A Curvilinear Relationship between Democracy and Administrative Capacity
We start out by presenting a scatterplot describing the variation in democracy and administrative capacity across countries in 2002, presented in Figure 2. This plot suggests that that the two traits are associated in a curvilinear manner. Countries such as Saudi Arabia (SAU) and Cuba (CUB) score very low on the democracy scale (0.5–1), but display a fairly high administrative capacity (about 4), whereas other countries, such as Liberia (LBR) and Haiti (HTI), with slightly higher democracy levels (3–4), feature a very low administrative performance (1–2). The highest levels in both regards are for the most part found not only in Western Europe (e.g., Finland (FIN) and the Netherlands (NLD)), but also in New Zealand (NZL). An extreme outlier is the case of Singapore (SGP), with a relatively low democracy score (4.08) and a high administrative score (8.75).
Figure 2. Democracy and Administrative Capacity in 2002 ICRG, International Country Risk Guide; FH, Freedom House; FIN, Finland; NZL, New Zealand; NLD, the Netherlands; SGP, Singapore; CUB, Cuba; SAU, Saudi Arabia; LBR, Liberia; HT, Haiti.
Download figure to PowerPoint
Table 1 contains two types of regression analyses, where the first three columns (models 1–3) present the results from the pooled time-series cross-section analyses, and where the last three columns (4–6) present the results from the country fixed-effects analyses. In model 1 we include our democracy measure and the three control variables, GDP, trade openness, and British colony. All variables, except trade openness, exert significant effects in the expected direction. Thus, administrative capacity is higher the higher the level of democracy and GDP per capita and in former British colonies.
Table 1. Democracy and Administrative Capacity (Regression Analyses)
| ||Administrative Capacity Pooled Time-Series Cross-Section||Administrative Capacity Fixed Effects Regression|
| Log GDP/capita||0.938***||0.731***||0.040**||0.469***||0.276*||−0.248***|
| Trade openness||−0.032||−0.025||−0.020||0.275*||0.298*||0.004|
| British colony||0.433***||0.438***||0.072***||–||–||–|
|Lagged dependent variable|
| Administrative capacity at t − 1||–||–||0.927***||–||–||0.885***|
|Number of observations||2,115||2,115||1,998||2,115||2,115||1,998|
|Number of countries||127||127||127||127||127||127|
In model 2 we add the squared democracy variable in order to gauge the existence of a curvilinear relationship. In this model, the democracy variable exerts a significant negative effect on administrative capacity, and the squared democracy variable exerts a significant positive effect. The inclusion of the latter variable also increases the adjusted R2 by three percentage points. In all, the outcome of this regression attests to the curvilinear character of the relationship; it holds, as we can see, under control for other explanatory conditions of importance.
In model 3, we add the lagged dependent variable to our model. In this model, thus, we control for administrative capacity at t – 1. The coefficients for our democracy variables are clearly smaller in this model, but remain statistically significant. In models 4 to 6 we present the results from regressions with country fixed effects. In these specifications, accordingly, only the variation across time is taken into account. For the democracy variables, the results are quite similar to the results found in the pooled TSCS analysis. This adds to the robustness of our findings.
In order to find out at what point democracy has a negative effect and at what point it has a positive effect on administrative performance, we have calculated the predicted effects at different values of the democracy variable, presented in Table 2. When the level of democracy is 0, the predicted effect of democracy on the administrative score is –0.055. This negative effect is significant at the 0.01 level, suggesting that a democratization of countries that are extremely authoritarian will lower the administrative quality. Looking at countries that are somewhat democratized, with democracy scores between three and four, the effect of further democratization does not significantly affect administrative capacity, whereas countries with democracy scores higher than five are expected to experience an improvement in the administrative capacity when they are further democratized. At these democracy scores, namely, the effect of democracy is positive and significant.
