MSc Economics
Track: Development Economics
Master Thesis
Democratic Capital in the 20
thCentury
An empirical study on the impact of colonial democratization
policies in Sub-Saharan Africa on the post-independence
development of democratic institutions
by
Conrad Jacobs 11408251
July 2017
Supervisor: Secondary Reviewer: P.C.R. Rossi, PhD. Prof. dr. M.P. Pradhan
2
Abstract
This thesis visits the question to which extent variations in colonizers’ policy explain differences in the development of democratic institutions after independence. By exploiting disparities in democratization policies the French and British imposed on their Sub-Saharan colonies I construct a variable named democratic experience. This variable is used as an instrument to estimate the effect of a societies’ democratic capital.
Here democratic capital is defined as an aggregate of societies’ valuation of democratic principles. Using a panel data set and controlling for economic development, ethnic and
cultural factors and colonizer characteristics I find a positive, small and significant relationship.
This document is written by Student Conrad Jacobs who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the
text and its references have been used in creating it.
3
Index
Introduction……….……… 5
2. Historical Background and Theoretical Framework ………... 10
2.1 Historical Background………... 10
2.2 Theoretical Framework……… 12
2.2.1 The importance of institutional determinants……… 12
2.2.2 Designing an approach to institutions………. 13
2.2.3 Including and defining democratic capital………... 15
2.2.4 Constructing the theoretical framework……… 16
2.2.5 Applying the theoretical framework to the Casus………… 17
3. Empirical approach and data………..………. 20
3.1 Data and Measurement………..…... 20
3.1.1 Panel data construction………... 20
3.1.2 Measuring a countries’ democratic capital……….. 22
3.1.3 Control variables………...……….. 23
3.2 Empirical Approach……….……..……… 25
4. Results………...….. ……. 26
4.1 Inspecting the Instrumental Variable………..………... 26
4.2 Final IV Results………..………….. 27
5. Discussion……….. 31
5.1 Threats to internal validity………. 31
5.1.1 Alternatives for the IV………... 31
5.1.2 Missing data and unbalanced panels……… 33
5.1.3 OLS regressions with time-invariant left-hand variables……… 33
5.2 Theory Assumptions and External Validity……… 34
6. Conclusion ……… 37
References ……….. 40
Appendix i ……….. 43
5
1
Introduction
As the lowering of the Union Jack represented the British withdrawal from Nigeria on the 1st of October 1960, the raising of the green white and green flag represented the
promise and high hopes that Nigerian Independence would bring. The British displayed what could be described as considerable enthusiasm, with the new constitution
enshrining the practice of a liberal democracy. In the following twenty-five years the country experienced six military coups, a three year long civil war, a corrupt and ineffective bureaucracy and a near bankruptcy of the state. Nigeria was no outlier: 25 years after the wave of African independencies civil wars, the rise of brutal dictators and economic malaise quickly put an end to the widespread optimism of the British- and Francophone independencies (Ake, 1993 ; Crowder 1987 ; Ndulu and O’Connell, 1999).
The African wave of independence (1955-1965) can justifiably be regarded as a political shock, severely changing the institutions and removing the colonial elite who built the extractive colonial states. British and French conditions for independence were often linked to implementing liberal and democratic practices, such as an independent judiciary and free and fair elections (Cooper, 2008). This endowment was preceded by a post-1945 push by Britain and France to introduce democratic practices such as mass political participation. Democracy wasn’t entirely new to the Sub-Saharan colonies: some had experienced democracy on a small scale be before in the first half of the 20th
century. Kenya held its first election in 1920 and Zambia in 1924 (Collier, 1982). Despite the introduction of democratic practices, we know ex post that Sub-Saharan regimes tended to follow a general trend, with most regimes consolidating their authoritarianism till the 1970’s (Ndule and O’Connell, 1999). This, to many historians, was no surprise. As Crowder points out in his 1985 seminal article ‘Who’s dream was it anyway?’ independence should not be regarded as a historical ‘dividing line’. Yes, African states were endowed with certain democratic political institutions at independence, but the ‘deeper’ practices, principles, norms and values of the colonial legacy were carried over. The custom to resolve disputes with violence rather than negotiation, the fact that the colonial state was not set up to be run in the benefit of its inhabitants, the unchecked powers of the colonial governor, the habit of judging a countries economic prosperity on the basis of its imports- and exports: all these factors
6
were transferred from colonies to independent states.
Crowder was on to something. Acemoglu, Johnson and Robinson would later refer to this as institutional persistence, the theory that clusters of economic and political structures of European colonies persisted (2001). Acemoglu et al. used the term ‘institutions’ to refer to Douglass North definition, namely that “Institutions are the rules of the game in a society, or more formally, are the humanly devised constraints that shape human interaction” (1990). But, if institutions are humanly devised, who determines these institutions?
Institutions are largely determined by political power, and Acemoglu and Robinson distinguish two components. The de jure political power is straightforward, and is determined by political institutions (such as a countries’ constitution, its electoral system or its democratic practices). On the other hand, the de facto political power emerges from the possibility to engage in collective action (Acemoglu and Robinson, 2006). De facto political power refers to the ability of a subgroup of the population to bypass its collective action problems and utilize their resources to undertake actions (such as lobbying, co-opting the military, use of arms, etc) to pursue their goals.
As colonial elites left, an existing African elite emerged and adopted the de facto political power structure the European colonizers left behind. In this view the
aforementioned deterioration of political institutions after the African Wave of Independence should be viewed as a successful attempt by the new de facto political power holders to diminish and retract the democratic and inclusive institutions that the newly independent colonies were endowed with.
However, this is not the entire story. The general trend of newly independent African countries did reflect a deterioration of democratic institutions, but if we inspect the data more closely we can detect large differences in the development of institutional characteristics between these states. Why were certain former colonies better able to uphold their endowed democratic institutions? And more importantly: what pre-colonial developments or policies upheld and/or stimulated democratic institutional quality and accountability?
Multiple scholars have explored this question from different angles, often focussing on colonial identity or length of colonial rule. Both Fayrer and Sacerdote (2009) and Olsson (2009) argue that longer colonial rule would have facilitated the transition of more inclusive, European institutions. However, this approach is flawed
7
due to two reasons. First of all, they disregard the often vital differences that lie beneath a colonizers’ identity and/or length of colonization. These aspects include the type of institutions the colonizers chose to transfer (inclusive or extractive) and how much of the colony was actually in (direct) control of the colonizer. A second reason the
approach is flawed is because variables such as colonial identity may act as a proxy for other attributes that are closely linked to a colonizers ‘identity, such as legal systems or religion.
