A How Different Is Tanzania Using the Synthetic WGI?

Figure 3.1 compares Tanzania with two sets of comparator countries and according to the six synthetic indicators present in the WGI database for 2018. The six indicators refer to the following institution-related areas: ‘Control of corruption’, ‘Government effectiveness’, ‘Political stability and lack of violence’, ‘Regulatory quality’, ‘Rule of Law’, and ‘Voice and accountability’. Comparator countries are of two types:

• Neighbour countries may share a close history, similar environmental conditions, comparative advantages, or political and economic organisations. The issue is thus whether such a common background does exist and, most importantly, whether Tanzania departs in any way from it, or on the contrary conforms with it. This group includes the East African community (Burundi, Kenya, Rwanda, Uganda), to which we add three countries on the southern border of Tanzania (Malawi, Mozambique, and Zambia).Footnote ²

• Another natural set of comparators are those countries that were at the same level of development, as measured by gross domestic product (GDP) per capita, as Tanzania twenty or thirty years ago and have done better since. These outperforming peer countries are all in Asia: Bangladesh, Lao and Vietnam have gained between 60 and 150 per cent in GDP per capita over Tanzania since 1990, and Cambodia substantially less (30 per cent). The issue is whether these outperformers present institutional features significantly different from Tanzania, which might explain their better performance or be a consequence of faster growth.

Before discussing the charts shown in Figure 3.1, a word must be said about the WGI database and the way these indicators are measured. As mentioned, each synthetic indicator results from the combination of those individual indicators in the original datasets that belong to each institutional area being considered – corruption, regulation, rule of law, and so on. Synthetic indicators thus capture the common information in the underlying set of individual indicators; that is, how they differ across countries. They are normalised with mean zero and unit standard deviation. As their distribution across countries is not far from being normal, their value, between −2 and +2, indicates where a country ranks in the global ordering according to a particular synthetic indicator. Roughly speaking, 0 would correspond to the median and −.5, around which most countries in Figure 3.1 tend to concentrate, would roughly correspond to the third decile from the bottom. Thus, most countries in the figure are in the middle part of the lower half of the global ranking – which comprises more than 200 countries.

Figure 3.1a

WGI: Tanzania and neighbour countries, 2018

Figure 3.1b

WGI: Tanzania and outperforming peer countries, 2018

A striking feature of Tanzania, taken in isolation, is the relative balance that is observed among the various indicators. If it were not for ‘government effectiveness’, its radar chart would be an almost perfect regular hexagon. An obvious conclusion is thus that most institutional areas described by the WGI in Tanzania are weak by international standards – that is, at the limit of the bottom third of the global ranking – but government effectiveness is a bit weaker than the others.

The comparison of Tanzania with neighbour countries shows both convergence and divergence. On the one hand, there are clearly two outliers in the region: Burundi with uniformly extremely weak WGI scores and, at the other extreme, Rwanda with scores high enough to reach the sixtieth global percentile in all institutional dimensions but ‘voice and accountability’, a clear reflection of its rather autocratic but otherwise effective leadership regime. On the other hand, Tanzania’s institutional profile turns out to be very similar to that of the other countries in the region. In Figure 3.1, Tanzania generally lies in the middle of the range defined by its neighbours – Uganda, Kenya, Mozambique, Malawi – in all areas except the control of corruption, where it apparently does less badly. Overall, if it were not for the very peculiar institutional quality profile of Burundi and Rwanda, two countries deeply marked, in opposite directions, by what has probably been the most tragic ethnic conflict in the history of the African continent, the left-hand chart of Figure 3.1 would suggest a rather homogeneous and moderately weak institutional quality profile for Tanzania and the Eastern Africa region.

When comparing Tanzania with outperforming peer countries on the right-hand panel of Figure 3.1, four features are noticeable: (1) the superiority of Tanzania over all countries in ‘voice and accountability’ and, to a lesser degree, the ‘control of corruption’; (2) the neat dominance of Vietnam in all other dimensions; (3) the relative disadvantage of Tanzania in the area of political stability – which is a bit surprising given precisely the stability of its democracy until quite recently; and (4) the similarity between Tanzania and other better performing countries in other areas. The main point, however, is that, despite those outperforming countries having grown considerably faster than Tanzania from the late 1980s to the mid-2010s, no strong differences seem to be present in their institutional quality profile, except for the superiority of Tanzania on the democratic front and the outstanding performance of Vietnam. Therefore, with the exception of the latter, growth does not seem to have brought a significant institutional advantage to the other outperformers. It is striking that Tanzania even dominates Bangladesh in all areas.

