In Western Europe and North America, the process of extending citizenship rights varied much with regard to both the time of initiation and the length and scope. Nonetheless, the developed or first world countries basically reached the same general terminus: liberal democracy concomitant with some kind of a welfare state. Or, to use the tripartite division of citizenship rights suggested by T. H. Marshall in his seminal essay, Citizenship and Social Class, they ended up with the bounded whole of civil, political and social rights. According to Marshall, there was a systematic sequencing to this development. In gist, civil liberties preceded political rights which then preceded social rights. This sequencing was so clear-cut that Marshall fixed it exactly on the temporal dimension. To quote: ‘it is possible, without doing too much violence to historical accuracy, to assign the formative period in the life of each to a different century – civil rights to the eighteenth, political to the nineteenth and social to the twentieth.’Footnote 2
Even though Marshall focused exclusively on the history of Great Britain in his essay, much indicates that he did regard this evolution as typical of the capitalist West as a whole.Footnote 3 Moreover, others have argued that the sequence is in fact an apt description of the evolution of rights in this part of the world.Footnote 4
Not so, however, in transformation and developing countries of the contemporary era. Here the extension of rights simply does not seem to follow the same sequence. As a number of scholars have pointed out in the recent decades, democratic elections often take place in the absence of civil and social rights.Footnote 5 Most noteworthy, Guillermo O'Donnell has argued that Marshall's route is the road not taken these days. Invoking Habermas, Weber and Marshall, O'Donnell's point of departure is that – with some important differences in scope and timing –‘the expansion and densification of civil rights in highly developed countries basically took place well before the acquisition of political and welfare rights’. In most contemporary Latin American countries, however, the situation is different. Though ‘the political rights entailed by this regime have become generally effective, the extension of civil rights to all adults is very incomplete’.Footnote 6 Or – to use a more general formulation of O'Donnell's – today ‘political citizenship may be implanted in the midst of very little, or highly skewed, civil citizenship, to say nothing of social welfare rights’.Footnote 7
This observation has been used to make the descriptive argument that the so-called third wave of democratization is intrinsically hollow. In its wake, it leaves few liberal democracies but many electoral or illiberal democracies. These illiberal democracies are – by definition – deficient with regard to civil liberties. But they are also unable and/or unwilling to implement policies reducing social inequalities.Footnote 8
ADAPTING MARSHALL'S SEQUENCE TO THE TWENTY-FIRST CENTURY
Important issues are thus at stake in the debate. But what is the theoretical logic underpinning these claims? The premise is to be found in the fact that most of the developing countries exhibit inauspicious structural bases for democratization. Diamond thus points out that the majority of countries involved in the recent wave of democratization are so socio-economically undeveloped that it is unlikely that they will be able to safeguard the non-electoral aspects of liberal democracy. These inauspicious structural constraints feed into what O'Donnell calls a severe incompleteness of the state, which undermines the impartial treatment of the citizens and occasions an inefficient implementation of policies.Footnote 9
For these reasons, political rights are likely to be more widespread than civil rights in present-day developing countries. Furthermore, social rights are likely to be even more precarious than the liberal rights because both a certain level of affluence and a capable state are needed to sanction such rights. Finally, another consequence of the unpropitious structures is that that civil society is likely to stay weak – a driver that has been emphasized as indispensible for the development of both civil and social rights.Footnote 10
To be sure, all of these rights may exist on paper in developing and transitional countries, not least because of the near universal ratification of human rights conventions. However, the fundamental point is that – beginning with political rights and ending with social rights – they are likely to be ever more deficient in reality.