The question on how domestic democratization policies of the colonizers
determined the current political institutions is central to this thesis. This thesis chooses to analyse how cross-country differences in valuations of democratic principles, defined as Democratic Capital, have had a lasting effect on a colonies post-independence
democratic development
This brings us to the thesis’s main question: To what extent have colonial policies that promoted and/or introduced democratic principles influenced the post-independence democratic institutional development? This research builds upon previous papers by offering several remedies that may mitigate previous issues. First of all, I exploit differences within colonial policies to accurately assess the effect of democratic capital on post-colonial institutional development. Second, I use an instrumental variable approach to accurately assess what amount of the variation in post-independent democracy scores is associated with differences in colonial democratization policies. Third, I use a panel data as to include time fixed effects, as previous papers have proved that omitting these effects can substantially warp results.
I hypothesize that there is a positive causal relation between democratic capital and democratic institutional development. This theory rests on three premises:
1. The introduction of policies aimed to introduce democratic practices and
principles by the French and British colonialists happened quasi-independently of the colonizers identity and colonial duration (Cooper, 2002).
2. Based on more recent theoretical work on democratic capital by Persson and Tabellini (2009) and the effects of de facto and de jure political power on institutional persistence by Acemoglu and Robinson (2006, 2010), colonial policies introducing democratic practices have contributed to higher levels of democratic capital at independence.
8
3. Colonialism has been an important factor in explaining current day differences in political regimes. This premise is based on the fundamental theories of path-dependency and institutional persistence which have been discussed at length by a whole swath of scholars (Nunn, 2009). The hypothesized linkage between democratization policies is summarized in figure1.
Figure 1: Schematic representation of hypothesized link, where t=0 is equal to independence
The scatterplot of a possible reduced-form relationship is represented in graph 1. This is a scatterplot where a former colonies’ democracy score 50 years after independence is correlated with its experience with democracy defined as year of independence minus year of first election to legislative council.
Graph 1: A Reduced-Form Graph of the speculated relationship Democratizional
policies at t<0 Democratic capital at t=0
Higher quality political institutions in t>0
9
In order to evaluate this relationship I combine data from multiple sources: (1) the ‘democracy’ index from the PolityIV data series, (2) data from Nohlen (1999) and Collier (1982) on colonial and post-colonial elections to construct our variable for democratic capital (3) a host of control variables that measure a countries’ economic development, measures for cultural and ethnic fractionalization and variables that take into account variations in colonization characteristics.
The database consists out of 23 former French and British colonies in Sub-Saharan Africa that gained independence between 1958 and 1966. The choice to keep the scope of this study confined is driven by the desire to keep the sample as
heterogeneous as possible. Furthermore a number of variables are incomplete as reliable data on former colonies are hard to come by, resulting in an unbalanced panel. These shortcomings will be adequately discussed.
This thesis will be constructed in the following order. I start by discussing the historical background and theoretical framework of the casus. In the subsequent chapter I will elaborate on the employed methodologic strategy and explore the data sources. This chapter will be followed up by a discussion of the main results and a inspection of possible threats to the studies internal and external validity. Finally, a conclusion will wrap up the thesis.
10
2
Historical Background and Theoretical Framework
2.1
Historical Background
Seeing that the European colonial powers carved up the African continent irrespective of historical, cultural or religious characteristics in what we today call the ‘scramble for Africa’, it is very remarkable that almost all of these African countries exist today. At independence almost all boarders remained intact and have stayed more or less the same to this day.
Apart from the Belgian Colonies in the heart of Africa, the bulk of the former colonies had been administered by Britain and France. This territorial dominance only increased after Germany was forced to distance themselves from their colonies after World War I, leading to the transfer of, among other colonies, Togo to France and Tanzania to Britain. Even though the goal of colonialism was broadly the same (i.e. resource extraction), both colonial powers opted for different strategies to achieve this. Where France relied on a centralized style of governance, Great Britain opted for a more decentralized and indirect form of governance (Ndulu and O’Connell, 1999).
Then World War II happened. It initiated the forces that would work to
dismantle European imperialism. Europe had been severely weakened, and both Russia and the United States supported decolonization (Cooper, 2002). Evidently, Britain and France realised that they couldn’t hold on to their colonies forever and subsequently agreed that their Sub-Saharan colonies would be independence-ready by the end of the century (Oliver and Atmore, 1994). This development was hugely influential, as after 1945 the decisions made by France and Great Britain did not concern if the colonies would be granted independence, but when and under which conditions (Smith, 1978).
However, things did not go as planned. For France, the transition from the French Union to the French Community in 1958 ushered in the end of France’s direct rule in Africa. This transition, which granted overseas colonies far greater autonomy and enabled self-governance, intended to turn Africans into French citizens by means of political liberalization. The effect was the opposite, as it ignited independence
11
For Great Britain a similar dynamic played out, as a push in the devolution of powers to local governments had a similar effect: empowering local movements that inevitably strived for independence (Cooper, 2002).
The looming ‘early’ independences impelled both Britain France to speed up their introduction of democratic principles and practices (Collier, 1982). However, it was too little too late: both colonial powers were never able to catch up. In the end, African institutions were ill-prepared for the momentous task that lay before them. The wish for an untimely independence eventually came at the cost of an abrupt and
incomplete transition.
The institutions of the former colonies were endowed with liberal and plural values, often mimicking their European counterparts (Ndulu and O’Connell, 1999). However, these inherited values and practices proved to be hollow. As shown in graph 2, both former French and British colonies experienced a substantial deterioration in their democracy scores, only to start recovering from the 1990’s. By 1988 all but five former Sub-Saharan colonies could be characterized by having an open democracy with a multi-party system. The institutional setback had been decisive.
12
Even though the rushed and unequal unfolding of events after World War II lead to considerable differences in how colonies chose to enact policies that introduced democratic practices, there already had been significant differences before the start of World War two. For example, in French West Africa (Afrique Occidentale Française, or AOF) Senegal, Benin (Dahomey) and Burkina Faso (Upper Volta) had experienced their first elections to legislative council in 1879, 1925 and 1946, respectively (Collier, 1982). These variations largely surpassed a colonizers identity. This doesn’t mean that the introduction was entirely unconditional on a colonizers’ identity, as for example the introduction of democratic practices to French Equatorial Africa (Afrique Equatorial Française, or AEF) was considerably homogeneous.
The central question to this research is if this institutional fall-back into
authoritarianism and one-party states could have been mitigated or even avoided if the former colonies were endowed with a higher level of democratic capital, i.e. a more thoughtful and paced introduction of democratic principles. I do want to be very careful in making clear that the relationship that is at the core of this study is between
democratic capital and the development of democratic institutions and not between colonizers and former colonies. It is the fact that colonization brought with it exogenous shocks, such as democratization, what makes this casus so suitable for answering this thesis’ research question.
2.2
Theoretical Framework
This section will provide this thesis’ theoretical underpinning by merging two theories on institutional change. First, Acemoglu and Robinson’s (2006 ; 2010) theory on the role of institutions on growth and development will provide a benchmark through which we can better understand the casus. Second, adding Persson and Tabellini’s theory of democratic capital (2009) will add a new dimension on how we conceptualize political power. But before I delve deeper into the theory, I will shortly explain the value of understanding the determinants of institutional development.
