruments for starting an institutional diagnostic. It is true that the aggregation procedure introduces some imprecision into the description of the quality of institutions, but it is not clear that one would get a better idea of this by considering the numerous and highly diverse individual indicators available, especially because their precision and mutual consistency is often uncertain. The above-noted congruence between the two sets of aggregate indicators is thus reassuring.

Several lessons can be drawn from the few aforementioned examples shown above. They can be summarised in the following way:

1. a. Not much is to be learned from the absolute level of aggregate institutional indicators when concentrating on low- and lower-middle-income countries. For these countries, scores tend to be low, thus reflecting the causal relationship running from the level of development to the quality of institutions.

2. b. When considering a single country, the possible asymmetry between scores in various institutional domains is of special interest since it suggests directions for further scrutiny of the functioning of institutions.

3. c. Benchmarking country A in relation to a group G of other countries requires distinguishing outliers. Comparisons between country A and median countries or against outliers have different meanings.

4. d. The quality of institutions or the use made of indicators may change over time, which points to the need to distinguish between permanent and transitory elements in the diagnostic to be established. Note that this has implications for the analysis of the empirical relationship between institutions and development. If the quality of institutions changes over time, it is difficult to relate it in a causal way to development indicators over a long period.

This section on indicators has relied on aggregate institutional indicators defined for a range of general domains and based on specialised individual indicators – most often based on the opinion of experts. This opinion may differ from the perception that insiders may have of the institutions in their own country, and most importantly on the practical implications of their possible dysfunctions. Surveying their views and identifying the weaknesses they point to is the objective of the second mechanical approach to an institutional diagnostic. The way it was implemented in the IDP is detailed below.

III Asking People: Opinion Surveys and Interviewing Knowledgeable People

Citizens of the country under analysis are insiders; they experience the functioning of national institutions on an everyday basis. If they are not necessarily equipped to compare their country to others, as the experts behind the global institutional indicators, they may be in some instances more knowledgeable, or provide a perspective that is closer to reality. A second important tool in establishing an institutional diagnostic consists therefore of simply asking nationals their opinion on the way institutions work in their country, the most patent institutional weaknesses, and how they think they could be fixed.

There are various ways of proceeding, depending on whose opinion is being collected. A representative sample of the population will mostly reveal how ordinary citizens feel about institutions in their everyday life. Even though their opinions may reveal differences across various types of institutions being considered, it is unlikely that these appraisals will be enunciated in terms of the obstacles to, or enablers of, economic development. Only that part of the population that is used to making decisions that are at the heart of the economic system, deep observers of the society and the economy, or people in positions that require an intimate knowledge of how institutions work, would be able to adopt such a perspective. Especially valuable in this respect should be the views of those personalities who have, or had, major responsibilities, such as government members, legislators, top civil servants, or managers of major firms.

The opinions expressed by these segments of the population must be seen as complementary, because of their different positions with respect to institutions. Eminent persons have the experience of top decision making: they are able to provide a rationale for the reforms they think necessary and/or those they try to implement, as well as past successes or failures. Yet they may not appreciate the nature and the strength of the constraints faced by more ordinary decision makers in running small and medium-sized businesses, or civil society organisations. Finally, these views may miss the way the ordinary citizens perceive institutional constraints.

This section elaborates on the experience gained in the IDP case studies in surveying individual opinions about the institutional context of a country at those three preceding levels. It first summarises the results obtained in IDP case studies from a specific survey that was specially designed for this project and intended for small samples of economic and social decision makers. It then offers a few remarks about the experience of the various country teams in interviewing top decision makers and other eminent persons.

A Using Opinion Polls: the case of the Afrobarometer

Opinion polls are conducted more or less regularly in most countries, including low-income countries. Their goal is to get a picture of: (i) individual well-being – income, health, life satisfaction; (ii) opinions on major current policy and political issues; and (iii) the most common appraisal of the functioning of society, including local communities and national institutions. Polls may be conducted for profit, or they may be implemented by non-profit organisations, like the Afrobarometer in Africa. Given the multidimensional scope of these surveys, however, they comprise few questions on institutions or governance per se.

As a way of experimenting with existing opinion surveys, this section makes use of a harmonised opinion poll run at certain time intervals in a rather large set of African countries – the Afrobarometer, nicely subtitled ‘A pan-African series of national public attitude surveys on democracy, governance, and society’. It is now in its ninth edition, covering the years 2019–2020, but country surveys have not yet been put together in a single database, as was done in the previous rounds. The rest of this section thus uses Wave 8, taken between 2016 and 2018, depending on the country, and covering some 34 countries in the region.

The questionnaire used in the Afrobarometer is common to all countries. It is rather long, since the codebook comprises some 350 variables, among which 80 questions are about the respondent’s evaluation of the country’s governance. They include the degree of democratisation, the efficiency of the government in providing services, the areas which the respondents see as the most problematic, their perception of corruption and their trust in the main actors in society (president, government, parliament, military, courts etc.).

Since it would have been too cumbersome to deal with all of these questions one by one, the same methodology as the one described above to aggregate individual indicators has been followed. Namely, five areas were predefined, closely mimicking the WGI and QoG synthetic indicators in the preceding section. Average question scores for each area were then summarised by their principal component. This yielded an aggregate indicator with a mean of zero and a unit standard deviation across the thirty-four countries present in the eighth wave of the Afrobarometer. Because of non-response in those categories that comprised a relatively small number of original questions, the category meant to represent ‘regulatory quality’ had to be dropped.

Figure 2.5 shows the results obtained with this procedure for a few countries. As before, attention to each country’s institutional profile should focus on two features: (i) the shape of the radar line (i.e., whether it is regular, implying comparable scores on its different branches, or asymmetric); and (ii) how it compares to other countries, keeping in mind that the zero line stands for the mean across all African countries – with no implication whatsoever regarding how African countries perform in comparison to other regions.