One could object to the preceding comparison with the outperforming peers that it should be carried out not in the most recent period but in the past, when income per capita in those countries was actually overtaking Tanzania’s. Figure 3.2 is the equivalent of Figure 3.1 for 2005. On the basis of the right-hand panel, it certainly cannot be said that outperformers were institutionally dominating Tanzania; it might even have been the contrary. However, what is striking is that, when comparing 2005 with 2018, all outperformers have substantially improved the quality of their institutions whereas little has changed in Tanzania, except for a slight improvement in the control of corruption, most likely the result of President Magufuli’s anti-corruption campaign, and a more sizeable worsening of government effectiveness. Faster growth among outperformers is thus associated with institutional improvement over time rather than some initial institutional advantage, which is an interesting observation.

Figure 3.2a

WGI: Tanzania and neighbour countries, 2005

Figure 3.2b

WGI: Tanzania and outperforming peer countries, 2005

The same can be said of the comparison between the left-hand panel of Figures 3.1 and 3.2. It appears there that neighbour countries in general have witnessed some improvement in the quality of their institutions, whereas this is not the case of Tanzania. As a matter of fact, it is noticeable that Tanzania practically dominated Burundi, Kenya, Rwanda, and Uganda in almost all areas in 2005, whereas it only dominates Burundi in 2018. It can thus be said that, in relative terms with respect to its neighbours and outperforming peers, the quality of institutions in Tanzania has somewhat deteriorated – except in the control of corruption – even though its ranking in the international scale may not have significantly changed.

B Exploring Alternative Synthetic Indicators

The conclusions from the comparison of WGI between Tanzania and comparator countries are interesting, and should somehow contribute to the institutional diagnostic of Tanzania: relative homogeneity of institutional quality at a low-middle international level across WGI areas, convergence with neighbour countries except Burundi and Rwanda, progress in the control of corruption, which may turn out to be less of a problem than in most comparator countries, less political stability but more democracy than outperforming peer countries, and limited improvement of institutional quality over time with respect to comparator countries. Yet the issue arises whether these conclusions may depend on the specificity of WGI synthetic indicators, in particular the way they are obtained from a variety of individual indicators and the fact that they are defined across the whole range of world nations.

Because of the growing interest in the relationship between development and institutions, many databases have been put together over the last few decades that rely on expert opinion to compare the quality of institutions across countries and in many different areas, be it the Polity IV database on the functioning of political institutions, Transparency International on corruption, Reporters without Borders on freedom of speech, the World Economic Forum Competitiveness index, the Bertelsmann Foundation Transformation Index, or Varieties of Democracy, to quote a few. As mentioned earlier, the WGI provides a statistical summary of those individual indicators found in a collection of these datasets, which presumably are related to each of the six areas that are considered in the WGI database. But even though they clearly make intuitive sense, do these areas provide the best analytical structure to study the relationship between institutions and development? Why not other areas, maybe more political or sociological, or possibly sub-areas?

The other question is whether a statistical summary based on the heterogeneity observed among all countries in the world is the best instrument to study the way institutions may affect the development process among countries at an early stage of economic development. Differences in institutional quality between advanced countries and low-income countries may not be of much relevance when trying to understand how institutions may be an obstacle to reach lower-middle income status. Would the synthetic WGI in the six institutional areas defined in that database be the same if they had been built on a sample of developing countries only?

To answer these questions, the Institutional Diagnostic Project has explored a set of alternative indicators based on developing countries and endogenously defined institutional areas. These are based on the Quality of Government (QoG) database managed at the University of Goteborg, which functions as a kind of repository of all databases gathering expert opinion in institutional areas (Teorell et al., Reference Sutton and Olomi2022). They boast today more than 2,000 individual indicators covering more than seventy years and most countries of the world, even though, of course, not all indicators are available for every year and every country – very far from it. Only a subset of developing countries and indicators were selected so as to avoid missing data and to strictly focus on institutional characteristics. As a result, the size of the country sample and the set of individual indicators were severely reduced, even when working on a single year.Footnote ³

Instead of predefining categories of individual indicators related to a single theme such as the control of corruption or the rule of law in the WGI database, a statistical procedure was used to regroup individual indicators by their informational proximity, or more precisely by their capacity to rank countries in roughly comparable order, while maximising the difference in rankings produced by distinct synthetic indicators. Each group or category of individual indicators is then summarised by a single synthetic indicator, in the same way as the synthetic WGI summarise all individual indicators behind ‘regulatory quality’ or ‘government effectiveness’. A statistical pseudo-cluster analysis permits us to endogenously define an arbitrary number of such categories with a methodology that is somehow equivalent to minimising the country-variability of individual indicators within categories and maximising differences between them.Footnote ⁴ To get a set of categories comparable with the WGI, it was arbitrarily decided to define six categories.Footnote ⁵

The novelty of this procedure lies in the statistical categorising of individual indicators based on how similar their variation across countries is, while not paying attention to what they represent. With the procedure used to summarise the informational content of all individual indicators in a category, the method extracts maximum information from the overall set of individual indicators in the database through a small arbitrary number of synthetic indicators.