2.2.1 The importance of institutional determinants
The importance of defining the determinants of institutional quality originate from its hypothesized link with growth theory. Usually, economists perceive aggregates of
13
human, physical and technological capital as the main determinants of economic growth. When we dig deeper and question what causes differences in these forms of capital we encounter a more fundamental question. What creates wealth? Why do some countries succeed in achieving prosperity, when others fail? In other words: if human-, physical- and technological capital are mere proximate causes for economic growth, what are the fundamental causes (Acemoglu and Robinson, 2010)?
The debate on institutional persistency and its relation with economic
development is not new. However, the idea institutions are not a solely a determinant but the fundamental cause of economic growth and development (the so-called ‘institutions rule’ view) is relatively new. Testing this idea is harder than it looks: endogeneity problems and reversed causality obscure otherwise straightforward relations, forcing economists to come up with innovative approaches to disentangle the web of causality. The most ingenious example is Acemoglu, Johnson and Robinson’s use of colonial mortality rates as an instrument for institutional quality (2001).
This relationship, between economic development and institutional quality, proves why achieving a better understanding of the determinants of institutional quality is so important. And even though the debate is far from settled, studies such as from Rodrik et al. (2004) provide us with an indication that institutions are a force to be reckoned with.
2.2.2 Designing an approach to institutions
This thesis’ initially follows Acemoglu and Robinson’s approach by modelling three types of institutional characteristics: (1) economic institutions, (2) political power and (3) political institutions (2010). As a starting point Douglass North’s definition of institution will be used as stated in the introduction (1990). “Institutions are the rules of the game in a society, or more formally, are the humanly devised constraints that shape human interaction” (1990). Acemoglu and Johnson unravel this definition and derive three key features (2010).
1- The words ‘humanly devised’ refer to the creators of the ‘rules of the game’: humans themselves.
2- Institutions are framed not in a way of what is possible, but what is not: institutions are set up as constraints on human behaviour.
14
Economic institutions define two major sets of rules that determine a countries’ economic outlook. First, it prescribes how a countries’ resources are distributed among a society (distribution). Second, it shapes key economic incentives and determines various investments in physical and human capital (performance). As different groups prefer distributions of income that are economically more advantageous to them, no one set of economic institutions is preferred by everyone. A conflict of interests has emerged.
The choice of economic institutions lay with those who wield political power. Here the concepts of de jure and de facto political power come into play. De jure political power is derived from the political institutions. For example: parliaments and judges (de jure political power holders) derive their power and legitimacy from democratic principles and a constitution (political institutions).
De facto political power refers to the possibility of a group or minority in a country to exert influence on decisions of economic and political institutions. This can happen through or outside political conventional institutions. Examples of de facto political power are: undertaking protests, lobbying, co-opting the military and revolting. De facto political power is derived from the underlying resources a group or minority possesses. These resources often refer to a groups’ financial position or the amount of capital they possess. However, exerting de facto power can also be linked to other types of resources such as control over the military or being the standard-bearer of an
appealing political philosophy (i.e. anti-apartheid with Nelson Mandela or the independence movements of Nkrumah or Nyerere).
Acemoglu and Robinson refer to the de jure and de facto political power holders as two state variables, from which an equilibrium is reached (2010). This equilibrium can be characterized as persistent: as those in power not only decide on the economic institutions but also the political institutions, they also decide on the future political institutions. The system maintains itself as power holders will always opt for economic and political institutions that benefit themselves. Through this lens we can see why it is so hard to achieve meaningful reforms: political power holders can easily persist change or enact comparable policies.
15
2.2.3 Including and defining democratic capital
Person and Tabellini provide a new approach to the dynamics of political and economic change. Mainly, they pursue a number of ideas on the relationship between democratic capital and economic development (2009). They argue that, as with social capital, democratic capital is a stock value which slowly accumulates with a societies historical experience with democratic practices. Not only that, experience with democracy is reinforcing in and of itself.
Based on the conclusion that long democratic experiences are a sufficient (but not a necessary) condition for economic development, the positive effect of democratic institutions on an economy would increase a societies’ expectations towards the future in the form of higher expected returns to investment. As returns to investment are safeguarded under democratic practices and institutions, anticipating future democracy becomes a key determinant for current economic performance.
Building upon these expectations one can see that a positive feedback loop emerges. This relationship, between the introduction of democratic practices, economic returns and democratic capital is summarized is figure two.
Figure 2: The positive feedback loop of Person and Tabellini (2009).
(1) Introduction of more democratic practices (2) Increase of economic performance (3) Increase of expected returns (4) Increase of democratic capital
16
The second part of Persson and Tabellini’s theory is concerned with the consolidation of democracy. Democratic capital should not be regarded as a flow variable, but as stock variable (2009). It is an aggregate of civic and social assets and builds up slowly over time. A society has to learn to appreciate the values of democracy before embracing it. And the only way a society within a country is able to accrue democratic capital is through learning from its own historical experiences.
2.2.4 Constructing the theoretical framework
By merging the ideas of Persson and Tabellini with Acemoglu and Robinson’s
framework we come to a result that can schematically be represented by figure 3 (the figure is based on Acemoglu, Johnson and Robinson, 2005). As discussed earlier, a countries political institutions has two determinants which form an equilibrium, its de facto and de jure power holders. This thesis chooses to append this theory by adding a third determinant: immaterial democratic values.
Immaterial democratic values are the underlying civic and societal forces from which political institutions are dependent on for what can be referred as a mandate. Even though a politician (a de jure political power holder) is elected (the political institution), if no one showed up to vote she or he still doesn’t have a mandate. Here voting or the valuation of democratic principles is the immaterial democratic value. In the same vain an appointed judge is not taken seriously if most disagreements are settled outside the law. Here trust and confidence in the judiciary are the immaterial democratic values.
So even though the de facto power holders are incentivized to pursue efforts that would diminish or co-opt a countries’ de jure political power holders, they aren’t the sole determinants of a countries’ political institutions. Naturally, if a society values democratic principles and practices than the de facto power holders have to exert more resources into its pursuit of its most preferred political institutions. To summarize: whereas de facto power holders strive for a more unequal distribution of de jure political power, immaterial democratic values are a force that strives for a more equal distribution of political power.
This thesis considers democratic capital as a quintessential part of immaterial democratic values as is encapsulates a societies’ experience and, through Persson and Tabellini, appreciation with democratic practices.
17
Figure 3: Rendering of relationship between the determinants of political institutions: de jure political power, de facto political power and democratic capital.
2.2.5 Applying the theoretical framework to the casus
When we take a look at the African wave of independence, we can identify we one large trend and one sudden shock. The large trend is the pursuit by colonizers to extend and expand democratic rights to larger share of a society by changing its political
institutions, thereby achieving a more equal distribution of power. The shock came with independence, where almost overnight the de facto and de jure power holders left and colonies were endowed with political institutions that mimicked their French and British counterparts.