Figure 2.5 Comparison of selected African countries, based on aggregated indicators elaborated on the basis of the Afrobarometer: Round 8, 2016–2018

Only two countries in the small sample represented in Figure 2.5 exhibit a regular pattern, meaning that there is no specific institutional domain with a noticeable weakness. These two countries are Senegal and Tanzania, although the rule of law indicator is particularly strong in Tanzania relative to other indicators, and relative to all African countries in the sample. All other countries show a bias in at least one or two domains. Thus, it is not surprising to see that ‘political stability’ and ‘absence of conflict and violence’ are weak in Kenya (remember the post-election killings in 2017) and Mozambique (on account of the resurgence of the Frelimo–Renamo conflict). This feature may not necessarily be considered as an institutional weakness per se, but it is a strong determinant of the context in which institutions function. More interesting from a diagnostic point of view is Ghana’s comparatively weak score on the front of corruption control, which contrasts with quite good scores along all other institutional dimensions. Benin also fares rather badly on the corruption axis, but also on government effectiveness, whereas Uganda fails on corruption and political stability. Finally, Malawi fails in regard to the opinion of the population on both government effectiveness in delivering services and democracy, that is, voice and accountability. If a diagnostic were to be conducted in these last four countries, the Afrobarometer would thus suggest clear directions of investigation.

Despite differences among them, it can be noted that the countries appearing in Figure 2.5 tend to do better than the average African country, since few scores are below zero (which is the mean for the whole sample of countries in every dimension). Equally noticeable is the similarity between the relative scores of countries in Figure 2.5 and comparisons made earlier using the WGI indicators or the indicators constructed from the QoG database. For instance, Tanzania tends to dominate its neighbour countries, as was roughly the case in Figures 2.1 and 2.2, when excluding Rwanda.¹² The similarity is not perfect, however. The ‘voice and accountability’ score appears to be low in Figure 2.1, based on WGI 2019, whereas it is relatively strong in Figure 2.5, based on the Afrobarometer. Interestingly, this difference likely reflects objective changes that took place between 2017 (the year the Afrobarometer survey was undertaken) and 2019, in regard to the freedom of the press and other media in Tanzania.¹³ Overall, it is interesting that a public opinion survey like the Afrobarometer delivers a message about institutional strengths and weaknesses in Africa that is similar to aggregate expert-based indicators.

Even though the discrepancy in the ‘voice and accountability’ score may have an objective explanation relating to changes in the control of the media by the executive in Tanzania, it raises several issues. First, it is another example of the kind of noticeable change that may take place during a short time span in the evaluation of institutional quality. Second, it may suggest that public opinion is more volatile than that of experts, or that the various factors that should be taken into account in the evaluation of the strength of (democratic) institutions covered under the heading ‘voice and accountability’ are not given the same weight by citizens and experts. Third, and more fundamentally, the questions asked in opinion surveys like the Afrobarometer bear upon limited aspects of institutions.

Public opinion surveys provide information on individual attitudes and perceptions that seem far away from the institution–development relationship but may nevertheless be of some indirect importance for development. Questions about people’s views on basic principles like democracy or justice, about their own moral values, about their degree of trust not only in formal institutions, which are accounted for in the above indicators, but also in relatives and neighbours, surely matter for the way a society – and therefore its economy – functions. Because they were not directly related to the way institutions work, they have not been included in the set of questions used to build the indicators analysed above. Of course, this should not prevent us from considering some of them, especially trust in others, if they appear relevant for a deeper exploration of the way specific sectors of a country’s economy work.¹⁴

B The Country Institutional Surveys (CIS)

Overcoming the limitations of opinion surveys in dealing with such specific issues as the role of national institutions in economic development requires two things: (i) restricting the sample to people with some direct experience in dealing with institutions, or with good knowledge about the way they work; and (ii) orienting the questionnaire towards the institution–development relationship while substantially broadening it to cover the full range of relevant institutional dimensions. The IDP has developed such a surveying tool, whose characteristics are now described, before showing the use that can be made of it.

1 The Structure of the CIS

The CIS implemented in the four IDP case studies is inspired by the Institutional Profile Database (IPD), an expert survey conducted jointly by the economic agencies of the French Embassies, the Centre for Prospective Studies and International Information, and the University of Maastricht (Bertho, 2013). Its questionnaire was taken as a basis for the CIS because of its rather remarkable degree of exhaustiveness. As adapted to the IDP project, the CIS questionnaire comprises some 320 questions, covering a broad range of institutional characteristics, structured into nine domains or areas:

1. 1. political institutions;

2. 2. security, law and order, and control of violence;

3. 3. the functioning of public administrations;

4. 4. the free operation of markets (ease of doing business, dealing with land rights);

5. 5. coordination of stakeholders, strategic vision, and innovation;

6. 6. the security of transactions and contracts;

7. 7. market regulation, social dialogue;

8. 8. relations with the rest of the world; and

9. 9. social cohesion, social protection, and solidarity.

Not surprisingly, this list of institutional areas is roughly consistent with the aggregate indicators used in the preceding section to describe the quality of institutions in a country and to make comparisons across countries, though it is slightly more detailed.

A questionnaire with so many questions is clearly impractical if applied to a sample of people who are busy with their own occupations, instead of the experts or observers surveyed in the IPD. Moreover, it is not clear that respondents would have the knowledge that would allow them to cover all the domains set out above. Two solutions were therefore implemented. Both meant a severe reduction in the number of questions – though one more than the other.

The first solution consisted of asking respondents to pinpoint three of the aforementioned areas that they would consider as the most constraining for development, and then to answer the corresponding questions. To make sure all domains were approached, however, respondents were also asked to answer questions in a fourth randomly chosen area. This solution thus yields two sets of information: (i) some ranking of institutional areas depending on the constraints they impose on development; and (ii) in each area, features that were seen as strengths or weaknesses. Overall, the total number of questions turned out to be similar to the original IPD questionnaire, even though many questions were adapted to make them as relevant as possible to the context of the country under analysis, and a few questions were added on country-specific topics. Given the choice of priority areas, the actual number of questions answered by CIS respondents was roughly a third of the total: that is, slightly more than 100 questions. Note that, given the specific structure of the questionnaire, the same question could be relevant under different institutional headings and thus could appear more than once in the full questionnaire. However, as the survey was implemented on tablets, it was possible to code the questionnaire in such a way that a respondent would not have to answer the same question several times.