Table 2. Predicted Effects on Administrative Capacity at Different Levels of Democracy
|Level of Democracy||Predicted Effect||95% Confidence Interval|
| 0||−0.055***||–0.095 to –0.015|
| 1||−0.042**||−0.074 to −0.009|
| 2||−0.029**||−0.054 to −0.003|
| 3||−0.016||−0.035 to 0.004|
| 4||−0.002||−0.017 to 0.012|
| 5||0.011||−0.003 to 0.024|
| 6||0.024***||0.007 to 0.040|
| 7||0.037***||0.015 to 0.059|
| 8||0.050***||0.022 to 0.078|
| 9||0.063***||0.028 to 0.099|
|10||0.076***||0.034 to 0.119|
Figure 3 illustrates the shape of the curve. The democracy measure has here been plotted against the predicted values from a model including democracy and democracy squared as predictors of administrative capacity. The relationship is clearly J-shaped, with a negative effect of democracy at low levels of democracy and a positive effect at fairly high democracy levels.
A way to picture the fit of our regression model is to graph actual levels of administrative performance against the predicted values. In Figure 4, we plot the predictions from model 2 for 2002. Figure 4 illustrates that the fit of this model is fairly good, with countries lining up on the diagonal.11 Thus, countries that our model predicts will perform well also do so. There are, however, exceptions to the rule. Some countries, like Italy (ITA), display a lower administrative score than would be expected according to their level of democracy and GDP per capita, and other countries, like Singapore (SGP), have higher administrative scores, than would be expected from their democratic and economic situation.
Two Methods of Steering and Control
Thus, a curvilinear relationship obtains between democratization and administrative capacity. As we see it, the curvilinear relationship results from the combined effect of two forms of steering and control that can be applied in public life. One is exercised from above, the other from below.
For efficient steering and control from above, the following instruments are needed:
centralized policy making,
top-down modes of implementation, and
statutory organs of control.
The direction and subsequent review of the public apparatus is easier if the more important policy decisions, together with their follow-up, are centralized to a single main arena. Coordination is enhanced and decision costs reduced thereby. Centralization, in turn, is facilitated if critical resources are controlled within one and the same decision making arena. This is particularly true with respect to tax revenues and the earnings from business activities in which the state is involved. In addition, the model involves top-down modes of implementation. Such an order is hierarchical: Powers of initiative and procedures for reporting to superiors are laid down from above. Alongside an internal administrative control in a variety of areas, there are special central agencies, such as auditing agencies, charged with the review of operations from without.
In all functioning states, instruments for steering and control from above are prominently present. If there were no such instruments, there would be no functioning state. State-ness presumes a substantial centralization of public activities. In the matter of state-ness there is, one might say, a zero-sum game: The central state institutions either exercise sovereignty within their geographical territory, or they do not. Sovereign state organs possess a basic capacity of supremacy, that is, they are able to hold competing power centers at bay (Krasner 1999).
Yet, the means that make state-ness possible are not always those that promote administrative capacity. As critics of the centralized model have noted, such a model has serious drawbacks at times—especially when it comes to the implementation and follow-up of various public measures. It is not always easy to get the arm of the state to reach out and to produce the intended effect in the field. The difficulty consists in being able (1) to adapt the measures taken to the conditions prevailing in different places and (2) to gather sufficient feedback on how the measures in question have actually worked. In other words, there is both an adaptation problem and an information problem. The cooperation of various actors in society can help to ease difficulties of this kind (Fox 2004; Ostrom 1990). It is such cooperation on the part of those affected that is the essence of what we here call steering and control from below. This form of direction and review is facilitated by the use of the following instruments:
divided/decentralized policy making,
implementation through cooperation, and
public review, through transparency and public participation.