The drawback of this methodology, compared with the WGI, is to make the labelling of categories less intuitive. As variables are grouped in an agnostic way, as a function of their informational content but not of their labelling, it may not be obvious a priori to find a common label. The intuition, however, is that, if the informational content across countries is similar, they must be related to some common institutional area. Experience shows that commonalities among indicators belonging to the same group are sufficient to encapsulate them under a single theme.

In our comparison of Tanzania with other countries, 160 individual indicators were selected from the QoG covering forty-five developing countries with no missing information. The preceding methodology was then applied to this subset of the QoG database, and resulted into six categories of individual indicators, each one being summarised by a synthetic indicator. Table 3.1 presents these six indicators, reporting the number of variables falling in each category and the common approximate theme they seem to cover. When needed, and to differentiate these indicators from the WGI, they will be labelled ‘QoG-DGC’ synthetic indicators (DGC for developing countries) in what follows.Footnote ⁶

It is interesting that this purely statistical categorisation of indicators led to a grouping that is not very different from the a priori grouping used by the WGI mentioned earlier. Yet there are noticeable and interesting differences. For instance, administrative capacity – or government effectiveness – and regulatory capacity are now a single indicator, suggesting that both are somewhat correlated across the developing countries in the database. This was not the case with the WGI. The same is observed with the control of corruption and the rule of law, which are now amalgamated as the issue of corruption. On the opposite side, voice and accountability in WGI are now separated into ‘voice and civil society’ and ‘democracy and accountability’. ‘Voice and civil society’ groups variables with a societal content. ‘Democracy and accountability’ describes more specifically the way political institutions work.

Overall, it is rather satisfactory to see that the institutional areas thought to be important play an important role in differentiating developing countries, and also that nuances need to be introduced, which are not present in the a priori categorisation used in WGI. That it is difficult to distinguish corruption and the rule of law, or that it makes sense in developing countries to distinguish between the autonomy of civil society and individuals on the one hand, and indicators describing the functioning of the parliament or the relationship between the executive and the judiciary on the other are useful warnings when embarking on an institutional diagnostic of a country.

Figure 3.3 is the replica with QoG-DGC indicators of Figure 3.1 built around the WGI. Both charts refer to 2018, and it can be seen they are convergent. The same regularity among the six axes is observed for Tanzania with some more weakness in ‘administrative and regulatory capacity’. In the comparison with neighbour countries, Tanzania still dominates Burundi but is close to other countries, except Rwanda – excluding ‘civil society and voice’ – a feature that was already present in Figure 3.1. As before, Tanzania does better than all countries but Rwanda in the control of corruption. When compared with outperforming peer countries in the right-hand chart, Tanzania appears a bit stronger than in Figure 3.1. It dominates Bangladesh – as before – but still appears weaker than other countries with respect to administrative and regulatory capacity and conflict and violence. Thus, the conclusion obtained earlier that institutional quality in outperforming peer countries was not overwhelmingly above that of Tanzania, and that Tanzania clearly dominated in terms of political institutions – that is, ‘voice and accountability’ in Figure 3.1, ‘civil society and voice’ in Figure 3.3 – is maintained. The main difference lies in the evaluation of Vietnam, which is relatively less favourable with the QoG-DGC synthetic indicators.

Figure 3.3a

QoG-DGC synthetic indicators: Tanzania versus neighbour countries

Figure 3.3b

QoG-DGC synthetic indicators: Tanzania versus outperforming peer countries

In sum, the alternative set of synthetic indicators derived in the present study from the QoG database and focused on developing countries does not lead us to modify the conclusions obtained with the WGI. This is clearly a test of their robustness. In particular, it is remarkable that ignoring the differences between advanced and developing countries, which are likely to strongly structure the WGI, does not really modify the relative institutional profile of Tanzania when set against those of the comparator countries considered in the present study. One could have thought that some institutions would differ across countries mostly because of the gap between advanced and developing countries but that this would matter less among the latter. Corruption may be a case in point. It clearly matters a lot when examining differences among all countries, as it is much less acute among advanced countries. It was not necessarily expected to be a differentiating feature when restricting the comparison to developing countries. It possibly reflects the importance that experts behind individual indicators put on that specific institutional feature.

C Tanzanian Institutions According to Other Indicators

Individual indicators in the databases used to build synthetic institutional indicators often originate from experts who presumably have inside knowledge about the way institutions work in a country and are able to make cross-country comparisons. Views may be different among people who are more directly exposed to the functioning of a country’s institutions, as citizens or firm managers. As a complement to the preceding analysis of synthetic expert indicators, this section compares Tanzania with the same set of countries using two surveys that are representative of users of institutions: the World Bank Enterprise Survey,Footnote ⁷ and the Afrobarometer (for the sub-Saharan comparator countries).

• World Bank Enterprise Survey

The Work Bank Enterprise Survey is a firm-level survey based on a representative sample of private firms, which collects the opinion of entrepreneurs on their working conditions and their daily experience with the institutional fabric of the country, including the government and public agencies. Their concerns are thus as much about the functioning of some particular institutions (law, regulation) as about the availability of key inputs or infrastructure. The survey asks, among other things, whether business owners and top managers identify a given topic as a major constraint.