The second solution was to ask respondents to answer all questions but to simultaneously and drastically reduce the number of questions in the original IPD questionnaire, while making them more consistent with the economic, social, and institutional reality of the country studied, and while maintaining the exhaustiveness of the areas covered and having respondents answer all questions. This choice did not prevent from proceeding with the initial ranking of institutional areas by perceived severity of the constraints they impose on development. It reduced the detail with which institutional areas were described but added to the representativeness by allowing all respondents to give their opinion on all institutional domains.

The first format of the CIS was implemented in Tanzania, Benin, and Bangladesh, whereas the second one was used in Mozambique. In all cases, variations could be introduced in the list of general institutional areas, depending on the specificity of the country. For instance, decentralisation was considered to be worth singling out in Mozambique, whereas ‘political institutions’ were split into features referring to the way the executive operates and features describing the functioning of the overall political system in the Bangladesh questionnaire. These variants were generally inspired by the intimate knowledge of the country held by the authors of the diagnostic, or by discussions with key informants within the country, as will be seen below.

In all questionnaires, answers to questions were formatted so that answers could fit a five-point Likert scale ranging from ‘not at all’ to ‘very much’, with ‘no answer’ as an additional option. In aggregating questions together, however, care was taken regarding whether the question being asked was formulated in a positive or a negative way. A high score on the Likert scale would then be taken as favourable in the former case but unfavourable in the latter.

Table A2.1 in the appendix to this chapter, taken from the Bangladesh case study, shows the structure of the questionnaire used in that instance. The complete questionnaires for all case studies are accessible on the Internet.¹⁵

A last important point to stress about the questionnaires is that answers are necessarily influenced by the current political, social, and economic context. It so happened that the CIS in Bangladesh was conducted at the time of the general elections, so that answers to some questions may have been biased by the arguments exchanged during the electoral campaign. An appropriate discounting of the significance of these answers is thus needed. This being said, most questions in the questionnaire refer to institutional features that are quite persistent. The same situation was found in Tanzania, as the survey was undertaken less than a year after a new president came into power with a rather ambitious anti-corruption programme. Respondents were thus asked to answer the questionnaire in the light of their experience over the last ten years, rather than on the basis of the last few months and the electoral platform of the new president. Still, when they had completed the questionnaire, they were then asked how their answers to the questionnaire would possibly be modified if they were to take into consideration the last few months since the presidential inauguration.

As should be clear by now, the CIS is not directed to the whole population but only to those people who are most likely to confront the institutional challenges of a country on a regular basis, either through their occupation or through observation from a particularly informative position. Stratified samples were used, with strata defined by occupation, sector, and high-level positions in several types of organisations. Typically, CIS samples comprised politicians from the ruling party and the opposition, top bureaucrats in ministries and public agencies, business executives in the main sectors of activity, academics, journalists, representatives of civil society, foreign diplomats, and heads of local offices of international organisations. To the extent possible, these strata, of different sizes, were combined with gender and geographic criteria.

The size of the sample differed across surveys. It was slightly more than 100 people in Tanzania and Mozambique, but triple that in Benin and Bangladesh. It is of course always better to deal with a bigger sample. However, because the CIS is meant to reveal the views that decision-making people may have on institutions, rather than to test the significance of such and such an answer to a specific question, sample size should matter mostly in order to make sure that the range of decision-making people who might have different views about institutions is fully covered.

2 Short Overview of Results and Lessons from the CIS in the IDP Case Studies

As the CIS was adapted to the context of the countries in which it was implemented, different definitions of the main institutional areas around which the questionnaire was built were used, while some items were added to or subtracted from the list of the nine areas mentioned above. The customising of the questionnaire also required inserting new questions, deleting others, and framing the remaining questions so that they fit the local context.

Regarding the institutional areas, experience shows, first of all, that for their ranking to deliver information it is important not to have too many or too few of them. In the former case, respondents may find it difficult to differentiate among all the alternatives. In the latter, they will tend to attach the same importance to most of them. The second lesson from experience is that it is important to provide respondents with a general description of the institutional areas they will have to rank, and of the questions they will be asked to answer. However, too general and generic wording may be insufficiently clear. For instance, ‘security of transactions and contracts’ may not be well understood if it is not specified that it refers to institutions that are supposed to guarantee contract compliance, especially debt contracts, to insolvency laws, to litigation procedures, and to business laws and courts. Likewise, it should be made clear that ‘political institutions’ without further precision should include not only constitutional matters but also how basic principles are obeyed, political life in general, the control exerted by the executive over political, economic, and civil society actors, electoral procedures, and checks and balances on the government. Incidentally, as this long list attests, this area had to be split into several sub-areas in some surveys.

Then comes the issue of how to articulate the ranking of institutional areas by respondents and their answers to the large number of questions in those areas, and possibly others. These questions are supposed to provide more detail on the reasons why an area is harmful to development. There are two ways of handling them. One way consists of simply ranking them according to the Likert scale and to examine the kind of institutional challenge the most unfavourable answers point towards. The other way consists of dealing with clusters of questions that may be considered as detailed institutional sub-areas – as shown in Table A2.1 – and checking what sub-area exhibits the lowest average Likert scale, bearing in mind the distinction between positive and negative question formulations. The first approach offers the advantage of focusing on extreme weaknesses, whereas the second reveals those sub-areas with a high frequency of mediocre scores.

One way or another, it is interesting that, despite the fact that the respondents to the CIS answered questions in the institutional areas selected by them as the most detrimental to development, the areas revealed by their answers to individual questions do not always fit their initial ranking. This was particularly the case in Bangladesh, where there was very little difference among areas in the initial ranking, whereas answers to questions very clearly singled out as strongly negative ‘land rights’, ‘civil service’, and ‘political institutions’ (in relation to the functioning of the executive). Likewise, in Mozambique, the lack of a ‘common vision of the national strategy’ and the ‘management of public administration’ appeared among the most detrimental areas, while answers to single questions suggested that ‘legal and constitutional matters’ and ‘political participation’ were the sub-areas where the lowest Likert scores were the most frequently observed. This seems to suggest that general institutional areas may indeed be too general to fully ground a diagnostic exercise because they comprise different dimensions, which may be appraised in different ways by the respondents. In other words, a general institutional area may be seen as mildly constraining for development despite some of its sub-areas being of the lowest quality.