The introduction of democratic practices not only built up democratic capital in the societies under colonial rule but also laid the foundation for the creation of a local elite. This new African elite would play a crucial role in the years to come. From Houphouet-Boigny in Cote d’Ivoire to Nkrumah in Ghana, almost all newly elected
18
leaders of the former colonies had consolidated their position before independence. This is in accordance with our theory. As these new African elites could be considered as the exclusive standard-bearers of the fight for independence, they had mustered considerable de facto political power. This power allowed the new African elites to step into the power vacuum that was caused by the sudden withdrawal of European elites.
However, even though both the de jure and de facto power holders had changed, the highly extractive institutions the colonizers had left behind had not (Crowder, 1987). This thesis will not speculate on the goals of the new African elites, but acknowledges that the existing institutions that incentivized extracting rents from society stayed in place (Acemoglu and Robinson, 2010). Simply replacing an elite would therefore do little to improve a colonies’ economic institutions.
This take-over by new African elites seem to contradict the accumulation of democratic capital that societies under colonial rule had experienced. Especially when you take into account the deterioration of democratic institutions former colonies experienced in the decades after independence. Wouldn’t an increase of democratic capital have led to a stimulus by the population for a continuation of democratic
principles? That depends on how you interpret the effects of the shock of independence. The aforementioned shock did not necessarily come from the takeover of a new elite. As the new African elite pursued the exact same extractive economic institutions, Crowder rightfully points out that independence was not a historical dividing line (1987). An example of this is the conservation the colonial-era currency boards. Even though these institutions were abandoned in the post-colonial era, its policies were continued by central banks thereafter (N’dulu and O’Connell, 1999).
On the other hand, we can consider the sudden endowment of liberal and democratic political institutions as a shock. As with the Nigerian independence
mentioned in the introduction, both Britain and France made independence conditional on the implementation of certain values and institutions. These values and aspects were enshrined in almost every countries’ new constitutions.
To summarize the conclusions so far: the exogenous shock of independence did change the de facto power holders, as it did not change its structure. Old colonial elites were simply swapped with new African elites. What did fundamentally change were countries’ political institutions, as this was a perquisite for independence.
19
between de facto and de jure political power. As we can derive from Acemoglu and Robinson’s characterization of political institutions, the natural response of the de facto power holders would be to restore the status quo (2010). A shock that would lead to a more equal distribution of power would therefore not be sufficient to trigger broader institutional change. A obvious response of the de facto power holders would be to ensure the extractive economic institutions are kept in place and exert influence to recreate the same or similar system political institutions. From this view the trend of institutional deterioration after independence is merely a readjustment period, in which the de facto power holders are exerting their power to restore the former political institutions.
How does the larger trend of introducing democratic principles interact with this shock? Though the policies that enhanced the creation of democratic capital were
implemented by the colonizers and therefore exogenous, the accumulated democratic capital was not. This thesis takes the view that democratic capital is an important
determinant a societies’ political institutions, thereby acknowledging that differences in democratic capital would have contributed to differing political institutions at
independence.
This variation in democratic capital would have acted as a counterbalance to the de facto power holders efforts to change its endowed political institutions. As
mentioned earlier: it would have been harder for elites to dismantle these endowed democratic institutions if society would attach a lot of value to these institutions. The democratization policies France and Britain pursued before independence would have increased a societies’ democratic capital.
This observation leads us to the hypothesis that there is a positive causal
relationship between the stock of a societies democratic capital at independence and its post-independence democratic development. When we take into account the broad movements of former colonies, we can hypothesize that in the immediate aftermath of independence countries with larger stocks of democratic capital were better able to withstand the political push back.
20
3
Empirical approach and data
3.1
Data and Measurement
3.1.1 Panel data construction
There have been justified criticisms on how indicators of institutional quality have been constructed in a way that makes them unsuitable for doing just that (Glaeser et al, 2004). This is mainly due to the fact that these indicators often measure choices of power holders rather than imposed constraints. As we work with a definition of
institutions that is based on the concept of constraints, these critiques should be taken seriously. I therefore choose to use the ‘democracy’ measure of PolityIV data series as the dependent variable (Marshall et al, 2016). This variable measures the openness of political regimes based on three interdependent elements1 on a scale from -10 to +10.
According to Glaeser et al., this data series “makes the greatest attempt at measuring the political environment” (2004). Each former colony that is included in the dataset has democracy scores assigned from the year of its independence till 2015.
The dataset includes 10 former British and 13 former French colonies in Sub-Saharan Africa that gained independence between 1958 (Guinea) and 1966 (Botswana). The decision to include only former French and British colonies is based on the
following reasons. First, the policies of introducing democratic principles weren’t universal and presupposes that the colonizers were democratic themselves, which wasn’t always the case for Portugal and Spain. Second, the underlying characteristics of other nations prohibited meaningful comparisons (e.g. many Portuguese colonies experienced decolonization through independence wars and Belgian colonies had mandatory voting laws). Third, the time span in which most British and all French colonies gained independence is relatively small, ensuring a higher degree of heterogeneity.
Furthermore, a number of former British colonies have been dropped from the database. Some achieved independence far too early (South-Africa) or too late
1 These elements are “the presence of institutions and procedures through which citizens can express
effective preference about alternative policies and leaders, the existence of institutional constraints on the exercise of power by the executive and the guarantee of civil liberties to all citizens in their daily lives and in acts of political participation”
21
(Zimbabwe). Others, like Sudan, weren’t official colonies (Sudan was Condominium governed by Egypt and Britain). )thers were arguably too little and too dependent on other countries to be taken into account (Lesotho and Swaziland).
3.1.2 Measuring a countries’ democratic capital
Measuring a former colonies’ democratic capital isn’t straightforward as there are no common, agreed upon statistic. The approach this thesis opts can therefore be seen as slightly avant garde and the author is aware of the potential pitfalls that come along with treading on new grounds.
The goal is to estimate the causal effect of a former colonies’ democratic capital at independence on its democratic development thereafter. A potential variable should therefore include two elements. It should measure a countries’ democratic capital at each countries’ independence, which naturally differs between former colonies.
Before independence almost each colony organized a territory-wide
independence election to elect a parliament (Collier, 1982). These elections, which were held on the eve of each colonies’ independence, contain information on former colonies democratic capital. These elections offer a wide range of potential variables, such as voter turnout rates or electoral results. For this thesis’ purposes the most indicative variable is voter turnout as a percentage of the total population2 (from now on this will
be simply referred to as voter turnout).
Voter turnout rates on parliamentary independence elections are a suitable proxy for democratic capital as it contains two elements that measure a societies democratic capital: the amount of citizens that are eligible to vote (suffrage) and the amount of citizens that choose to utilize it (electoral participation). Both elements should be included as solely focussing on one element can lead to distortions. For
example, in Malawi’s independence election voter turnout rates as a percentage of those eligible to vote was very high, at 95,2%. However, those eligible to vote only made up 2,6% of the population (Nohlen et al, 1999). Solely focussing on one measure would significantly decrease the internal validity.