The structural constraints, however, only explain why citizenship rights, and in particular civil and social rights, are often curtailed, not why the countries nevertheless democratize. More technically, the inauspicious structures are constants across most of the developing countries. Not so with the zeitgeist and external pressure, the nature of which shows significant variation even across shorter spans of time. This is where the more operative variable – what has been termed ‘liberal hegemony’– enters the storyline. This hegemony owes its existence to the fact that the Western democracies emerged victoriously from the Cold War and it has been used to make a two-sided claim. On the one hand, the liberal hegemony means that virtually no developing country today can completely ignore the pressure for democratization. On the other hand, it is underlined that – due to the unpropitious structural constraints described above – the responses often take the form of lip service to democratization rather than reforms that stick.Footnote 11
The liberal hegemony reinforces the structurally induced asymmetry because of the prevailing tendency to identify democracy with elections.Footnote 12 The domestic as well as international failure to appreciate these constitutive elements of the Western bounded whole of citizenship finds expression in the fact that many Western governments have been willing to grant political bona fides to countries that hold elections, even when civil liberties and social rights are clearly violated.Footnote 13 If liberal hegemony is indeed of the essence, an important consequence – which we will return to at a later stage of this article – follows: it is only with the advent of the post-1989 liberal hegemony that the adverse combination of political rights in the absence of civil liberties and social rights has spread like wildfire.
On the basis of the theoretical points made above, it is possible to formulate a very simple hypothesis (henceforth H1):
In transformation and developing countries of the contemporary era political rights are at least as effective as (and often more effective than) civil liberties which are at least as effective as (and often more effective than) social rights.
We assess this hypothesis vis-à-vis the conventional sequence based on Marshall (henceforth H0):
In transformation and developing countries of the contemporary era civil liberties are at least as effective as (and often more effective than) political rights which are at least as effective as (and often more effective than) social rights.
Notice that the two hypotheses only differ with regard to the respective commencing of political and civil citizenship rights – in both cases social rights make up the third and final part of the sequence. Consequently, some of the combinations of scores across the various kinds of citizenship rights are covered by both hypotheses. We come back to this below when we describe the conceptual typology used to test H0 and H1.
More generally, the reason that we set out to test these two hypotheses is this: what is still conspicuously absent in the discussions described above is a systematic empirical investigation of the contemporary sequencing between Marshall's three generations of rights. The only partial exception is Latin America, where OxhornFootnote 14 and Foweraker and KrznaricFootnote 15 have demonstrated that the sequence described by H1 seems to be operating. However, a systematic large-N appraisal is still pending. In the words of O'Donnell: ‘At the present stage of our knowledge, these are no more than hypotheses that remain to be empirically explored.’Footnote 16
In this article, we set out to deliver such an appraisal. This is the most important contribution of the article. But the analysis also allows us to push at the theoretical factors described above. Thus, after the analysis proper, which is confined to the 2000s due to data limitations, we venture further back in time to deliver a cross-temporal check on our main finding. This allows us to present some conclusions about the validity of the claims concerning the onset of liberal hegemony – as well as some derivative conclusions about Western distinctiveness.
THE SCOPE OF THE INQUIRY
As indicated above, our proposition (H1) is bound in both space and time. Regarding the former demarcation line, it does not claim validity for the Western countries, most or all of which is today characterized by the bounded whole of political, civil and social rights. Regarding the latter demarcation line, the temporal scope condition of liberal hegemony is of the essence. Only after 1989 has the external pressure to democratize become so salient that political rights are likely to be encountered in the most distant of countries.
More precisely, the analyses we carry out include transformation and developing countries in the period 2003–7. Whereas the set of countries fits the first scope condition neatly, the choice of time period requires some further justification, which also allows us to introduce the dataset used to test the two hypotheses. Arguably, one of the reasons that Marshall's sequence has not been systematically revisited so far has to do with data limitations. The conventional indices of democratic rights, such as Polity's POLITY IV and Freedom House's Freedom in the World Survey, do not provide data on social rights.Footnote 17 More generally, measurements of social rights are extremely hard to come by and the few extant indicators – such as the Human Development Index or the Physical Quality of Life Index – are of an entirely different ilk from the standard procedural measures of political rights and civil liberties. This is so because they measure the (expected) outcomes of rightsFootnote 18 rather than the actual presence and respect for welfare rights. Hence, it has been very difficult to include all three attributes in one and the same empirical analysis.