Table 2.1 summarises in a synthetic way the information conveyed by the CIS in the four case studies of the IDP project. As far as the general institutional areas are concerned, whether their ranking was made a priori by survey respondents or based on the questions with low scores, it is not surprising to see so much commonality across countries since the options were similar and of a limited number. Yet there are interesting differences. Institutions that affect the business environment are mentioned in Benin and Tanzania, but not in Bangladesh and Mozambique. Land issues never appear in the a priori ranking, but they are present in the question-based rankings in three countries. This divergence may be taken to mean that the handling of land rights is quite bad but that their influence on development is not considered to be of great importance. On the other hand, it is striking, but certainly not unexpected, that the low quality of the public administration, or civil service, appears everywhere as a crucial institutional challenge, the same being true of the way political institutions work. (Remember, however, that this area may in some cases be too broadly defined.)

Table 2.1 Synthetic summary of institutional weaknesses and strengths revealed by the CIS in the case studies

	Bangladesh	Benin	Mozambiquea	Tanzania
Three worst general institutional areas (a priori ranking)	Political institutions: system	Political institutions	Lack of vision	Public institutions
	Justice and regulation	Public administration	Public administration	Public administration
	Political institutions: executive	Unfriendly business environment	Management of macro and sectoral policies	Unfriendly business environment
Three worst general institutional areas (based on all questions)a	Management of land issues	Public administration	Legal and constitutional matters	Public administration
	Civil service	Management of land issues	Limited freedom of information and political participation	Management of land issues
	Political institutions: executive	Unfriendly business environment	Constraints imposed by donors	Unfriendly business environment
Weak institutional sub-areas	Weak checks and balances on executive	Generalised corruption	Corruption	Management of land issues
	Excessive centralisation	Opaqueness of executive	Constitutional breaches	Generalised corruption
	Limited freedom of information	Imperfect knowledge of the law	Elite capture	Opaqueness of executive
	Corrupt elections	Mismanagement of SOEs	Public goods delivery	Weak regulation
	Nepotism	Lack of vision	Land conflicts	Pressure of lobbies
	Public services		Aid dependence	Excessive centralisation
	Land conflicts		Separation of powers
Strong institutional sub-areas	Capacity for informal secure deals with government	Civil liberties	Availability of foreign aid	Civil liberties
	Role of donors	State free from traditional norms	Civil liberties	Security
	Rigorous macro policies	Role of donors	State free from traditional norms	Sense of national identity
	Sense of national identity	Anti-corruption efforts	Sense of national identity

^a The question-based ranking of general areas in Mozambique bears upon a slightly different set of areas than the a priori ranking.

Comparison with the aggregate governance indicators analysed earlier is not easy because congruence in the definition of institutional areas is limited. Yet it is interesting to note that poor management of the public administration emphasised in the CIS conforms well with the relatively low score of ‘government effectiveness’ in Figures 2.1–2.3 for the four case studies.

More detailed information is revealed when scrutinising institutional sub-areas, some of which are reported in Table 2.1 according to their especially low or high Likert score. Many of these institutional weaknesses, or strengths, are analysed in depth in the case studies. Yet even at this aggregate level some interesting features appear. It is indeed at that level of the CIS that corruption is unanimously mentioned, providing another source of consistency between survey results and aggregate institutional indicators. However, the survey gives more detail about where corruption practices are the most salient – that is, at the political level, between business and the executive, and between the population and the state bureaucracy. Imperfect knowledge of the law, mismanagement of state-owned enterprises (SOEs), elite capture phenomena and weak regulation of big business, aid dependence, and excessive power centralisation, all uncover precise institutional weaknesses or their consequences, and provide useful directions for further inquiry.

Some of the strengths emerging in Table 2.1 are instructive, too. That the capacity to strike secure informal deals with the executive is found to be a strength in Bangladesh unveils an important characteristic of the institutional framework in that country. That meritocracy – in effect, the recognition of academic credentials in the bureaucracy – is mentioned among the favourable institutional features in Tanzania is also worth stressing, for this feature coexists with some signs of elite capture and generalised corruption. In both cases, the survey respondents demonstrate a rather flexible conceptualisation of institutions.

The role of donors is stressed at different stages of the survey and arouses ambivalent reactions by the respondents. They generally agree that this is an important aspect of the economic management of their country. In some cases, they emphasise the positive effect of development assistance on national budgets, or the usefulness of advice provided by donors. In others, they see aid dependence as severely compromising the long-run development of the country, and donors as constraining policy options too much. This two-sided role has long been underscored in the aid literature, but it is interesting that it is also very present in the minds of decision makers in recipient countries.

Specific questions in the CIS, in contrast to whole areas or sub-areas, may also deliver information that could be relevant in a further stage of the institutional diagnostic procedure. In one country, they may concern public procurement or the reliability of public statistics; in another, they touch upon the presence of discrimination, or the lack thereof; and, in still another, the issue is the unequal geographical coverage of public services.

Another valuable advantage of the CIS is its capacity to differentiate answers by the characteristics of the respondents. Of special interest are differences according to occupation, and especially between business executives and others, in view of the crucial role of the business sector in the development process. In Bangladesh, which is the only case study that systematically exploits that dimension of the survey, it is remarkable that business executives are more severe than politicians, bureaucrats, and academics with respect to the judicial system and the public administration.