Unlike the situation with the synthetic indicators reviewed earlier, the Tanzanian institutional context of firms is felt to be very constraining. Figure 3.4 shows how various areas are felt as more constraining by firms in the same set of countries as earlier. Firm managers in all neighbour countries but Burundi feel much less constrained than in Tanzania. Compared with outperforming peers, the difference is even more striking. Less than 15 per cent of firms feel constrained in those countries, except in Bangladesh where, as in Tanzania, corruption and electricity shortages appear to be a major constraint for more than half of the firms.

Figure 3.4a

Perceived constraints in World Bank Enterprise Surveys: Tanzania versus neighbour countries

Figure 3.4b

Perceived constraints in World Bank Enterprise Surveys: Tanzania versus outperforming peer countries

The perception of Tanzanian entrepreneurs, however, appears more negative than their actual experience. If corruption is reported as a major constraint by almost half of firms in Tanzania, only a fifth effectively experience the payment of bribes, a value substantially lower than the sub-Saharan average (a quarter) and lower than Burundi (almost a third), or Kenya and Malawi (around a quarter). The dimension in which Tanzania clearly underperforms is in the share of firms that expect to give gifts to secure contracts with the government. On this specific question, two-thirds of Tanzanian firms answer positively, much more than in neighbour countries but at a level comparable with Cambodia, Laos, and Vietnam. It suggests that in some contexts corruption is institutionalised in such a way that firms fully internalise it and do not perceive it as a constraint, while they are more perceptive in other contexts where corruption looks more like rent extraction.

The relatively pessimistic perception of firms in Tanzania and the contrast with their practical experience appear again in the relationship of firms with the tax administration. Senior Tanzanian managers report that, on average, they spend 2 per cent of their time dealing with the tax administration. This is below most of the comparator countries. Still, it translates into the worst perception of the tax administration compared with all other countries. The length of procedures may explain these differences. Interaction with public officials might not be that costly in monetary terms or in actual time spent, even if things do not move forward.

An interesting conclusion that comes from this brief review of the World Bank Enterprise Surveys in connection with the deeper analysis of synthetic institutional or governance indicators made earlier is that the context in which people assess the quality of their institutional environment matters. Experts may be right that, practically, corruption and rent-seeking in Tanzania tend to be milder than in other developing countries since Tanzanian entrepreneurs altogether seem to pay fewer bribes. Yet entrepreneurs may be more sensitive to the fact that some of them make such payments. Whether facts or perception matter more for development is an open question, but perception does drive actual behaviour, at least partially.

D Insights Gained by Comparing Tanzania with Other Countries

The main conclusion from the comparison of Tanzania with other countries is that Tanzania does not show any clear specificity in terms of institutional quality among neighbouring countries when obvious outlier comparators – that is, Burundi and Rwanda – are ignored. This conclusion has several possible explanations. One is that the indicators used in the comparison are too vague and too aggregate to show how specific the institutional landscape may be in a given country. More detailed indicators could show deeper differences, but, by their construction, they would refer to one, possibly limited, side of the landscape. The comparison with those countries that outperformed Tanzania’s growth does not show a clear institutional disadvantage of the latter in recent years. However, it is clearly the case that outperformers have been able to substantially improve their institutional quality in the last fifteen years – that is, between 2005 and 2018 – whereas Tanzania did not in any significant way. Neighbour countries also improved, albeit by less than outperformers – Rwanda being from that point of view a clear outlier.

Representative surveys conducted among firm managers and citizens yield additional insights. More than in the case of expert-based synthetic indicators, however, the problem of the reference point emerges when comparing countries. It is not clear whether differences between Tanzania and comparator countries are driven by intrinsic differences in institutional quality or by distinct reference points among respondents living in different environments. Both the World Bank Enterprise Survey and the Afrobarometer suggest that Tanzanians are more demanding of their formal institutions. This could ease up institutional reforms but does not say much about how constraining the quality of institutions may be for development.

A last remark is in order about the comparison exercise conducted in this section, in the spirit of so many studies of this kind. As already mentioned, the choice of comparator countries is crucial. Observed differences may possibly reveal a particular challenge in a country, which then needs deeper investigation. In the present case, however, care must be taken because comparator countries as well as Tanzania have in common an institutional context of relatively low quality. It is not because the control of corruptions is estimated to be slightly better in Tanzania than in the comparator countries used in the present analysis that corruption may not be detrimental to its development. In other words, the often-heard argument that corruption or another symptom of institutional deficiency ‘is as bad here than among neighbours or even outperformers’ in no way reduces their deleterious potential impact on development.

A The Survey: Design of the Questionnaire

The questionnaire has four intertwined components: (1) the personal characteristics of the respondents; (2) institutional areas perceived as the most constraining for the development of Tanzania; (3) the perception of the functioning of institutions; and (4) current (at the time of the survey) institutional developments in the country.