The introduction of the disaggregated data compiled in the so-called Bertelsmann Transformation Index (BTI) has altered the picture, however. The BTI has been published since 2005 (so far covering the years 2003, 2005 and 2007) and includes 116 (2003) to 125 (2007) developing and transformation countries with more than two million inhabitants. Providing disaggregated data on many aspects at a low level of abstraction, the dataset allows us to capture each of the three attributes described by Marshall. Furthermore, the BTI scores are linked directly to narrative qualifiers which we will employ when establishing categorical cut-off points.
In what follows, we first create a conceptual typology that makes for distinguishing systematically between the respective attributes of political rights, civil rights and social rights. Second, we order the transformation and developing countries at the three points in time that the BTI includes, i.e. 2003, 2005 and 2007.Footnote 19 Some might object that any appraisal of sequencing requires a systematic analysis over longer time spans. After all, an established methodological point is that a synchronic relationship cannot corroborate or disconfirm the existence of a diachronic development.
Here we make two counter-claims. First, and logically, the synchronic existence of Marshall's hierarchy is a necessary (though not a sufficient) condition for the cross-temporal sequence that he describes. Second, and empirically, we will observe whether any countries move within the typology in the period 2003–7, thus allowing some appreciation of cross-temporal developments.
One problem, which does not hold consequences for the test of the two hypotheses but only for the underpinning theoretical arguments, must be squarely confronted, however. The temporal scope condition basically means that in so far as H1 is supported, the phenomenon captured by it is explained by the onset of liberal hegemony (after 1989). This will not do as it is obviously something that remains to be tested – not least because it is not uniformly accepted. Oxhorn, for instance, points out that political rights preceded both civil liberties and social rights in the Latin American (re-)democratizations in the 1970s and 1980s.Footnote 20 As already mentioned, at the end of the article we attempt to pursue this point by venturing back in time – to the period prior to the present liberal hegemony.
A CONCEPTUAL TYPOLOGY OF CITIZENSHIP RIGHTS
Before operationalizing the different kinds of rights using BTI, however, we need to clarify their meaning. As we set out to update Marshall's sequence, we stay zealously close to his description of the three kinds of rights:
The civil element is composed of the rights necessary for individual freedom – liberty of the person, freedom of speech, thought and faith, the right to own property and to conclude valid contracts, and the right to justice.…By the political element I mean the right to participate in the exercise of political power, as a member of a body invested with political authority or as an elector of the members of such a body.…By the social element I mean the whole range from the right to a modicum of economic welfare and security to the right to share to the full in the social heritage and to live the life of a civilised being according to the standards prevailing in the society.Footnote 21
Bearing these distinctions in mind, the BTI contains nine sub-components which are relevant for our analysis. We have distributed them under the three respective headings in the following ways: (1) political rights (free elections, democrats rule); (2) civil rights (association/assembly rights, freedom of expression, independence of judiciary, civil rights ensured and property rights); and (3) social rights (social safety nets, equal opportunity).Footnote 22
The systematic distinction between political, civil and social rights paves the way for the construction of a conceptual typology. As we wish to provide a relatively fine-grained overview we have chosen to trichotomize each of the three attributes using the subdivisions of the BTI. The BTI scores a given country on an index ranging from 1 to 10, with 10 indicating the highest level of accomplishment. The four-fold hierarchy of linguistic qualifiers guiding the expert assessments distinguishes between ‘no defect’ (9–10), ‘moderate defect’ (6–8), ‘severe defect’ (3–5) and ‘extreme defect’ (1–2). To trichotomize the attributes, we collapse the latter two into a common category of severe defect (1–5), which basically reflects that the attribute in question, say civil rights, is absent.
ANALYSING THE GENERAL RELATIONSHIPS
The consequent typology for 2007 with empirical referent is illustrated in Table 1. The property space consists of 33 (27) types, in which the 125 countries included in the BTI for that year are situated.