There are disadvantages that come with using voter turnout as the independent variable. The primal concern is if voter turnout accurately measures a societies’
2 This data is available in to Nohlen, Krennerich and Tibaut’s published work ‘Elections in Africa: A Data
22
democratic capital. Ruth Berins Collier rightly points out that high voter turnout rates reflect either two sorts of electoral participation, a distinction she derives from Marxist literature (1982). First, there is mobilized participation. This displays the ability of the de facto power holders to persuade and influence the masses. On the other hand, there is autonomous participation. Autonomous participation refers to the interests and capacity of citizens to autonomously articulate their interests.
A lot of the evidence that independence elections were primarily driven by mobilized participation can be provided by the percentages the largest parties stacked up in the independence elections (Nohlen et al, 1999). Outright majorities were the norm, and often voters only had one choice. For example, Houphouet-Boigny’s PDCI (Parti Democratique de la Cote d’Ivoire) won a whopping 100% of the votes in its first parliamentary election in 1959. This trend fits our theory nicely: the new African elite wielded their de facto political power to claim and pursue de jure political power, which had to be achieved through elections
To mitigate this problem and accurately measure the impact of democratic capital we need to find an exogenous source of variation. As mentioned earlier in
chapter 2, a possible exogenous variable is the aggregate of the efforts of the colonizers’ to introduce democratic practices to their colonies. The question is now: how do we measure the countries democratic experience ?
To answer this question we can take a look at Persson and Tabellini’s measure of democratic capital and simplify it by assuming that democratic capital in a society under colonial rule does not depreciate (2009). This leaves us with the following
measurement:
𝐷𝐸𝑖 = ∑ 𝑡𝑣
𝑡𝑖,0−𝑡𝑖,𝑣
𝑣
With 𝐷𝐸𝑖 a countries’ 𝑖 democratic experience, 𝑡𝑖,0 the year country 𝑖 achieves independence and 𝑡𝑖,𝑣 a vector which represents the year country 𝑖 experiences the democratic milestone 𝑣. In total, there are four different democratic milestones which are all derived from Colliers (1982). These four milestones are: the first election of a legislative council, the first African that is elected to a legislative council, the first
23
territory-wide election and the first election with universal suffrage. If for example a country achieves independence in 1960 and it experiences the four aforementioned milestones in 1940, 1945, 1950 and 1955 its democratic experience is 20+15+10+5 = 50.
Even though democratic milestones can slightly overlap (for example, if the first election where an African is elected to legislative office happens with a countries’ first nationwide election) it is important to take all four in consideration. This is because omitting one factor may crudely distort a countries democratic capital. Kenya was one of the first former colonies to experience democracy on a small scale in 1924 in an election where the vote was highly restricted. However, the country has not had an election with universal suffrage before coming independent (Collier, 1982).
3.1.3 Control variables
Even though this thesis instruments its independent variable, reverse causality and endogeneity problems impend its validity. Therefore a number of additional control variables have been added to uphold conditional mean independence. These variables can be categorized in three categories: economic development, ethnic and cultural factors and characteristics of colonization.
Controlling for economic growth is based on Lipset’s hypothesis: that economic progress leads to better political institutions (1959). If Lipset’s hypothesis would be correct, the extent to which a countries’ institutions are integrated with its economy is highly endogenous (Rodrik et al, 2004). Even though multiple studies have provided mixed evidence at best (Acemoglu et al, 2008; Glaesar et al, 2004) this thesis takes the approach to include two measures that proxy for economic development on the basis that it might mitigate other potential sources of omitted variable bias. First, data from the World Bank Indicators on annual GDP growth will be transformed into logarithms and lagged by one year. Second, I include the scaled absolute value of latitude, as temperate zones enjoy healthier climates and have more productive soils (Landes, 1998). Measures on malaria infection rates, which authors like Gallup and Sachs see as a key factor in determining variations in economic growth rates between countries, are not included due to the fact that all included countries scored equally on the Malaria Index (which was set up by Gallup and Sachs (2001).
24
proposed association between a countries’ ethnic heterogeneity and democratic institutions. This idea stems from Putnam’s theory, which states that trust in strangers is a prerequisite for facilitating collective action and is therefore essential to a
accountable, democratic and efficient government (1993). In ethnic heterogeneous societies those in power are often incentivized to subdue or restrict political freedoms of other ethnic groups. The Rwandan Genocide of 1994 is an example of a worst-case scenario. This distrust incentives intolerance towards governance, and Putnam concludes empirically that trust in others is a good predictor of good governance (1999). Including measures of ethnic fractionalization may mitigate these theorized effects, and this thesis wields two different variables. First, I choose to include Posners’ Politcally Relevevant Ethnic Groups (PREG) from 2004 as scholars have found links between it and government performance (La Porta et al, 1998). This measure is an improvement on Easterly and Levine’s measure of Ethnic-Linguistic Fractionalization (1997). Both ELF and PREG measures are negatively correlated with our voter turnout and democratic experience variables. Second, variables measuring the percent of the population that worships Catholicism, Islam and Protestantism are taken from La Porta et al (1999). The theory behind this theory is that Islam and Catholicism can be
regarded as ‘hierarchical religions’ and seem to incorporate structures that exhibit a negative effect on government performance and democratization (Landes, 1998). For example, both Islam and Catholicism are regarded as more interventionalist than Protestantism. This is supported by empirical evidence as La Porta et al find that countries with large amounts of Catholics or Muslims are associated with below-average performing governments (1999).
Our third category of control variables include two variables that are added to control for potential characteristics of colonizers. First, a binary indicator that indicates if a country has been colonized by the French or the British is added to control for characteristics that are linked to colonizer’s identity such as the legal structure (Civil law vs Common law). These are possible determinants of government performance and are correlated to our variables due to France’s ‘one-size-fits-all’ policy for a few of their territories (namely the AEF). Second, Fayrer and Sacerdote and Olsson conclude that the length of colonization had an significant positive effect on a countries’
democratization after independence, a measure of years under colonial rule is added as well (2009; 2009).
25
3.2
Empirical Approach
I will follow a straightforward econometric analysis in an attempt to estimate the causal relationship between democratic capital and democratic institutional development. This will be achieved by employing a time-series cross-sectional analysis in which the
dependent variable is the democracy score from the PolityIV data series of former colonies 𝑌𝑖 in year 𝑡. The dependent variable, voter turnout, will be instrumented by the measure of democratic experience. The reduced-form regression is as follows.