The questionnaire first collects information about personal characteristics of the respondents, including nationality, gender, level of education, place of birth. In a final part, it gathers more sensitive information on the past and present occupation of respondents as well as on their political affinity.

The second section of the questionnaire enumerates ten broad institutional areas listed below in Table 3.2 and respondents were asked to select the three areas that, according to them, most constrain development in Tanzania. Respondents then had to allocate twenty points among these three areas – the higher the number of points, the more detrimental the area for development. The selected areas are important for the analysis but also for the subsequent part of the survey because they determined the set of questions presented to the respondent in the main part of the survey.

The core section of the CIS comprises 345 questions on the perception of institutions. All rely on a Likert scale, ranging from ‘Not at all’ and ‘little’ to ‘moderately so’, ‘much’, and ‘very much’. Responses are then converted into discrete numbers, ranging from one to five, for the analysis. The questionnaire is inspired by the Institutional Profile Database (IPD), an expert survey conducted jointly by the Economic Services of the French Embassies, the Centre for Prospective Studies and International Information in Paris, and the University of Maastricht (Bertho, Reference Barrett, Mtana, Osaki and Rubagumya2013). The last wave of that survey taken in 2012 covered 143 countries in 2012. Respondents were staff members of the Economic Department of French Embassies or country offices of the French Agency for Development. The CIS questionnaire differs in several dimensions, mostly because many questions were adapted to the Tanzanian context. Yet about 40 per cent of the CIS questions remain similar to their IPD counterpart.

From a practical point of view, administering the whole questionnaire was not an option owing to its length. To shorten the time needed to complete the questionnaire, every question was associated with at least one of the ten general institutional areas in Table 3.2, and respondents were asked to answer only the questions related to the three institutional areas they selected in the previous step of the questionnaire, as well as questions related to a fourth area, randomly drawn from the remaining ones. This original feature of the questionnaire guaranteed that all institutional areas are at least partly covered at the end of the survey. In practice, respondents had to answer around half of all questions, as some questions appeared under several institutional areas.

Because the survey was taken only a year after the election of a new president whose mandate had been announced as rather disruptive, respondents were explicitly asked to answer the questionnaire as if no change had yet taken place in the institutional framework of the country. Enumerators were specifically trained to convey that message to the respondents. As institutions are persistent, there is little doubt that answers to the survey describe the way in which decision makers of various types in Tanzania perceived the institutional landscape that prevailed before the election of President Magufuli and the kind of influence it had exerted on the development path of the country.

Because of this potential disruption in some institutions or in the perception of them, a last section of the questionnaire was devoted to the most recent institutional changes. Respondents were asked to identify the questions to which they would have answered differently if they had been about the recent past or the near future of Tanzania. In this open-ended part of the questionnaire, respondents also had the possibility to mention institutional features that they thought were important for development and were not covered in the survey.

B Execution of the Survey

The Tanzania CIS survey was conducted between the end of January and early February 2017 in a collaborative effort between Institutional Diagnostic Project researchers, OPM, and REPOA, a Tanzanian think-tank. A total of 101 individuals were sampled in a purposively stratified sample aimed at collecting the views of people involved in, or in close contact with, institutions. Respondents had been or were in first- or second-tier positions in the decision-making structure of public, private, or civil society organisations. Their selection followed two steps.

First, survey designers defined sample strata in terms of occupation, position level, geographical constraints, and, tentatively, gender balance. By design, half of the sample were surveyed in Dar-es-Salaam, with the remaining half divided into five major cities: Dodoma, Morogoro, Mwanza, Mbeya, and Arusha. The sample also had to include a quarter of respondents from economic spheres, another quarter from the political sphere, a third quarter from the civil society in a broad sense, and the remaining quarter from various areas including the donor community, diplomats, the police and military forces, or the judiciary.Footnote ¹² Note, however, that many respondents occupied other positions in the past and thus had experience in more than one area.

It must be stressed that the CIS survey sample design had no intention to be representative of any particular population, a fortiori of the whole population. From that point of view, it is not an opinion survey, as for instance the Afrobarometer could be. People with direct experience of the way institutions work in Tanzania were targeted, and a stratification was built in within that population so that a diversity of viewpoints could be obtained. The reason for that choice is that the goal of the CIS survey was to learn from the experience of people with knowledge of the state of institutions in Tanzania rather than what a majority of these people would think about the functioning of the judiciary or the work of the auditor general.

As we targeted top-tier decision makers, they turned out to be different from the standard profile in the whole population. They were older and more educated than the general population. A majority of respondents were in their forties, eighty of them had a university degree, while twenty-nine reported to have studied abroad. Almost all of them lived in urban areas, but half were born in rural areas. Even though not in the sample stratification procedure, political diversity was achieved: eighteen respondents declared a political affinity with the ruling party, seventeen with the opposition, and forty-five reported no political affinity.Footnote ¹³ Such a diversity is reassuring as it avoids excessively laudatory or critical views in questions addressing the role of the government.