Ordering of the Countries, 2007
The empirical ordering provides a pivot for testing the relative merits H1 and H0. In Table 1, we have denoted the combinations (types) which fit Marshall's logic with an ‘M’. The types that instead conform to our expectation, stated in H1, have been shaded in two different nuances of grey. We will come back to the distinction between these two nuances below. Suffice to say here that H1 is supported by referents situated in any of the shaded cells. Notice, furthermore, that the graphical presentation of the typology mirrors the hierarchical logic of H1, meaning that a clustering along the diagonal indicates that political rights are at least as effective as civil liberties which are at least as effective as social rights.Footnote 23 Notice, finally, that six types are covered by both H0 and H1. It is therefore entirely plausible that the fit of both hypotheses may be high. However, the relative fit should still differ due to the fact that each hypothesis also covers four additional types which are not shared. It is these relative differences that are of the essence.
But how can we, on this basis, assess the goodness of fit of the two competing hypotheses systematically? As Bailey has demonstrated, the most perfect empirical specimen of a typology is one that conforms to a so-called ‘perfect simple order scale’.Footnote 24 In such an instance, the referents are classified according to a strictly hierarchical pattern, meaning that there are no tie scores on the properties (more on this later). To evaluate the extent to which a typological ordering fits the hypotheses we simply calculate the proportion of cases classified in accordance with the theoretical expectations.Footnote 25 We use this measure to assess and compare the relative fit of H1 and H0. To further establish the absolute strength of the two propositions, we calculate the value produced by a random distribution as a frame of reference.
Regarding the 2007 numbers used in Table 1, 58 per cent of the cases (73/125) conform to the logic of H0. A random distribution would make for a value of only 37 per cent (10/27). At first sight, the evidence thus seems to render unequivocal support to the contemporary relevance of Marshall's sequence. But looks may be deceiving. The equivalent proportion for H1 is an impressive 89 per cent (111/125), which is the same as saying that it finds much stronger empirical support.
How about the corresponding empirical orderings for the twin years of 2003 and 2005? To keep things simple, we refrain from reproducing the typological presentations for these years in the text and only take stock of the respective coefficients of reproducibility (cf. fn. 25). Interested readers may find the two typologies in Appendix 1. In both 2003 and 2005, the proportion of cases supporting H0 was somewhat higher than in 2007, namely, 0.70 (81/116) and 0.64 (76/119), respectively. However, the proportion supporting H1 was once again much higher, namely 0.89 (103/116) in 2003 and a stunning 0.92 (110/119) in 2005.
To conclude, H1 has a remarkably better fit with the empirical reality than H0. Furthermore, H1 is strongly supported in an absolute sense since the proportion associated with a random distribution would be 37 per cent. Finally, it is worth recalling that H0, too, has a better fit than such a random distribution. There are two reasons for this. First, because the part of the sequence that is similar across H1 and H0, i.e. the proposition that social rights are placed at the end of the chain, holds up strongly. Second, because a large number of countries exhibit the same status on civil liberties and political rights; a combination covered by both hypotheses.
Notice, however, that the combinations which are unique for H0 hold very few referents over the period. In fact, they only contain two in 2007. In 2003 and 2005 the equivalent numbers are two and one, respectively. The same is not the case for H1 which houses a substantial amount of referent in the types which are unique to this sequence. In 2007, no fewer than 40 countries are situated in such types. In 2005 and 2003 the equivalent numbers are 36 and 25.
Most of these types are interesting for one further reason: they represent different values on the properties and thus make up more theoretically interesting combinations than many of the types shared by H0 and H1 which instead combine the same values on two out of three or three out of three properties. Seen from this perspective, there are significant differences between the respective fits of the two hypotheses. One may, on this basis, argue that the high fit of H0 is somewhat misleading. That said, what the analysis shows is basically that the alternative sequence of H1 is the more prevalent one in the contemporary era.
As regards the cross-temporal movements in the, admittedly, short period under consideration, we find 37 changes from 2003 to 2005 and 30 changes in the period 2005–7. Out of these, only four and six cases, respectively, disconfirms our expectations by moving beyond the shaded types, whereas the corresponding numbers for H0 are no less than 20 and 18. In this respect, too, H1 is therefore supported in both an absolute and a relative sense (i.e. vis-à-vis the random distribution and vis-à-vis H0).