To conclude this short synthesis of the CIS expert opinion survey undertaken in the four IDP case studies, it is fair to say that this instrument discriminates better among institutions than the aggregate institutional/governance indicators discussed in the first part of this chapter, and is considerably more instructive than general opinion surveys like the Afrobarometer. This is basically because of its stronger focus on institutions, its more complete inquiry into how well or badly they work, the fact that its set of respondents have some real experience and expertise in local institutions, and the explicit request that they evaluate institutions in regard to how they affect economic development. Despite these advantages, however, the CIS survey must still be seen as a mechanical exploratory tool that suggests areas or sub-areas where institutions do not function well and may be detrimental to development, as well as possibly sub-areas where the opposite may hold. Yet nothing is revealed about what may explain such situations, nor about the channels through which dysfunctional institutions may impinge on, or benefit, development. Executive decisions may be judged ‘excessively centralised’ or ‘opaque’, land disputes may be found to be too frequent, or the business elite too powerful, but what are we to infer from these statements that might point to appropriate reforms? It will be the task of the analysts to figure this out at a later stage of the diagnostic.

The format of the CIS evolved over time, from its first edition in Tanzania to the last ones in Bangladesh and Mozambique. In the latter case, the research team opted for a shorter questionnaire common to all respondents and adopted a slightly broader range of institutional areas than in the other countries. Along the initial lines of a longer and, to some extent, customised questionnaire, the Bangladesh survey appears as the most accomplished one, partly because it was able to integrate the experience acquired in the previous editions. The questionnaire was more systematically organised, not only in the main institutional areas but also, within an area, in sub-areas, or ‘clusters’, and even sub-clusters. This seems to have enhanced the legibility of the questionnaire and eased its statistical treatment. The questionnaire for Bangladesh should therefore be used as a template for a further edition of the survey, if any, unless there are reasons to prefer a shorter common questionnaire.

C Asking Key Informants

The last group of people to be approached for their personal insights into the role of the nature and quality of institutions in their country are those persons who exercise significant responsibilities as politicians in power or in active opposition, top bureaucrats, high-level academics, and personalities of the civil society. Numerous such key informants were interviewed as part of the initial exploratory phase of every IDP case study. In this essentially methodological chapter, the point is to summarise what was learned from them about each country, as this is fully reported and then developed in each case study. The main purpose of this section is to reflect on the way these interviews were conducted and, possibly, on some common features in the opinions expressed by the key informants across countries.

The identity of the key informants varied across case studies, but the choice was made at the outset not to interview high-powered members of the current executive – that is, presidents, vice-presidents, or prime ministers. This choice reflects not so much the difficulty of approaching them, but the concern that their opinion would necessarily be biased, partisan, or too much influenced by current challenges. In this category of informants, interviewees were most often personalities who were in this kind of position in the past and could thus have developed deeper insights into institutional constraints on development-oriented action when they were in charge, as well as today.

Different formats were used to gather the opinions of these particularly knowledgeable persons: seminars, open-ended conversations, or a predetermined set of questions. With retrospect, the latter formula proved the most effective. Yet it requires already having some good intuitions regarding the most relevant issues, so as to avoid losing time on commonalities. From this point of view, the Mozambique experience, with a set of well-chosen questions in each interview, delivered particularly interesting indications.

Several common problems were noticeable among these interviews, which often limited what could be learned from the interlocutors. The first one is that the very concepts of ‘institution’ and of their role in development were uneasy to convey to the interlocutors. For instance, the view that corruption per se is only the symptom of ill-functioning institutions, the issue being not only the detection and then the punishment of corruption but the circumstances that create the possibility of extracting rents, was not always uniformly shared. Respondents were often satisfied to cite corruption as the main source of problems in the way their country operates and is governed, rather than identifying its deep causes and, possibly, how it could be remedied. A second related common problem was the tendency of key informants to rely on an ideal normative framework without much relevance to the analysis of institutional problems in their country and solutions to them. For instance, the view was often expressed that the reason something does not work well is because it departs from mostly theoretical norms, like a full-fledged democracy with perfect transparency, effective checks and balances on the executive, and complete separation of the executive, the legislative, and the judiciary. Using such a norm as a reference to think about reforms is fine but illusory, since it misses key political economy constraints that precisely explain the persistence of weak institutions and their consequences. The difficulty is that political or political economy issues are still too sensitive for people who have been closely involved in them, whereas opposition members are generally biased, and people who are not directly involved in politics do not always realise the nature of these constraints. A third difficulty experienced during the interviews was the tendency for the conversation to focus exclusively on current public concerns or concerns which left their mark on the minds of interviewees, rather than on what they thought may be key persistent institutional weaknesses in their country.

Another interesting observation that results from these interviews is the similarity across interlocutors and across countries regarding the institutional fields cited as possible sources of hindrance to the process of development. Beyond corruption, practically all informants touched upon the de facto functioning of the political system and the judiciary, and the excessive centralisation of power and public decision making. Yet the link was not always drawn with the pace and structure of economic development. Closer to the issue of development, issues like a dysfunctional civil service, limited state capacity, the lack of coordination between public entities, or the management of land issues, were also almost unanimously cited. If the symptoms are clear and were widely shared, however, their causes were rarely discussed and the remedies proposed were not always realistic, often boiling down to wishful thinking: for example, ‘eradicate corruption’, ‘decentralise decisions’, and ‘have parliament play its role’.

Being what it is, the whole exercise is nevertheless of utmost interest, not only because it allows us to establish a kind of ranking of the most serious symptoms of institutional weaknesses as seen by informed players, and to sometimes have a glimpse into the political-economic factors behind them, but also, and most importantly, because these weaknesses were usually depicted and discussed in a particular sectoral context, be it a specific ministry, local government, the education sector, tax collection, or banking regulation. At a later stage, this observation of institutional dysfunctions within specific economic or social contexts is what may allow for a better understanding of their causes and possible remedies. In that sense, the direct contact with present or past high-level decision makers or observers sheds a different light on institutions than the general description of the quality of institutions and governance that is obtained from the experts consulted in the construction of international indicators, from opinion surveys, and even from the lower-tier decision makers polled in the CIS.