C Critical Institutions for the Development of Tanzania

Figure 3.5 shows the institutional areas most frequently mentioned by respondents as constraining development. Institutions behind public administration came first and political institutions second. Business-related institutions were in third position. On the other side of the spectrum, only four respondents chose ‘security of transactions and contracts’, possibly because this area was considered to be more specific and technical than others. Similar conclusions are obtained when taking respondents’ weighting into account, yet political institutions then rank first.

Figure 3.5

Choice of institutional areas as most constraining for development

A framing bias could partially drive the ranking of the areas, with the first areas in the list appearing more often among the choices of respondents. Still, the allocation of the twenty points by respondents over the three selected field is less sensitive to this sorting as respondents have to focus on three fields only when they allocate the points. It has remarkably little effect on the ranking. Last but not least, qualitative insights collected in the preparation of the survey and described in the next section are very much in line with the current conclusions.

Critical area choices vary by the characteristics of respondents and yield contrasting stories. For instance, female respondents gave very little weight to political institutions, but they consider that social cohesion and protection, solidarity, and relations with the rest of the world matter more for development than do the male respondents. Political affinity also played a role, with respondents closer to the ruling party, CCM, emphasising the difficulties arising from land rights, while the opposition stressed the constraints related to political institutions and to the public administration.

The choice of the top three constraints to development, according to respondents’ opinions, is a piece of information in itself, but it also determines most of the questions asked of each respondent. Areas that were not chosen by many respondents may actually work well, or they may work imperfectly without constraining development. The fourth area, randomly selected among the remaining field and imposed upon the respondents, permits an examination of that issue. Indeed, the less critical institutional field, namely ‘security of transaction and contracts’, ended up being covered by a fifth of respondents. There is thus enough statistical power to understand why this area was considered as less critical by respondents.

D The Perceived Functioning of Institutions in Tanzania

Within and across areas, the CIS aimed to identify, as precisely as possible, which specific institutions were perceived as constraining by respondents. The subsequent analysis first evaluated questions by their mean response on a scale ranging from 1, ‘most negative’, to 5, ‘most positive’. For questions asked in a negative way, the Likert scale was inverted to make sure that a higher value always meant a better perception. Questions were then ranked according to the top weaknesses and strengths of Tanzanian institutions. The last part of the analysis explored the heterogeneity of answers across sub-samples and tried to determine whether the perception of institutional weaknesses was correlated with some salient characteristics of respondents.

As in many opinion surveys, there was a mass of answers around the central position, which may reflect the default choice of respondents if they were unsure about their position. It is therefore more relevant to look at the tails of the distribution, namely questions with clearly positive or negative answers. Forty-six questions – or 13 per cent of all questions – had an average score below 2.5, while only twenty-seven scored above 3.5.

An alternative to asking respondents which were the problematic institutional areas is to look at the proportion of low-score answers among all questions that fell under that area. This is equivalent to comparing the distribution of low-score answers by institutional area to the distribution of all questions as done in Figure 3.6.

Figure 3.6

Proportion of questions by institutional areas according to their average scores

The first part of the graph shows the distribution of all questions across the ten areas.Footnote ¹⁴ For instance, the first bar in Figure 3.6 shows that 15 per cent of the 345 questions of the CIS fall under ‘political institutions’. The first bar in the second group reports that 13 per cent of the forty-six questions with a score below 2.5 are related to political institutions. In the third group, which plots the distribution of questions with an average response above 3.5, 14.8 per cent of questions are part of the political institutions cluster. Not all areas exhibit such a balanced pattern, however. ‘Public administration’, ‘ease of doing business’, and ‘land rights’ are largely overrepresented among low scores, which suggests that they comprised relatively more obstacles to development than others. This conclusion agrees with the identification of critical areas in Figure 3.5. This is not the case for land rights, of which treatment was almost unanimously perceived as unsatisfactory. On the other side of the spectrum, the ‘social cohesion, social protection, and solidarity’ area represents 20 per cent of all questions, but only 4 per cent of low-scores and 48 per cent of high scores.

A closer look at the questions that collected the lowest average scores permits us to bring more precision to the identification of institutional weaknesses by respondents. In this perspective, issues related to land come at the forefront. What comes out of detailed questions is that land laws do not seem well understood at the local level, and at this level it is common to have operations outside the legal framework, with limited transparency, and eventually involving corruption. Respondents qualified land tenure as insecure, leading to frequent land disputes and eventually feeding open conflict. Overall, the unequal and fractionalised distribution of land was often found to be a constraint for development.