DISCUSSING OUTLIERS
Even though the fit of H1 was impressive, some outliers do exist. The same of course goes for H0. In fact, the most salient outliers to H0 are exactly those countries that conform to the first part of the H1-sequence, i.e. countries in which political rights are more effective than civil rights.
Returning to H1, one unexpected combination (i.e. type) in particular stands out as it contains a significant number of referents over the period. This is type 24 where (in 2007) we encounter the countries Cuba, Kazakhstan, Libya, Malaysia, Oman, Singapore, Thailand and Tunisia. These countries combine a status of severe defects on both political and civil rights with a status of moderate defects on social rights, meaning that they turn the expected sequence upside-down. As regards the distributions of 2003 and 2005, the same pattern is evident, as type 24 contains a majority of the outliers to H1.Footnote 26
Generally speaking, then, this combination of attributes seems to be a viable one. More particularly, three different clusters of countries can be identified in this type. First, there are those formerly or presently communist countries in which a historical legacy of ideological commitment to welfare programmes has gone hand-in-hand with harsh repression of political and civil liberties. Second, there are the Asian developmental countries that have used their newly gained wealth to introduce some degree of social rights but still have not – as have South Korea and Taiwan – moved away from holistic authoritarianism. Finally, we find a number of Arab countries that can afford social benefits which are used to pacify (‘bribe’) the population so as not to demand political and civil freedom.
ARE THE FINDINGS ROBUST?
The presence of some outliers does not alter the main conclusion: that the empirical regularities of the typological ordering are extremely salient. But, some might object, could this be a result of tailoring the threshold or of the choice of dataset? Anticipating such speculations, and to support the reliability of the analysis, we test the robustness of the findings in two steps. First, we alter the employed BTI-thresholds. Second, we reorder the countries using the Freedom in the World (FH) survey provided by Freedom House.
To keep down the complexity of the presentation, this time around we only scrutinize the data for 2007 and refrain from reproducing the additional typologies graphically. The BTI robustness test is straightforward as the dataset has already been described. This time we use the following thresholds: ‘no defect’ (8–10), ‘moderate defect’ (5–7) and ‘severe defect’ (1–4), meaning that this new ordering is more lax than that carried out above. In this lax BTI-ordering, the proportion of cases supporting H0 and H1 is 0.63 (79/125) and 0.85 (106/125), respectively. Even though the differences are less stark than in the principal analysis, the relative goodness of fit still differs significantly.
Unfortunately, the FH robustness test only allows us to appraise the first part of the sequence, i.e. the distinction between political rights and civil liberties. Such is the case because the FH – as already mentioned – does not include measurements of social rights.Footnote 27 This means that this second robustness test cannot corroborate or question the two hypotheses in toto. But this is not too much of a problem as the FH data allows us to test the part of the sequence that actually differs between H1 and H0. To unfold the entire typology, we simply add the original BTI 2007 measurements of social rights to the FH measurement of political and civil rights.
Two additional elaborations are warranted, though. First, to mirror Marshall's two attributes of political and civil citizenship (and the BTI operationalization) as closely as possible, the political rights attribute is covered by the sub-components Electoral Process and Political Pluralism and Participation. The civil liberties attribute is covered by the sub-components Freedom of Expression and Belief, Associational and Organizational Rights and Rule of Law.
Second, we have recalibrated the scores from 0 (no freedom) to 100 (full freedom) because the different sub-components are not on the same footing.Footnote 28 A threshold of 85 is used to demarcate the status of no defects and the threshold of 49 is used to separate moderate defects (50–85) from severe defects (49 and below). Using these thresholds, and ordering the same 125 countries as in the previous analysis, the percentage of cases supporting H0 is 64 (80/125) whereas the analogous proportion for H1 is no less than 90 per cent (112/125). These numbers thus also support our previous findings. Taken together, we therefore conclude that the results – and hence the established fits of H0 and H1 – are indeed robust in the period placed under scrutiny.
A PERFECT HIERARCHY?