IV Resorting to History and Economics

After consulting the opinions of others on their perceptions of the institutional obstacles to, or constraints on, the pace or the sustainability of development, the issue must then be approached from the point of view of economics, and in a more inductive way. The objective of this new stage of the diagnostic methodology is to identify the economic development challenges faced by a country, in order at a later stage to investigate whether and how they may be related to institutional weaknesses. Preparing the ground for this exercise involves more than analysing the current economic situation of a country, as well as its assets and liabilities for future growth. Because development is an evolutionary process, and because present economic challenges most often have some of their roots in the economic, social, and political past of a country, ascertaining their nature and their origin also requires a careful review of the country’s political and economic history.

The point is not to propose a methodology for such a review. On the political and social side, we can rely on the existing literature about the country concerned. On the economic side, if available in the recent literature, we may make use of economic diagnostics highlighting the constraints that bear on the acceleration, the sustainability, and the inclusiveness of economic growth. It is not clear, however, that such a diagnostic satisfactorily incorporates all of the roots of these constraints in the past or in recent history. If this is not the case, such a deeper economic diagnostic will have to be established.

Growth diagnostic exercises along the lines of Hausmann, Rodrik, and Velasco (2005) are a useful reference when they are available for the country studied, if they are not outdated. Based on a rather standard model of economic growth, this diagnostic methodology consists of identifying those constraints on economic growth which are likely to be the most binding in the pursuit of faster economic growth.

The idea is simple. Within a standard neo-classical framework, the determinants of growth are the level of investment, the productivity of these investments, and possibly other sources of productivity gains, like a more educated labour force or the adoption of better techniques, or organisation, of production, although the Hausmann, Rodrik, and Velasco approach concentrates on the first two factors. In turn, each of these factors may be hindered by various limitations. Investment may be too low because returns are insufficient or because the cost to finance them is too high. Returns may be low for physical reasons, like the geographical context, lack of human capital, or bad infrastructure, but also due to low appropriability, like excessive taxation, unsecure property rights, macroeconomic volatility, or simply a lack of information on technology or markets. On the other hand, access to finance may be limited because of poor savings, weak financial intermediation, and the unavailability of foreign financing. All of these possibilities form a kind of tree, the top of which is the rate of growth of the economy, with the branches being the immediate determinants of growth, the sub-branches being the channels through which these determinants may fail, and the bottom of the tree being all the potential constraints just listed. The growth diagnostic approach then consists of finding some quantitative measure of the strength of these constraints and looking at those that depart most from some norm. For instance, a higher return to schooling relative to other countries would suggest that human capital is scarce and thus binds economic growth. Likewise, a disproportionately high borrowing rate of interest reveals either insufficient savings or dysfunctional financial intermediation, the same being true of a large gap between the marginal product of capital and borrowing rates. Another example illustrates a deficiency at the level of infrastructure: firms have sometimes to rely on their own generators to palliate frequent electricity outages across the grid, which increases the price they pay for energy. The gap between this price and the posted price of electricity is a measure of how constraining the supply of energy is for firms. Comparing the level of these ‘shadow prices’ of the various potential constraints on growth with the levels observed in benchmark countries, it is then possible to establish a list of the most binding constraints.

The actual implementation of the growth diagnostic framework goes beyond a few indicators of the type just mentioned. This can be seen, for instance, in the kind of user manual provided by Hausmann et al. (2008) and others. If it is a useful instrument, it has limitations, and practical applications do not always reveal more than what mere intuition would suggest. Among these limitations, one may cite its quasi-exclusive focus on private investment, the inherent difficulty of detecting price or non-price signals, the extreme reliance on inter-country comparisons without clear criteria to select benchmark countries, and the lack of attention to the interaction across constraints and their time dimension (i.e., which one should be handled first).¹⁶

Another limitation of the standard growth diagnostic approach is its aggregate approach to economic growth and the too-little attention that is given to the structural aspects of development, and especially the structural transformation of the economy that causes and is caused by development, along the lines of the well-known analyses by Kuznets (1955) and Lewis (1954), and the key distinction between formal and informal sectors. This aspect of development is particularly important when dealing with low-income or lower-middle-income countries.

Judging from a few recent applications in the countries covered by the IDP case studies, the conclusions from growth diagnostic exercises are not always very enlightening, even though they are relevant. In the case of Tanzania, for instance, such a diagnostic undertaken under the auspices of the United States Agency for International Development (USAID) in 2010¹⁷ pointed to the following major binding gaps: infrastructure, appropriability of returns (due to unsecure land rights for investors), technical skills, and small and medium-sized enterprises’ (SMEs’) access to finance (including agriculture). A similar study by the Organisation for Economic Co-operation and Development (OECD, 2013) added the weak regulation of business and trade to this list. With a little hindsight, it cannot be said that the lack of infrastructure, of human capital, and of financial resources for small firms were unexpected constraints on growth. In effect, they are common to most low-income or lower-middle-income countries. The issue of land rights may be more specific, and therefore may warrant further investigation. The same would apply to regulation, if the authors had something else in mind than the way the administration deals with the private business sector.

Equally disappointing is the executive summary of an ‘Inclusive Growth Diagnostic of Bangladesh – again under the auspices of USAID (Davidson et al., 2014) – which points to electricity and governance as the most binding constraints on faster economic growth at the aggregate level, and, again, to energy and human capital as the most binding constraints on the growth of the garment sector – although education is explicitly mentioned as not binding at the level of the whole economy.

Of course, there is much more than these general conclusions in the two reports just mentioned, especially in the Bangladesh ‘inclusive’ growth diagnostic, with its focus on specific economic sectors and social issues like women’s entrepreneurship. The main point is simply that the approach is too mechanical and too static to take into consideration the past structural evolution of the economy, which may have left heavy sequels in the current working of the economy. Also, it does not anticipate future constraints for which remedies should probably be put in place today. Moreover, it is largely based on information drawn from enterprise surveys, which tend to over-emphasise the practical aspects of business, rather than deeper constraints, and to underplay the macroeconomic aspects of development, despite their utmost importance.