Cooperation on the part of the social groups affected is easier if the public apparatus is divided both vertically and horizontally—so that decentralization and checks and balances characterize every level. Opportunities for access and for public oversight are increased thereby. But for such things to be achieved, control over critical resources must be dispersed in a similar fashion. Furthermore, implementation must be accomplished to a great extent through cooperation with the groups most heavily affected. It is a matter of working out appropriate measures through dialogue and of inducing those affected to participate as partners in the practical implementation of said measures. This can make policy adaptation easier. The object, furthermore, is to involve different groups of citizens in the process of review and assessment: to get them to work as inspectors and whistle-blowers. The advantage of such an arrangement is that those who are close to the operations of a given public program tend to have a good sense of how well it has worked. This “social auditing” can take many different forms. At a general level, the aim is to create arenas and channels wherein popular reactions to public programs can come to expression. Such involvement by the stakeholders could make up for the well-known inability on the part of central organs to effectively direct and review public activities out in the field (Ackerman 2003; Fukuyama 2004; Tendler 1997).
Such direction and review cannot replace the top-down kind; rather, it furnishes a complement. In the best case a synergy between the two forms can be created: The direction is applied in dialogue and cooperation, while the more professional type of statutory auditing is combined with a more popular and social form. It is when an interaction of this kind can be created that the highest level of administrative capacity can be reached. Thus, interaction between state and society is to the benefit of both (Evans 2005; Hadenius 2003; Lange 2005; Wang 1999). In contrast to the case with basal state-ness, we here have a positive-sum game: The state's bureaucratic capacity can be enhanced if other actors in society are invited to take part in directing and reviewing the actions of the administrative apparatus.
Such positive interaction between state and society can only be created, as a rule, in certain institutionally and organizationally developed democracies. These states not only have well-institutionalized organs for steering and control from above, but they are also able to make use of functioning channels for societal participation in the policy process (Hadenius 2001). It is for this reason the administrative capacity is highest among developed democracies. The establishing of well-functioning democratic institutions could involve several advantages with respect to administrative performance. Factors like political competition and freedom of assembly and the press further transparency and accountability. This makes it easier to detect and to actively combat abuses in the public agencies.
At the opposite end—in an authoritarian setting—a satisfactory measure of direction and review from above can at times be accomplished, especially if the state is able to draw on a flourishing resource base and to take harsh measures against low performance. Such advantages can make up, to a great extent, for the meager measure found in such societies of steering and control from below. The monarchies of the Middle East would seem to be successful examples hereof. With access to vast and highly profitable natural resources, these states have been able to establish not just a large-scale repressive apparatus, but a fairly well-functioning bureaucratic system as well (Bellin 2004; Herb 1999; Ross 2001). The administrative performance of these states does not reach top-notch level, but it stands out as relatively good in comparative perspective.
When authoritarian structures are relaxed, however, administrative capacity tends to decline. The reason for the downturn is that centralized forms of management, which could be used before, are not so easily maintained when the system is loosened up. Changes of the latter sort normally involve, after all, the appearance of numerous new actors on the scene, making control more difficult. At the same time, the most brutal forms of retribution can no longer be employed, weakening the forceful arm of the state. But this weakness is not made up for—at least not to a sufficient extent—by the development of institutions and social networks able to serve as vehicles for steering and review from below. At this stage, accordingly, none of the instruments that can do the job are present in a satisfactory degree.
When a state has just begun its democratization, then many of the institutional requisites for democracy are still lacking. Competition between parties, for example, is still highly restricted as a rule. The governing side may still be greatly favored in respect of economic and organizational resources, and it may yet be able to influence the electoral process to its own advantage. Freedom of assembly and of the press may also be highly restricted. At this stage, the effects conducive to state capacity that democratization can yield do not appear. The administrative capacity accordingly falls. As can be seen in Figure 3, it reaches its lowest point at about level 4 on the scale of democracy. After that it moves sharply upward. It is first when the level of democracy is between 7 and 8, however, that we find an administrative quality clearly exceeding that of the strongly authoritarian states. It is a level of around 7.5 that is often taken as the categorical boundary between democracy and its absence (Hadenius and Teorell 2005b). It is first above this level, among the full-fledged democracies, that the real profits of democratization appear.