The goodness of fit of H1 is so remarkable that it makes sense to push further at the hierarchical logic underpinning it. Our hypothesis can – based on the stipulated trichotomous understanding of defects – be translated into a more demanding hierarchical proposition, which we term H2:
Transformation and developing countries of the contemporary era are characterized by a perfect hierarchical relationship with regard to their fulfilment of the criteria underpinning the three attributes of political rights, civil rights and social rights.
H2 points back to the basic formula for a perfect simple order scale (also known as a Guttman scale) mentioned earlier in the article. To demonstrate this, it is suitable to convert our ordinal distinctions into numbers. On each of the three attributes, we therefore let 2 denote no defect, 1 denote a moderate defect and 0 denote a severe defect. The combinations which conform to the notion of a perfect simple order ranking are the following: (2,2,2), (2,2,1), (2,1,1), (1,1,1), (1,1,0), (1,0,0) and (0,0,0) – i.e. the scale is characterized by the fact that there are no tie scores. Relating this to Table 1, these combinations are indicated by a darker shade of grey (presented by types 1, 4, 13, 14, 17, 26, 27), whereas the lighter shade of grey are those conforming to H1 but not to H2.
Using the strict BTI-ordering covering 2007, 85 per cent of the cases (106/125) corroborate H2. This time around, a random distribution would imply a value of only 26 per cent (7/27). The empirical fit is therefore still very high. In fact, it exactly passes the 0.85-criterion normally used as a rule of thumb for establishing the presence of a strong simple order (a Guttman) scale. What is more, in the distributions for 2003 and 2005, respectively, the proportion of supporting cases is even higher, namely 0.88 (102/116 and 105/119). In gist, the perfect hierarchical logic underpinning H2 resonates strongly in the data.
The presence of such a systematic pattern presents us with various methodological advantages. First, as already hinted, a simple order scale is strongly one-dimensional. Consequently, when moving from the typological property space tout court to the scale one may resort to simple, functional reduction (i.e. deleting the empty cells), thus retaining the notion of a hierarchy among the attributes. Second, and related to this, the systematic sequencing in attribute fulfilment allows us to create an ordinal scale of citizenship rights – or of political regime forms if the combinations are conceptualized with reference to democratic theory. Hence, based on the tripartite division of citizenship rights the regime forms would run from full repression of citizenship rights at one end of the spectrum to their complete and joint fulfilment at the other end of the spectrum. Notice that what would basically be created is thus an ordinal scale extending from the thinnest (1,0,0) to the thickest (2,2,2) types of Marshallian notions of citizenship; what could also be understood as a scale capturing ever-thicker concepts of democracy.
EXTENDING THE ANALYSIS BACK IN TIME
The prior analysis has clearly demonstrated that the pattern described by H1 is more prevalent than that described by H0 these days. Yet this conclusion only applies to the period of 2003–7, for which we had adequate data to measure Marshall's tripartite division of citizenship rights. A fortiori, we have so far been unable to test the underlying theoretical assertion about the onset of liberal hegemony after the fall of communism. By the same token, we have not been able to assess whether the developing countries have only recently departed from the Western route described by Marshall or whether they have not followed it at all.
As already mentioned, we do not have solid data on cross-national social rights for the period prior to 2003. Recall, however, that H0 and H1 differ only with respect to the sequencing between political rights and civil liberties. In so far as we can trace the sequencing on these two attributes back in time, it will thus be possible to deliver a tentative test of the validity of the factors or scope conditions on which H1 is based. Equally importantly, such an analysis would make up a robustness test of the prior analysis as it would give us a glimpse of the cross-temporal robustness of H1.
It is indeed possible to extend (part of) the analysis back in time if we once again turn to the FH's Freedom in the Worlds Survey. Whereas we picked and chose on the sub-component of the FH in the prior robustness test for 2007, this is not possible with the cross-temporal analysis as the scores on the sub-components are not available for this period. Instead we are forced to rely on the general distinction between ‘political rights’ and ‘civil liberties’.Footnote 29 On each of these components, the FH scores the countries between 1 (best) and 7 (worst). To replicate our prior analysis, we have trichotomized the indices so that the scores 1 and 2 correspond to ‘no defects’, 3 and 4 to ‘moderate defects’ and 5–7 to ‘severe defects’. The Freedom House ratings reach back to 1972. As we have more adequate data from 2003 onwards, we have tested the two hypotheses in the period 1972–2002.