𝑌𝑖,𝑡 = 𝛽1(𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐_𝑒𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒𝑖) + ∑ 𝛽𝑧𝑋𝑧,𝑖 + 𝜆𝑡+ 𝜖𝑖𝑡
Here 𝛽1 is the variable of interest: the effect of an increase of 1 in our democratic experience variable on a country’s post-independence democracy index. 𝑋𝑧,𝑖 is a vector of control variables and 𝜆𝑡 are time fixed effects. Time fixed effects are included to control for variables that are constant across entities but evolve over time. An example of this is the changing of policies of donor countries/the World Bank towards African countries, such as the Washington consensus. Entity fixed effects are not included: as our dependent variable is time-invariant, including entity fixed effects would lead to perfect collinearity. 𝜖𝑖𝑡 is the error term.
I use a two stage least squares approach to estimate the impact of the effect of democratic capital on the democracy scores. The second stage regression is as follows.
𝑌𝑖,𝑡 = 𝛼 + 𝛽1(𝑉𝑜𝑡𝑒𝑟_𝑇𝑢𝑟𝑛𝑜𝑢𝑡𝑖 = 𝐷𝑒𝑚𝑜𝑐𝑟𝑎𝑡𝑖𝑐_𝐸𝑥𝑝𝑒𝑟𝑖𝑒𝑛𝑐𝑒𝑖) + ∑ 𝛽𝑧𝑋𝑧,𝑖 + 𝜆𝑡+ 𝑢𝑖𝑡
As our instrument is time invariant, it will not withstand the
within-transformation of a fixed effects estimator. Every regression will therefore be estimated using GLS random effects. With regards to the conditions for valid instruments,
instrument relevance (𝑐𝑜𝑟𝑟(𝐷𝐸𝑖, 𝑉𝑇𝑖) ≠ 0) and instrument exogeneity (𝑐𝑜𝑟𝑟(𝐷𝐸𝑖, 𝑢𝑖) =
26
4
Results
4.1
Inspecting the Instrumental Variable
The results of the first stage regression are presented in table one, specification (1) through (4). Every specification is highly significant and has F-stats above 30.00. The standard errors are displayed under the coefficients in parentheses. As we only have GDP growth data from the 1960’s and onward only latitude is included in the economic development category. Regression specification (4) is the most-preferred. This indicates that a one-point increase in a countries democratic experience is associated with a 0.15 increase in voter turnout as a percentage of the total population.
To better assess if our instrument is valid in the sense that it measures a societies’ autonomous participation and not mobilized participation we run a few additional regressions. If we look at table one specification (5) and (6) we can carefully conclude that a one-point increase in democratic capital is associated with a 0.14 increase in a former colonies’ democracy score at the year of independence and a decrease of 0.275 percent in the share the largest party won at the independence election. This fits the theory, as a countries’ experience with democracy is positively related to the extent to which an independence election was driven by autonomous participation, and not mobilized participation.
Dependent Variable Vot. Turn. Vot. Turn. Vot. Turn. Vot. Turn.
Democ. t=0
Largest part.
(1) OLS (2) OLS (3) OLS 4 (OLS) 5 (OLS) 6 (OLS)
Democratic Experience 0.087*** 0.14*** 0.141*** 0.15*** 0.14*** -0.275*** (0.007) (0.008) (0.008) (0.006) (0.004) (0.03) Intercept 19.286** * 15.455** * 15.304** * 17.953** * 9.487** 61.45*** (0.562) (0.671) (1.544) (1.59) (0.716) (3.949)
Cultural and Ethnic
Factors? - + + + + + Economic Development? - - + + + + Colonizer Characteristics? - - - + + + Observations 1262 1206 1206 1262 1206 1206 R² 0.069 0.324 0.324 0.402 0.661 0.307 SER 10.892 8.512 8.515 8.012 3.682 17.083
27
As table one provides support for the instrumental relevance condition the same cannot be said for the instrumental exogeneity condition. This is condition is equal to the argument that democratic experience should affect a countries’ democratic institutions solely through democratic capital and will be assessed in the discussion (chapter 5).
4.2
Final IV Results
The main results have been compiled in table two. F-stats are not mentioned due to them being above 30.00 for every specification. The standard errors are displayed under the coefficients in parentheses.
The outcomes seems to support the theory that democratic capital is a key determinant of democratic development. In the most-preferred specification (11) a 1 percent increase in voter turnout is associated with an increase in a countries’ post-independence democracy score of 0.127. As the dependent variable has been identified as a proxy for a societies’ democratic capital, the results are in line with the hypothesis. However, the effects seem to be relatively small. To put it in perspective: in our dataset the democracy score has a standard deviation of 5.8. To accomplish an increase of one standard deviation, voter turnout should increase by almost 50%. This is a very
profound and almost unattainable increase, as the standard deviation of voter turnout is 11.58 and no former colony has attained a value above 50%.
Another notable feature is the interaction between cultural and ethnic variables and the variable of interest. By comparing specification (8), which has no ethnic and cultural control variables, with specification 10, which does include these variables, the coefficient drops from 1.207 to 0.157. A significant decrease.
The rest of the analysis of the results in table two will be subdivided into 4 parts in which the discusses 1) time fixed effects, 2) factors of ethnicity and culture, 3)
economic development and 4) colonial characteristics are discussed.
First of all, the importance of time fixed effects can be recognized whilst
analysing the subsequent improvements when these effects are added to specification (9) to create specification (10). The regression R² improves substantially, from 0.187 without time fixed effects to 0.516 with time fixed effects.
Second, the importance of adding factors that account for culture and ethnicity prove to be worthwhile. The politically relevant ethnic groups measure from Posner is consistently significant at the 1% and 5% level across all specifications (2004). All
28
coefficients are negative, which is in line with the findings La Porta et al (1999). The coefficients of the variables measuring the Protestant, Catholic and Muslim populations are in line with Putnam’s theory. In the most-preferred specification (11) both Catholic and Muslim variables have a significant and negative effect on a countries democratic development (1993).
Third, the coefficients of the economic development indicators are relatively stable, pointing to a positive, significant relationship. However, the observation that its’ coefficients drops from 0.55 (specification 6) to 0.346 (specification 11) when we simply add the countries’ former colonizer tells us that there is more to this story. This indicates that economic growth’s effect on the democracy scores has for a large part been driven by its association with the identity of the former colonizer.3 This is in
line with earlier findings by Acemoglu et al. who found that adding country fixed effects (what our colonizer identity variable is) diminishes the statistical relationship between income and democracy (2008).
The question now is how to interpret the significant positive coefficient of 0.346 in specification (11). On the one hand this positive relationship is fundamental to Perssons and Tabellini’s theory on democratic capital (2009): as societies develop economically as a result of implementing democratic practices, countries’ higher expected returns of investment would drive a societies’ anticipation that democratic institutions are upheld in the future. This relationship is summarized in figure two. On the other hand, it also can be interpreted as a vindication of Lipset’s hypothesis: that economic growth positively effects democratic development. Even though disentangling both effects is difficult, the potential threat of reversed causality (if democracy causes economic development) means that these results should be approached with caution. The observation that the variable latitude often is positively associated with a countries’ democratic development is in accordance with Landes (1998) conclusions that temperate zones enjoy more conducive environments. However, when adding variables that control for ethnicity and culture (from specification 4 to 6) the measure of latitude diminishes substantially and loses significance. This is no surprise, as latitude
3 The correlation between the colonizers identity (which is 1 if it was France) and the democracy score 5,
29
and religion are highly correlated4 (Northern Sub-Saharan countries tend to have higher
Muslim populations).