To the extent that growth diagnostics are available, they should be used and updated. If none is available, then a similar approach has to be developed. In both cases, however, it is essential to give more depth to the analysis by incorporating it within a reflection on the long-run evolution of the economy, its potential growth engines, and its current and future likely challenges. Political history and the current state of the political game or the structure of political power are other essential factors that will need to be considered at a later stage of the diagnostic when the causes for the persistence of weak institutional equilibria and the political economy of reforms will be the focus of attention. The overall review of the economy and its political context must rely on the existing literature about the country, but also, where necessary, on original work by the diagnostic team. It is also important that the review covers all aspects of the economy, at macro, meso, and sectoral levels, and that it looks at societal issues too, insofar as institutional weaknesses may be more or less visible, depending on the perspective one adopts with regard to the economy.

The Bangladesh case study provides a good example of the need to go deeper into the analysis of the economy than what a simple-minded growth diagnostic approach does, and to combine it with a review of economic and political history. Bangladesh has grown at a rather rapid rate over the last twenty years or so, very much – but not exclusively – thanks to ready-made garment (RMG) exports. Doubtless, a growth diagnostic exercise would make it possible to identify constraints to be relaxed in order to accelerate growth under this RMG-dominated growth regime. However, an in-depth review of the economic evolution of the country suggested that the long-run sustainability of growth requires a diversification of exports beyond the garment industry. This is unlikely to result from private initiative and would call for an adequate sector-based public policy, such as existed in the past when the textile sector was seen as worth of priority efforts by the government. Lessons can be drawn from this past experience, including not only the policy instruments which were used but also the whole decisions process – by which we mean in particular the relationship between entrepreneurs and the state that allowed for the implementation of the policy drive. The same type of diversification issues arose in the review of the Tanzanian economy.

V Preparing for Thematic Studies

At this stage of the diagnostic process, it can be considered that most of the more easily accessible materials needed to begin an institutional diagnostic have been gathered. To recap, these are the following:

1. a. Institution/governance indicators: Which institutional areas appear weaker than the others? How does the country being studied differ from benchmark countries (i.e., neighbour or faster-developing countries)?

2. b. Which institutions are perceived as weak or most constraining for development by: the whole population, people who are most exposed to the working of institutions, including business managers, politicians, and the civil society, or observers of the way institutions function, and, finally, top decision makers, including past members of the executive, high-level politicians, top bureaucrats, and big business and civil society leaders?

3. c. A review of the political and economic history of the country, with an emphasis on current and future economic challenges for the acceleration or the sustainability and inclusiveness of growth.

Based on this set of information or evidence, an attempt can be made to bring them together by asking, for instance, whether the identified development challenges relate to specific features revealed by institutional aggregators, or to patterns in the perceptions of people, including experts, about the way institutions hamper faster or more inclusive growth. Digging deeper than simple associations to uncover some logical relationship between these various pieces of evidence might be difficult, however. Some clues will be available in certain cases, particularly when some key informants and analysts concur in the identification of the development challenges confronting their country and provide converging institutional explanations. In most cases, however, further scrutiny will be needed to make the link between development challenges and institutional weaknesses.

The experience accumulated on the occasion of the case studies suggests that this essential step in the diagnostic calls for a more detailed approach than when reviewing general economic development challenges. Some economic challenges will still be too general to be directly related to certain institutional areas, like the rule of law or the quality of regulation in the WGI aggregate indicators, to the problem of corruption in opinion surveys, or to weak state capacity in interviews with key informants. Even in those cases where there apparently is more proximity, such as, for instance, when a dysfunctional public administration is shown to truly exert a major drag on development, a more detailed analysis is needed to determine what makes it dysfunctional. Is it the lack of skill of civil servants, their rent-seeking behaviour, the overlapping of responsibilities, or inefficient management? And then, in every case, what prevents the relevant authority from taking action to remedy those flaws?

Answering these types of questions, as well as addressing the institutional factors behind major economic development challenges, requires getting into more detail about the institutional context in which the economy and the process of economic decision-making works. This cannot be done at the aggregate level – except perhaps when studying possible flaws in macro policymaking – but calls for attention to specific sectors. To take an example, shedding light on the role of institutions behind the absence of firm policies aimed at pushing export diversification in Bangladesh or Tanzania demands a better understanding of the relationship between private business and the state. Likewise, understanding the pervasive infringement of property rights in relation to land, which is found to be a binding constraint in a growth accounting exercise and is stressed by expert opinions, necessitates that we look at the way land allocation issues are resolved through market mechanisms or through the administrative machinery, including the judiciary.

More detailed analysis defines a second step of the institutional diagnostic methodology: thematic studies aimed at identifying the role of the institutional context in precise circumstances or sectors, chosen based on the results of the three preceding mechanical steps of the diagnostic (see the two examples just mentioned – export diversification and land rights). This new stage, which is intended to add information for the final phase of the diagnostic, is no longer mechanical. Thematic studies demand research instruments that are adapted to the area being studied and should be left to specialists in that area. These experts will be able to observe, in situ so to speak, the role of specific institutional features in producing observed results, including, possibly, the political economy factors that block solutions to the detected problems.

The choice of these thematic studies is best left to the authors of the diagnostic, relying on what has been learned in the mechanical steps: that is, the most salient governance indicators, the institutional features most frequently cited by the people and the experts, or the particular areas highlighted by key informants, and, above all, the main development challenges revealed by the review of economic development and policies. Right away, however, some subjects appear hard to avoid. Think, for instance, of the institutional context of the relationship between the business sector, which is essential for economic development, and the state, the main policy actor. Another key thematic area is the functioning of the public administration, possibly in some specific sector of activity like education, taxation, or land management. Likewise, some space must necessarily be devoted to the strategic sectors of the economy, possibly the export sectors.

VI The Final Diagnostic and The ‘Diagnostic Table’

Based both on the general approach to the way institutions may affect development (see the first sections of the present chapter) and on a closer look at how institutions actually affect the functioning of the economy and the political economy of policymaking in certain thematic areas, analysts should then be able to propose a diagnostic of the institutional setup that governs development in the country being studied. Beyond pointing to institutional weaknesses, or possibly strengths, and their implications, they should also be able to speculate on the nature of the reforms to be undertaken and, most importantly, the political economy of these reforms.