This time around the developed countries are also included in this ordering, meaning that the analysis is truly global. This should not have any impact on the relative fit of H0 and H1 as the clear majority of these countries are likely to be characterized by the bounded whole of political, civil and social rights – a combination covered by both of the hypotheses.
The logic of the temporal robustness test is illustrated in Table 2. Once again, the shaded cells (all of which are shaded in the same nuance) are those that correspond to H1 whereas those marked with an ‘M’ corresponds to H0. Notice that this time around, the two hypotheses are each other's mirror image, i.e. they are completely symmetric around the diagonal. This means that the fit of the two hypotheses may not differ much in absolute terms (as they share the entire diagonal).
Ordering of the Countries, 1972
Table 2 shows that in 1972, H0 actually had a higher consistency than H1. Of the 149 countries, 90 per cent (134) fall as predicted by H0 whereas only 87 per cent (130) fall as predicted by H1. It is not surprising that the consistencies are higher than in the prior analysis, as the random fits of the two hypotheses have risen from 37 per cent (10/27) to 67 per cent (6/9). The numbers thus cannot be directly compared across the two analyses – it is once again the relative fits of H0 and H1 that are interesting as well as the fits compared to that of a random distribution.
What happens as we move closer to the present? Figure 1 shows the relative consistencies of the hypotheses in the period 1972–2002.
Consistency of H0 and H1, 1972–2002
An interesting pattern is laid bare. Until the late 1970s, H0 retains the better fit. But then the relationship is turned upside-down as H1 shows a much higher consistency. By the mid-1980s the difference is that between a consistency level of 0.97 and one of 0.87. Generally speaking, this difference is then maintained throughout the period, the only exception being the years around 1990 in which the gap between the fits of the two hypotheses narrows strikingly. This hiccup is obviously related to the fall of communism, which created a state of flux in which the extension of the franchise did not always precede the extension of civil liberties (e.g. in the Baltic countries where the Russian minorities were long barred from the vote). Notice, also, that the fits of both H0 and H1 are consistently higher than that of the random distribution.
What, then, can we say about the onset of liberal hegemony? On the one hand, with respect to the two first links in Marshall's chain, H0 was supported to a higher degree than H1 as recently as the late 1970s, meaning that the alternative development described by H1 seems to have kicked in only recently. On the other hand, this shift occurred a full decade before the breakdown of communism and the concomitant onset of liberal hegemony.
More must be said on this issue, however. The 1990s did in fact evidence the largest gap between the fits of H0 and H1 found over the period. Moreover, if attention is directed toward the third link in the chain, social rights, the full sequence described by H1 was probably strengthened much more after 1989–91 than is apparent in Figure 1. Such is the case because the countries of the Eastern Bloc – notorious for their combining of social rights with the absence of political rights and civil liberties – changed their ways: some went democratic while retaining a welfare state (e.g. the new EU-members), others remained autocratic but with more deteriorating levels of social provisions (e.g. the countries in Central Asia and the Caucasus).
Where does this leave us? Most obviously, the cross-temporal analysis lends further support to the general conclusion of this article, namely that Marshall classic sequence has largely been replaced by one in which political rights precede their civil counterparts. Or, more precisely, that the sequence described by H1 has become more prevalent than that described by H0. Here it is once again worth noting that the combinations unique to H0 contain very few referents through most of the period whereas such is not the case for H1. An example is illustrative here. In 1974, 23 referents are situated in the types that are unique to H0 (and therefore represent distinct values on the two properties). Only 11 referents are found in the types that are unique to H1 (and therefore represent distinct values on the two properties). Twenty years later (1994), the equivalent numbers are 7 and 32.
