8 (high violence but moderate restrictions). These very different contexts

represent extremely different scenarios within which democracy is expected

to have an impact. To address this situation, I consider in Table 4.1 the most

114

Democratic Paci¬cation

likely value of current repression, at time t (in the column), given a speci¬c

value of previous repression at time t ’ 1, and Suffrage set to 0 (in the row).

This is compared with the most likely value of current repression, given a

speci¬c value of previous repression at time t ’ 1 and Suffrage set to 100.

It is clear from these ¬ndings that electoral participation is fairly weak

at decreasing state repression. For example, the second column reveals that

in situations where there is no Suffrage at time t ’ 1, there is a natural drift

downward in repressive behavior within the following year. This acknowl-

edges that it is dif¬cult for states (even relatively authoritarian ones) to

sustain repressive behavior and that they reduce the magnitude of this appli-

cation soon after it is employed. Thus, when prior repression was at low

to mid-range values (levels 3 and 4), the next period repression is relatively

low (in category 2); when previous repression was in the middle to upper

range (categories 6 and 7), repression in the next year is likely moderate (in

category 5); and when prior repression was quite high (category 8), the next

year repression was likely to be closer to moderate values (in category 6).

Revealing the weak pacifying in¬‚uence, movement in Suffrage from its

lowest value (0) to its highest (100) in column 3 does very little to alter the

results. Indeed, only three categories of repression are reduced by such a

change: when previous repression was in categories 5 (moderate violence

and restriction) and 7 (high violence, low restriction), these are reduced to

category 2 (low violence, moderate restriction); and, when previous repres-

sion was in category 9 (high violence and restriction), this is reduced to

category 8 (high violence, moderate restriction). Suffrage does have an in¬‚u-

ence, therefore, but compared with what simply happens naturally over time

in the most autocratic of contexts, the magnitude of this in¬‚uence is quite

weak. In line with this, it should come as no surprise that during a period

when Suffrage was particularly high in El Salvador (the legislative elections

of 1976 and constitutional assembly of 1982), observers noted a consistent

application of political assassinations, disappearances, and state prosecution

of citizens. This led to numerous individuals referring to “demonstration”

elections being nothing more than a show for foreigners.

Of course, other aspects of Voice may be relevant to repressive lethality.

An additional measurement is considered later.

Competition and Participation

To investigate the impact of electoral competition and participation on

repression, I use the variables collected by Vanhanen (2000) and modi¬ed

115

State Repression and Domestic Democratic Peace

Competition/Participation

0.4

Change in Probability of Repression

0.3

0.2

0.1

0

1 2 3 4 5 6 7 8 9

-0.1

-0.2

-0.3

Categories of Repression

Figure 4.2 The In¬‚uence of Competition/Participation on Repressive Lethality

in Gates et al. (2003); this is referred to hereafter as the Vanhanen/Gates

democracy measure, or Competition/Participation.3 As with Suffrage, I

include this variable in the basic model provided earlier. This model as

well as some ¬t statistics are provided in Table 1, column 2 of Appendix 2.

Once again, the likelihood ratio statistics and their corresponding

p-values suggest that the model including Competition/Participation is better

than the model excluding it (i.e., the base model). Similar to the Suffrage

model, Hypothesis 1 is again supported, and Voice is deemed statistically sig-

ni¬cant in determining the form repression takes. Indeed, ¬ndings reveal

that, across categories of repression, the in¬‚uence of this variable is sev-

eral times larger than Suffrage as well as most variables in the basic model.

Figure 4.2 presents the predicted probabilities.

From the results, Competition/Participation exhibits an in¬‚uence that is

generally similar to that identi¬ed with Suffrage, but with some important

differences. First, only repressive categories 1 and 2 are in¬‚uenced posi-

tively in this analysis whereas the others (3“9) are decreased, supporting

the weak variant of the domestic democratic peace (Hypothesis 1). Second,

when one increases Competition/Participation from its minimum to its maxi-

mum, the magnitude of change in predicted probabilities is much higher

3 Recall that this measure privileges the competition component but hedges it by the propor-

tion of the population enfranchised (over 30 percent).

116

Democratic Paci¬cation

than that identi¬ed when Suffrage is similarly altered. For example, in the

context of enhanced Competition/Participation one observes an increase of

approximately 26 percent in categories 1 and 2 repression (where violence

is low but restrictions are respectively low and moderate). This amounts

to an increase of approximately six times that provided by Suffrage. Third,

the in¬‚uence of Competition/Participation is negative in higher values of the

dependent variable as suggested by the domestic democratic peace, but

the magnitude of in¬‚uence declines as the lethality of repression increases

from category 5 to 9. Thus, the change in probability of achieving category

5 repression (where violence and restrictions are both applied at moderate

levels) is decreased by about 20 percent when Competition/Participation is

increased from its minimum to its maximum. When one considers repres-

sion where moderate values of violence but high values of restrictions are

found (category 6), the change in probability is 12 percent; when one con-

siders repression where violence is high but restrictions are moderate (cate-

gory 8), the change in probability is 5 percent. Finally, at the highest value

of repression (where both violence and restrictions are high), the change

in predicted probability is 1 percent, indicating a very limited in¬‚uence.

Voice thus paci¬es quite signi¬cantly, but the in¬‚uence is substantially more

important with regard to the least repressive techniques as well as with

regard to those that emphasize restrictions (supporting the argument that

democracy is better at diminishing violence than at diminishing restrictions

on civil liberties).

It is useful to consider the results in a different form (see Table 4.2).

When one considers the most likely value of current repression when Com-

petition/Participation is 0 and given a speci¬c value of previous repression,

the results show a more robust pacifying in¬‚uence.

For example, we see that when there is no Competition/Participation

(column 2), there is still a drift in repression in relatively autocratic gov-

ernments away from higher values in the previous year to lower values

in the subsequent period. Thus, when prior repression was in category 3

(where violence is low but restrictions are high), the next period repression

is likely in category 2 (where violence is low but restrictions are moderate);

when previous repression was in category 4 (where violence is moderate but

restrictions are low), in the next year repression is likely in category 2; when

prior repression was in categories 6 and 7 (where, respectively, violence is

moderate but restrictions are high and violence is high but restrictions

are low), in the next year repression is likely moderate. These results are

similar to those in the Suffrage model. Findings differ when one considers

117

State Repression and Domestic Democratic Peace

Table 4.2. The In¬‚uence of Competition/Participation on the Most Likely Current

Value of Repression

1 2 3

Most Likely Value of Current Most Likely Value of Current

Repression When Competition/ Repression When Competition/

Participation = 0 Participation = 100