Fourth, including measures that control for colonial colonization characteristics has mixed results. The a binary indicator for a colonizer’s identity (where British = 0 and French = 1) often is negative and significant. This is in line with numerous findings in the literature that support the theory that characteristics that are associated with French institutions, culture and tradition (such as a powerful executive branch and Civil Law) are less conducive for democratic development than their British counterparts (La Porta et al, 1999). The remarkable large and significant positive effects of French in specification (8) are probably driven by the exclusion of our PREG variable, as this is strongly negatively correlated with our binary variable French5. This means that the
French colonial empire ruled over areas that were less ethnically fractionalized than the British.
In contrast to colonizer identity, length of colonial duration only seems to have a negligible effect on democratic development as its coefficient is very small and highly insignificant in every specification. On first glance this is contrary to the conclusions from Fayrer and Sacerdote (2009) and Olsson (2009) and in line with the shortcomings put forward in chapter 1. But, as Olsson indicates, his results are primarily driven by former British colonies and in which period (mercantilist vs imperialist) a society is colonized. The inclusion of a dummy variable of colonizer identity may therefore subvert its effects.
To summarize it all up: the effect of our variable if interest is significant and highly dependent on cultural and ethnic control variables. The following chapter will delve deeper into this relationship, explore threats to its internal validity by studying alterations in the use of instruments and discuss how these results fit into the
overarching theory.
4 When we correlate our variable of latitude with our measure of Muslim, Catholic and Protestant
populations we attain 0.7477, -7.064 and -0.3036 respectively.
5 Corr(𝐹𝑟𝑒𝑛𝑐ℎ
30
Table 2: GLS random effects regressions with instrumented dependent variable. Notes: All IV estimates are based on two stage-state least-squares regressions. *** is significant at the 1% level, ** at the 5% level and * at the 10% level.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)
Dependent Variable is Democracy from Independence (t=0) to 2015
Variable of interest
Voter Turnout 0.076 0.087 -0.13 0.175*** 1.215 -0.06 0.107 1.027*** 0.117 0.157*** 0.127**
(instrumentend by democratic
captial) (0.229) (2.515) (0.127) (0.062) (32.776) (0.037) (0.226) (0.0356) (0.157) (0.05) (0.059)
Culture and Ethnicity
PREG -7.303** 6.976*** -8.17** -6.106** -6.958*** -7.061*** (3.171) (0.972) (3.907) (2.942) (0.955) (0.972) Catholic 0.129*** 0.154*** -0.109* -0.096** -0.113*** -0.112*** (0.044) (0.013) (0.057) (0.047) (0.015) (0.015) Protestant -0.013 0.013 -0.008 -0.007 0.003 0 (0.057) (0.016) (0.073) (0.05) (0.016) (0.016) Muslim -0.056 0.051*** -0.019 -0.014 -0.022** -0.025*** (0.030) (0.008) (0.046) (0.03) (0.009) (0.009) Economic Development
Log lagged annual GDP Growth 0.573*** 0.55*** -0.249 0.351* 0.333** 0.346**
(0.199) (0.155) (0.465) (0.188) (0.162) (0.16) Latitude 22.361*** 5.529 31.598*** 12.529 12.905*** 12.709*** (2.764) (3.604) (6.402) (12.328) (4.027) (4.02) Colonization Characteristics French -16.981 -4.91* 15.249*** -4.509** -4.577*** -4.214*** (383.592) (2.832) (3.364) (1.912) (0.622) (0.681) Duration of Colonization 0 -0.003 0.004 -0.002 -0.001 (0.303) (0.008) (0.004) (0.005) (0.002)
Year fixed effects? - + + + + + + + - + +
Observations 1262 1262 1206 937 1262 915 1206 957 915 915 915
31
5
Discussion
5.1 Threats to internal validity
5.1.1 Alternatives for the IV
The instrument relevancy condition has already been demonstrated in chapter 4.1 but the instrument exogeneity condition has not yet been addressed correctly. As is
displayed in table three, our instrumental variable is lightly correlated with the
residuals of our most-preferred specification (table two, specification 11). Clearly, this is a violation of the instrument exogeneity (𝑐𝑜𝑟𝑟(𝐷𝐸𝑖, 𝑢𝑖) = 0) condition.
To better be able to inspect the four pillars that make up the aggregate of
democratic experience, table three presents variations of specification (11) of table two. Here the dependent variable is instrumented by each pillar separately. It seems that every specification, and therefore every component of the democratic experience variable, does not satisfy the instrumental exogeneity condition.
One possible explanation would be that colonizers’ chose to implement (limited) democratic practices in societies’ that naturally possessed more democratic capital. For example, societies in Kenya and Senegal were able to pressure their colonizers’ to allow highly restricted, small scale election in 1920 and 1879. If this were correct, democratic capital would simply measure ‘naturally occurring’ variation in democratic capital between African societies. This would mean that democratization policies weren’t fully exogenous. The data seems to support this hypothesis as a lot of former colonies experiences democratic practices on a small scale before World War II, especially in West-Africa (Collier, 1982).
A second potential theory is that the control variables dot not uphold conditional mean independence. Table two showed us that including factors that measure cultural and ethnic variations significantly diminished the effect of our variable of interest. It is plausible that colonizers would be more careful granting ethnic heterogenous societies’ democracy, which establishes a link with the instrument. As Putnam’s theory predicts a negative relationship between ethnic/religious heterogeneity and the development of democratic institutions, the threat to instrument exogeneity is becomes visible.
32
solution is clear. As will be discussed in chapter 5.2, both measures of voter turnout and democratic experience have been chosen due to more reliable data not being available. Developing a more suitable dependent variable, directly measuring democratic capital on democratic development, would greatly enhance the internal validity of comparable studies in the future.
Table 3: Alterations of the GLS random effects regression of specification (11) in table 2. Each specification has a different instrument. Notes: All IV estimates are based on two stage-state
least-squares regressions. *** is significant at the 1% level, ** at the 5% level and * at the 10% level.