But the supplementary analysis also presents ammunition for questioning the theoretical underpinnings of this new sequencing. The change favouring H1 clearly took place before the onset of liberal hegemony. Thus, this factor becomes less critical, although it probably reinforced the new sequencing. Of course, it is plausible that the liberal hegemony was in ascendancy even by the early 1980s. After all, the communist model was not delivering in this decade as evidenced by the fact that the countries of the Eastern Bloc were lagging more and more behind the Western world with respect to economic performance. Nonetheless, the pattern described by H1 holds up more strongly than does its postulated causes.
CONCLUSIONS
We set out to appraise the contemporary sequencing of Marshall's three citizenship rights, namely, civil, political and social rights. The scope of the initial empirical inquiry was restricted to the developing countries and to the period 2003–07 (the latter scope condition owing to data-limitations). We proposed that under the auspices of the third wave of democratization, Marshall's sequencing has been altered so that political rights are at least as effective as civil rights which are at least as effective as social rights. The empirical analyses carried out – including the first two robustness tests – strongly supported this hypothesis. To be sure, Marshall's sequence also exhibits a relatively impressive fit (compared to a random distribution). But this is first and foremost a consequence of the fact that many countries exhibit the same status on some or all of the properties.
The fit of H1 was so remarkable that it made sense to reformulate it into an even stronger hierarchical proposition, H2, mirroring the logic of a perfect simple order scale. The establishment of such a scale, which can be understood as an attempt to scale democracy in a very broad sense, is interesting in itself as it allows for more sophisticated analyses of the causes and consequences of the new sequencing. But we first and foremost included it to refine and further underpin the empirical conclusion: in today's developing world, civil rights seldom precede political rights, thereby turning upside down the historical sequence of Western Europe. Constant across time, however, social rights generally make up the last part of the sequence.
Finally, we attempted to trace the part of the sequence that differs between H0 and H1 back in time. The cross-temporal analysis indicated that the departure from Marshall's itinerary appears to be a rather novel development, which kicked in a mere generation ago. That would still place it before the onset of liberal hegemony, at least if this is taken to begin only with the fall of communism. This is another interesting observation that calls for further research.
Until such research has been carried out, it is difficult to say much about which itinerary leads to the thicker combinations of citizenship rights (or democracy) over time. Yet one conclusion suggests itself. In the introduction we mentioned that the proliferation of thinner types of democracy has led some scholars to make the descriptive argument that the third wave of democratization is intrinsically hollow. But that is not all. Some have gone further and offered the prescriptive argument that the developed world should use its muscles to revert the sequence to the historical pattern known from the first world. In particular, Zakaria has argued that the democracy promotion should place emphasis on what he terms constitutional liberalism, relegating elections to a later stage where the rudiments of this constitutional infrastructure are in place.Footnote 30
Our analysis shows that Zakaria's preferred route to liberal democracy – the purely constitutionalist route or ‘authoritarian transition’– does not really exist in the developing world today.Footnote 31 Needless to say, this does not in itself prove that this way is closed. That conclusion would rest on a non sequitur as the case may simply be that Zakaria's route has not yet been tried out. But it is relatively plausible that Zakaria's path is indeed strewn with obstacles as a number of theoretical arguments can be wielded against it. Carothers has thus convincingly argued that the development of civil liberties and the rule of law is practically impossible in the absence of some kind of electoral competition. Zakaria therefore commits what Carothers terms a ‘sequencing fallacy’.Footnote 32
More generally, the historical advent of liberal constitutionalism has required either some form of external pressure, convincing illiberal elites that it is necessary to strengthen the centripetal power and constitutional infrastructure of the state or some kind of elections to check the power holders. The ‘law-abiding autocrat’ is virtually an oxymoron and it is very difficult to imagine a country moving towards constitutional liberalism in the absence of such pressures.
In the world of today, where the international society has largely eliminated the geopolitical anarchy known from the Western past,Footnote 33 electoral competition thus seems to be a prerequisite for moving toward both civil liberties and, arguably, social rights. That theoretical lesson is supported by the empirical analysis presented in this article. However, we do not feel sanguine about the extent to which this itinerary is in fact open to countries with inauspicious structural constraints. We simply note that the opposite itinerary seems solidly closed.
Appendix
Ordering of the Countries, 2003
Ordering of the Countries, 2005