The second most cited negative item is corruption. This is thought to permeate many institutions of the country, whether at the political level or in the relations between the bureaucracy, the citizenry, and business. The delegation of missions by the state to public monopolies such as electricity production and distribution (the Tanzania Electric Supply Company, TANESCO) or natural resource extraction (gas) is found to be not very transparent. Corruption in the privatisation process of public companies that took place in the recent past is also denounced. Respondents estimated that transfer prices were too low and that promised gains in efficiency were not achieved. These points remain relevant in the current management of natural resources and respondents anticipated that they will be so in the future. On a smaller scale, the role of corruption in increasing the cost and the hardship of starting a business was also mentioned.

In the agricultural sector, respondents complained about low price levels and their volatility, which they imputed to the role of intermediaries. Access to physical and financial inputs was also felt as being restricted. Both were thought to constrain the development of the agricultural sector.

Other weak points reported are more scattered, including the absence of independent trade unions, the non-indexation of wages on inflation, the low prospect of university graduates getting a position in line with their training, and the dependence of Tanzania on foreign stakeholders.

Strengths are also worth mentioning. Respondents praised the limited discrimination based on geographical origin, religion, and ethnicity, for instance, in access to public services such as school and education. More generally, the sense of Tanzanian identity appeared to be quite strong. These positive statements should, however, be put into perspective. First, given the peculiar format of the questionnaire, less than a third of respondents had to answer these questions. Second, the risk of internal conflict based on regional differences, religion, or ethnic lines is nevertheless seen by respondents as moderately high, which seems somewhat contradictory. In a different perspective, respondents consistently emphasised the feeling of security. They were also satisfied that people were free to form associations of a varying nature, violence against political organisations was limited, and the executive had strong control over the police.

Although respondents complained about the role of foreign stakeholders, they underlined the positive role of foreign aid. It was widely recognised as a source of funding for infrastructure and a driver of improvements in the health and education sector. However, its impact on corruption was also emphasised.

At the grassroots level, respondents underscored the traditional solidarity links (family, neighbours, associations, religious organisations, etc.), which provide support to those in need, as well as the role of informal microfinance institutions such as rotating saving and credit associations. On the other hand, respondents were confident that formal social protection mechanisms, such as the Tanzania Social Action Fund, would act as a complement to rather than as a substitute for informal instruments.

Perceptions of institutions varied across groups of respondents, as evidenced by an analysis of the heterogeneity of answers. For instance, women, who, in our sample, disproportionally came from the civil society, criticised the Tanzanian state for discriminating along gender, religious, ethnic, and regional lines in terms of access to the judicial system, health care and administrative services. They were also more concerned about the influence of interest groups in the design of policies.

The same disaggregation was implemented in many subgroups. It yielded results that fitted expectations. Respondents who positioned themselves closer to the opposition party had rather negative perceptions about the independence of the judiciary, the army, and the police. They also felt civil liberties were restricted. Unsurprisingly, being close to the ruling party yielded opposite views. Respondents who studied abroad were more sensitive to matters related to trade and to the influence of foreign stakeholders in national policies. They perceived Tanzania as being very exposed to competition from foreign firms, whether from neighbouring countries, other developing countries, or advanced economies. They were also concerned by the fact that foreign firms, governments, and multilateral organisations are an obstacle to the implementation of autonomous policies and reforms.

Being involved in business raised the awareness of respondents about foreign firms having an easier time establishing themselves in Tanzania and gaining access to funds from local banks. Despite being active in the private sector, these respondents found that access to information about the ownership structure of large firms was quite difficult. Overall, business managers were rather pessimistic about the progress of the middle class and considered that networks were important for accessing top official positions, compared with merit-based promotion.

E Prospective Assessment of Institutional Changes

It should be kept in mind that the CIS survey intended to capture the perception of institutions as they operated during the five to ten years prior to the time of the survey, which was about one year after President Magufuli was sworn in with an ambitious reform programme, most importantly concerning corruption. The timing of the survey is therefore quite interesting for gaining some insights into the respondents’ anticipations about the new regime. At the end of the interview, respondents were thus asked how Tanzanian institutions had recently evolved and whether their answers to the questions on the core part of the questionnaire would have been different if reference had been made to the recent past or the near future of Tanzania. In total, 90 per cent of the respondents wished to express their opinion, even though in some cases very briefly.

As many as 28 respondents explicitly mentioned a fall in corruption and increased transparency and accountability as major recent changes in the Tanzanian institutional landscape. They explained that civil servants abided by the law more, and side payments and bribes had been drastically reduced. If questions had been about the recent past and not on a longer timeframe, most felt that corruption would probably be less frequently mentioned as a major institutional weakness.