More will be said on the methodological framework to be used in this last step of the diagnostic when we summarise the conclusions of the diagnostic performed on the IDP country studies and when we draw broad lessons from the literature dealing with two miracle development experiences of Southeast Asia, South Korea, and Taiwan (see Chapters 3 and 4). Meanwhile, however, it may be useful to indicate the general approach that has been followed in the country case studies, as a way of structuring the elaboration of the final diagnostic.

This approach is summarised in the ‘diagnostic table’; an example, drawn from the Benin study, is shown in Table 2.2. This table tries to relate the basic institutional weaknesses identified in the study of a country as practically ubiquitous in all aspects of the functioning of the economy with general economic consequences, on the one hand, and proximate causes, on the other. These proximate causes, which are amenable to changes through policies and reforms, must themselves be related to ‘deep factors’, which may be responsible precisely for whether those policies and reforms can be undertaken or not.

Table 2.2 The diagnostic table of the Benin case study

Deep factors	Proximate causes	Basic institutional weaknesses	Economic consequences
– Political game (neo-patrimonialism, with multiple oligarchs) – Geography (small country with a big resource-rich, overly protectionist, neighbour) – Multiple ethnic groups and a regional divide – Role of donors	– Policy instability (1): frequent law changes – Policy instability (2): frequent changes in the organisation of key economic sectors (e.g., cotton sector) – Lack of long-term development planning – Elite capture of key state functions – Weakness of state, reflected in its inability to exert control over all its public administration – Existence of rent opportunities in illegal trade with big neighbour	– Widespread corruption (e.g., in business and politics, lack of independence of tax collectors and magistrates) – Weak enforcement (and complexity) of laws – Weak regulation (domination of big business) of key sectors – Lack of state coordination (e.g., fierce competition between ministries) – Low state capacity (e.g., under-staffing of key public administrations, low quality of education) – Low prioritising of critical public goods (e.g., education or power generation) – Opacity of policymaking and economic management; unaccountability of public agencies in key sectors – Pervasive informal practices, magnified by illegal cross-border trade	– Low quality of education – Weak sustainability of the growth pattern: * low productivity growth; * low diversification; and * low level and pace of industrialisation – Poor investment climate – Lopsided spatial development – Increasing inequality and slow reduction of poverty – Chronic aid dependence – Lack of citizens’ trust in key institutions – Vulnerability to external shocks

In examining Table 2.2, it is quite important to realise that there is no one-to-one relationship between the elements of the various columns. One institutional weakness does not have a unique general consequence, and has more than one, unique, proximate cause. The relationship between the four columns is essentially multivariate. The important point is essentially the chain of causality. The whole set of institutional weaknesses is the consequence of the whole set of ‘proximate causes’, which depend themselves on the whole set of ‘deep factors’. At the other end of the chain, the set of institutional weaknesses affects how the economy works. Of course, looking at the whole chain, it can be said that the ‘deep factors’ are the ultimate determinants of economic performance. This would be correct, but not necessarily interesting from a diagnostic point of view. The important element in the whole chain is the proximate causes because they are amenable to changes through policies. This is much less the case for deep factors. Yet they are essential in order to understand why policy reforms are not taking place or why certain policy choices are made and, as such, they are an intimate part of the diagnostic. For instance, if the structure of political power prevents a reform that will help resolve some institutional weakness being undertaken, the only thing the analyst can do is, on the one hand, to identify the winners and losers of the reform and understand the nature of the blockage, and, on the other hand, to take firm notice of it in the diagnostic.

VII Conclusion

In concluding this chapter, it is important to stress the radical difference between our all-encompassing approach and the purely mechanical approach based solely on more or less disaggregated governance indicators or specialised surveys. The shortcoming of the latter comes from the fact that it is implicitly based on relatively loose relationships between institutions and development, as can be derived from the empirical cross-section growth literature. Equally striking is the difference between our approach and theoretical approaches to the role of institutions in development, which are necessarily simplified and rely on rough stylised empirical facts for confirmation. By deliberately probing the details of how the institutional context of a country affects the functioning of its economy, or at least some key aspects of it, including economic decision making at all levels, and how it interacts with political economy factors, we hope that a finer diagnostic can be achieved that improves our understanding of the institution–development relationship in the case of specific countries.

A last remark is in order. Its purpose is to dispel the idea that the institutional diagnostic approach described in the preceding pages is holistic. The approach starts with a long exploratory phase aimed at: (i) getting a rather comprehensive view of a country’s economic achievements and failings; and (ii) uncovering some salient aspects in which it differs from a priori good comparator countries, possibly emphasised by knowledgeable citizens. To help articulate the various ingredients of this exploratory phase, a structural standpoint is adopted, which looks at how resources are moved from one sector to another, privileging the Kuznetsian and Lewisian distinction between low-productivity (generally informal) and high-productivity (typically formal) sectors. It also looks at the intra-sector dynamics and the way both inter-sectoral transfers and intra-sectoral changes affect and are affected by macro-level economic policies and constraints.

From there, the analysis proceeds by delving into the key issues identified so far, whether they pertain to specific sectors or to the more general functioning of the economy. It is at this stage that attention is deliberately focused on the institutional underpinnings of these issues. In dealing with them, all kinds of possible intervening factors are subjected to scrutiny: economic, demographic, social, historical, and political. In other words, the eyes are kept wide open, and all disciplinary boundaries can be traversed in order to get a deep and complete grasp of the roots and the proximate causes of institutional failures or dysfunction. In searching for the ultimate or near-ultimate causal factors, the possible role of politics is not eschewed, as is typically the case in conventional country diagnostic studies (see, for instance, the World Bank report, ‘The East Asian Miracle’, where little is said about the political context of the ‘miracle’, despite its obvious relevance). Furthermore, in addressing politics care must be taken to go beyond cursory or perfunctory mentioning of the broad issues at play. This means that effort is undertaken with a view to elucidating the precise ways in which a political system functions and interacts with economic and social agents or groups.

As should be evident from the above summary, our approach is structured, and its encompassing and transdisciplinary character manifests around privileged axes of analysis that are not predetermined but that emerge from a methodologically constructed exploratory phase.