From about the middle of the scale, then, higher levels of democracy bring gradually with them a higher administrative quality. But it is first when a high level of democracy has been attained that such political changes tend to bring their reward in the administrative realm. However, even relatively well-established democracies sometimes display a mediocre and sometimes declining administrative performance. This is true to a great extent in Latin America (Argentina and Brazil are striking examples). Such differences in performance reflect differences in the quality of democracy above and beyond its purely institutional requisites. To function properly, democratic institutions must not only be accompanied by a broad array of political resources in society, such as individual resources relating to knowledge and political activation, but also collective resources, such as the development of a vital press, and of well-rooted parties and organizations, able to operate as tools of communication and representation. In many countries that have undergone democratization, the development of such features has been uneven.12
Thus, we suspect that the picture becomes clearer if we consider not only the development of democratic institutions, but also the development of various political resources. In Table 3 we present some analyses aimed at evaluating the proposed explanations of the curvilinear relationship documented above. In model 1 we present a base model, including the democracy variables and some control variables.13 In the following model we also include a measure of political activity, namely, electoral participation, using data provided by Vanhanen (2000). This of course is a less than perfect measure of political activity, since it is restricted to electoral activity and thus excludes other forms of political participation. Yet, this is the only measure that could provide a reasonable cross-country and cross-time coverage. In this model we also add a variable tapping the daily circulation of newspapers per capita, drawing on Banks (1995).14
Table 3. Democracy and Administrative Capacity—Explaining Curvilinearity (Regression Analyses)
| ||Administrative Capacity (TSCS Data)All Observations||Adm. Capacity (TSCS) Only Democracies|
| Newspaper circulation||–||0.274***||–0.625***||–||0.326***|
| Newspaper circulation × democracy||–||–||0.104***||–||–|
| Log GDP/capita||0.800***||0.666***||0.768***||1.126***||0.885***|
| British colony||0.436***||0.506***||0.603***||0.670***||0.805***|
|Number of observations||1912||1912||1912||1240||1240|
|Number of countries||124||124||124||105||105|
The newspaper circulation variable exerts a positive and significant effect in model 2, suggesting that countries with a high level of circulation of newspapers also are marked by higher administrative scores. This could be due to the fact that it is easier to oust bad officeholders in such countries, since the public becomes more aware of bad administrative practices through their media usage. The effect of the electoral participation variable is positive as expected, but not significant. At the same time, in model 2, the two democracy variables still hold their own; the coefficients are only slightly reduced when the resource variables are introduced. This implies that the effect of the resource variables could only be complementary to the effect of the democratic institutions.
To further evaluate this effect, we have constructed a variable that represents the interaction between newspaper circulation and the level of democracy. The idea is that the relationship between democracy and governance is conditional: When democracy is associated with the presence of certain capacities in society, this adds to the quality of administrative practices. More concretely, democratization should have a greater impact when there are favorable media conditions, enabling the citizenry to obtain sufficient information about the public activities, and to hold the persons in charge accountable. As can be seen in Table 3, the combined democracy/newspaper circulation variable has an evident individual impact. What is telling, moreover, is the effect on the two democracy variables. For both of them, the coefficients are reduced considerably. This implies that the influence of democracy on administrative performance is due to an important degree to an interaction between democratization and improvements in the field of newspaper circulation.
In the remaining models (4–5) we concentrate on the upturn-side of the J-curve. As it is here a matter of a mainly linear relationship between democracy and governance, we only apply the original democracy variable. Moreover, only observations higher than four on the democracy index become relevant in this analysis.15 Using this sample, we present a new base model (4). In model 5 we add the intervening variables. In this test the variables measuring electoral participation and newspaper circulation both exert significant effects, suggesting that countries that have a high level of electoral participation and a high level of newspaper circulation perform better administratively. As before, the democracy variable is strongly significant, while the coefficient is slightly reduced. This illustrates, again, that there is no clear-cut trade-off between democracy development and resource development. For administrative improvement, these features should be developed in conjunction.