A corollary of the reduction in corruption was the improvement of tax collection. Fifteen respondents said that the recent surge in tax collection efficiency would have changed the way they answered the core part of the survey. They felt that taxpayers had a harder time bypassing their tax duties. According to a few respondents, changes in tax collection had pushed some businesses into financial trouble. They mentioned that some firms had to close operations, that many of them faced liquidity constraints, and that it had become harder to make money. On the public spending side, more effective tax collection was viewed as raising the capacity of the state to accomplish its mission. It was expected that, combined with greater accountability, this would be a guarantee of better use of public resources. Eighteen respondents mentioned that public service provision was improving, especially in the dimensions related to education, health, and infrastructure. A few of them thought this was the result of a change in the work spirit of civil servants and would eventually lead to more equal coverage of public services, to less discrimination, and to less importance of social networks when applying for a position in the public service. Clearly, however, these perspectives of a more equal and meritocratic society were aspirations and hopes, rather than what respondents had already experienced. In effect, some respondents questioned the depth and sustainability of current changes, and whether they could alter the development path of the country.

These positive prospects were somewhat counterbalanced by concerns about the transparency and accountability of the new regime. More than 10 per cent of the sample explicitly pointed out that it had become hard to express views challenging the government, although free press, free media, and even free demonstrations were essential for the accountability and transparency of public affairs. The independence of the judiciary system was also mentioned as crucial for the credibility of the executive towards citizens and firms, a view that was not limited to respondents aligned with the opposition. Actually, several respondents expressed their fears that the new administration could depart from these principles. The risk of an autocratic drift was even mentioned in a few cases.

F Discussion and Conclusions

From the CIS, a broad consensus emerges pointing to several institutional challenges. As far as general institutional areas are concerned, the major concern is about political institutions, public administration, and the ease of doing business. The judiciary system comes just afterwards. Other areas are further down, but one may also consider that they are included in the areas at the top – for example, land right management may be covered by public administration and security of contracts by the judiciary. The problem here is that it is difficult to define institutional areas that do not overlap with each other. To a large extent, this difficulty is also present in the definition of synthetic institutional indicators. It is unavoidable when institutional areas are defined in too broad a way, but breaking them down would lead to a large number of sub-areas among which it would be difficult to decide which is more constraining than others for the development of the country.

As general as it is, the institutional decomposition used in the CIS survey and the ranking given by respondents is nevertheless instructive, even though it clearly requires further analysis to grasp its meaning and its implications. In a way, this will be done throughout the rest of this volume. At this stage, it is worth stressing the convergence of the CIS-survey and the analysis of synthetic indicators in pointing to administrative capacity as a major obstacle to faster development in Tanzania, and to the lack of competitiveness of the production sector that is potentially due to a suboptimal business environment.

Individual questions in the core part of the questionnaire yield more precise insights about respondents’ perception. Summarising them leads to the following list of consensual institutional challenges:

• the management of land rights and, more generally, the allocation of land;

• corruption at the level of both politics and the public administration;

• the regulation of the economy, in particular of infrastructure;

• the lack of transparency and accountability of the state.

Note that these challenges fit in their own way the ranking of broad institutional themes by respondents. Corruption, and the transparency and accountability of the state clearly affect how the functioning of both political institutions and the public administration are perceived. On their side, the management of land rights and the regulation of the economy also cut across broad themes such as public administration and the ease of doing business.

On the strength side, the survey again shows some consensus around the sense of national identity and security, which implicitly seems to point to political stability.

The open-ended discussion with the respondents at the end of the interview made it possible to check that the recommendation to complete the questionnaire bearing in mind the institutional Tanzanian context during the last five to ten years had been complied with. This did not prevent optimistic expectations and hopes about the way the new administration would address some of the preceding challenges.

Stepping back from the analysis of results, the question then arises of whether a survey which relies on the country’s political, economic, and social decision-making population leads to a different evaluation of institutional quality than the expert-based institutional indicators found in international databases. As a way of testing this, use has been made of the fact that many questions in the CIS overlap with the IPD questionnaire submitted to some French diplomats posted in Tanzania. Based on common questions between the two questionnaires, it is possible to measure the degree of correlation between the opinions of a sample of 100 economic, administrative, or academic actors in Tanzania and those of a few close foreign observers. There is some convergence between the two surveys, but it is very partial.

If we select the 130 questions that are identical in the CIS and the IPD, the correlation of answers between the two surveys is only 0.30.Footnote ¹⁵ On the same set of questions but within the CIS, the correlation between Tanzanian respondents and foreigners is also limited, reaching 0.5. This rather low degree of correlation, combined with the heterogeneity analysis, shows the importance of the identity of respondents to this type of survey. Many studies rely on few respondents per country, who often share a similar position in society. They have their own view of institutions, which may not be shared by Tanzanian or even other foreign diplomats active in Tanzania (the correlation between French diplomats and foreign respondents is only 0.22).

By enlarging the sample of respondents, the CIS survey is innovative and offers a more diverse view on institutions. Within broad areas, the CIS yields more precise answers on what is found to go wrong and for whom. Most importantly, it allows us to analyse the diversity of perceptions across population groups in the society, which is essential in interpreting sample averages. From that point of view, the Tanzanian experience suggests that a substantially larger sample of respondents would have yielded more precise estimates of cross-averages.