(1) (2) (3) (4) (5)
Dep. Var. is Democracy
Voter Turnout 0.127**
instrumentend by democratic
experience
(0.059)
Voter Turnout 0.162***
instrumented by (independence - First
Election to Legislative Council)
(0.124)
Voter Turnout 0.089
instrumented by (independence - First
African elected to Legislative Council)
(0.108)
Voter Turnout 0.094**
instrumented by (independence - First
Territory-Wide Election)
(0.04)
Voter Turnout 0.027
instrumented by (independence - First
Election with Universival Suffrage)
(0.043)
Correlation between instrument and residuals
0.0601 0.2488 -0.0249 -0.3227 -0.1056
Culture and Ethnicity? + + + + +
Economic Development? + + + + +
Colonizer Characteristics + + + + +
Year fixed effects? + + + + +
Observations 915 915 915 915 915
33
5.1.2 Missing data and unbalanced panels
That fact that data is missing is entirely due to the fact that not all data is available. This problem is most persistent in the annual economic growth data from the World Bank Indicators. These generally start in 1960, meaning that countries that achieved independence before 1960 have no data computed. For Tanzania, Guinea and Mali economic data is completely missing till well into the 20th century. With regards to other
variables: Mali does not have a PREG score, the dependent variable of Benin (voter turnout) is missing and a number of countries did not have data at hand on how the voter turnout rate as a percentage of the total population.
If the missing data is a threat to this thesis’ internal validity is based on the following aspects: is data is missing (completely) at random, relatively how much data is missing and what possibilities there are to resolve the problem.
The data does not seem to be missing completely at random, which is not
peculiar as achieving a random distribution is dependent on the amount of observations (i.e. central limit theorem). However, the missing data problem seems to be relatively minor: only a very small percentage of the data is missing (about 2 percent).
Omitting the countries with missing data (Benin, Mali, Mauritania, Tanzania and Guinea) does not significantly alter the results which were presented in table two. The instrumented effect of our variable of interest with specification (10) from table two is 0.136 and is significant at the 5% level.
For some missing data I computed them myself or utilized the next best thing. If for example voter turnout as a percentage of total population was unknown but the number of votes, voter turnout as a percentage of registered voters and total population was known, these measures were computed by myself. For the rest of the remaining data (such as economic growth before 1960) there isnt’t a reliable method to estimate these. As there aren’t any significant alterations in the results if the countries with missing data are omitted, I choose to not to actively mitigate problem.
5.1.3 OLS regressions with time-invariant left-hand variables
Often scholars on institutional theory do not wield panel data but instead select a specific year in which they measure a countries democratic institutions to serve as the dependent variable. For example, Acemoglu, Johnson and Robinson use a measure of
34
constraints of the executive in 1990 and Olsson uses a countries’ democracy score in 2007 (2001; 2009). Both measures are extracted from the the PolityIV dataset.
As a robustness check a number of OLS regressions with Eicker-Huber-White standard errors are presented in table four. These regressions are alterations of specification (10) in table two, with two major differences. First, the dependent variables are countries’ democracy score at independence and 5, 10 ,25 and 50 years after independence. Second , annual GDP growth is omitted due to the variable being highly insignificant across all specifications.
The estimations represented in table four are generally similar to the
estimations of specification (10) in table two. The dependent variable tends to be more significant and larger, and the control variables have the expected signs and are more significant as well. One relatively large difference is the coefficient of determination (R²) which is near to or equal to zero for specification (1), (2), (3) and (5) in table four. The sudden increase in R² in specification (4), where it reaches 0.6212, is highly remarkable.
The estimation results in table four suggest three findings. First of all, the signs, significance and quantities of the coefficients are in line with our panel data regressions. Second, that excluding time fixed effects enhances the extent to which our specification is able to predict values of democracy. And third, that arbitrarily selecting a year for the dependent variable can significantly distort results.
5.2 Theory Assumptions and External Validity
Due to careful considerations this thesis has restricted itself to a panel of 23 former British and French colonies in Sub-Saharan Africa. The literature presents a consensus on the idea that colonialism is important for explaining current day institutional differences between countries. However, the fundamental question through which channels colonialism has shaped current day institutions is still up for debate. This thesis presents a relative new approach on how to think about democratization by drawing a link with the concept of democratic capital and expressing it as a sum of collective expectations and experiences.
On a more fundamental level this thesis’ hypothesis may seem trivial. Naturally a society has to experience democracy at first hand to be able to value it. Even though democracy has to be implemented, a society has to grow towards a collective appreciation to ensure that it is self-containing. However, the assumption that
35
experiencing democracy automatically leads to a never-diminishing appreciation (or: positive valuation) may be strong. There are countless examples scattered through history where democratization has backfired, offten through what has earlier been described as mobilized participation.
Then there is the difficult approach to measuring democratic capital, and even more with interpreting it. As this thesis’ viewed voter turnout rates of independence election (as a % of total population) as a suitable proxy for democratic capital, this is a second-best approach at best. Logically one would prefer a direct measure, through for example a survey or account of political participation.
The second-best approach this thesis has opted for has another drawback:
interpreting its results. Whilst instrumenting with a countries’ democratic experience, it may seem slightly peculiar to interpret an increase in voter turnout rates as having a causal effect on a countries’ democratic development. Wouldn’t mandatory voting laws than also have a causal effect?
As noted there are shortcomings with regards this thesis’ methodology. The idea that democratic capital is a major determinant of a countries’ democratic development is rather new, so a more out-of-the-box approach was needed to come up with
applicable measures. And the results do look promising. The positive and significant coefficients have proven that this approach should be investigated further. At this stage it would be a step too far to conclude incisively that democratic capital is a key
36
Table 4: OLS fixed effect regressions with robust standard errors. Notes: All IVS estimates are based on two stage-state least-squares regressions. *** is significant at the 1% level, ** at the 5% level and * at the 10% level.
(1) (2) (3) (4) (5)
Dependent variable is Democracy at t=0 Democracy at t=5 Democracy at t=10 Democracy at t=25 Democracy at t=50
Variable of interest
Voter Turnout 0.937*** 0.711*** -0.271*** 0.106*** 0.344***
(instrumented with democratic captial) (0.047) (0.032) (0.024) (0.015) (0.037)
Intercept 1.061 -2.654* 9.37*** -6.425*** 6.994***
(2.165) (1.503) (1.105) (0.747) (1.315)
Culture and Ethnicity
PREG -13.929*** -9.716*** -11.847*** -6.03*** -9.902*** (1.607) (1.015) (0.866) (0.609) (1.058) Catholic -0.127*** -0.15*** -0.042*** 0.010* -0.17*** (0.025) (0.019) (0.009) (0.006) (0.014) Protestant -0.17*** -0.024* -0.067*** -0.002 -0.077*** (0.026) (0.014) (0.008) (0.007) (0.013) Muslim -0.048** 0.006 -0.007 0.081*** -0.137*** (0.023) (0.014) (0.008) (0.006) (0.013) Economic Development Latitude 19.908*** 18.303*** -14.240*** -3.018* 46333*** (5.156) (3.58) (2.811) (1.599) (3.933) Colonization Characteristics French -21.182*** -17.597 -1.61*** -4.955*** -7.398*** (1.00) (0.706) (0.443) (0.277) (0.698) Duration of Colonization -0.26*** -0.026*** 0.003*** 0.001 0.005*** (0.001) (0.001) (0.001) (0.001) (0.001) Observations 1156 1156 1156 1156 1156 R² 0 0.0113 0 0.6212 0 SER 7.708 5.264 4.138 2.181 5.414
