HomeDocuments on SitePersonalDemocratic PeaceDemocide20th C. DemocideMegamurderersLesser MurderersWhy DemocideDimensionsConflictMethodsTheoryPolicyLinks PHOTOS OF DEMOCIDEGalleries

Other Democratic Peace Documents On This Site


What is the "democratic peace"?

"Waging denuclearization and social justice through democracy"

"The rule of law: towards eliminating war"

"Freedom of the press--A Way to Global Peace"

"Convocation Speech"

Freeman Interview

City Times Interview


Bibliography on Democracy and War

Q & A On Democracies Not Making War on Each Other

But What About...?

"The democratic peace: a new idea?"


"Libertarianism and International Violence"

"Libertarian Propositions on Violence Within and Between Nations: A Test Against Published Research Results"

"Democracies ARE less warlike than other regimes"


Vol. 2: The Conflict Helix (see Chapter 35)

Vol. 4: War, Power, Peace (see e.g., Propositions 16.11 and 16.27

Statistics of Democide

The Miracle That Is Freedom

Power Kills


By R.J. Rummel

In my Understanding Conflict and War I concluded that the more freedom a state accords its citizens, the less likely it is to be involved in foreign violence; and the more freedom within two states, the less likely there will be violence between them. Moreover, war between free--libertarian--states will not occur, and other violence between them is very improbable. These strong conclusions were based on the theoretical analysis and empirical work detailed in the aforementioned work and since subjected to additional and highly positive tests.2

This conclusion about the violence-reducing effect of freedom also extends to conflict within states: the more libertarian a state, the less intense its violence can and tends to become. The purpose of this paper is to explain this proposition, to test it against all violence within states for the five years 1976-1980, and to suggest a principle underlying all these and the aforementioned results.


As will be more precisely defined later, freedom within states means at a minimum the existence of basic civil liberties and political rights, or what is often called a liberal democracy, and at its fullest extension also includes economic freedom, a free market. In either case, I will term such freedom prescriptively or descriptively as libertarianism. As to why libertarianism should minimize intrastate violence, I give a detailed theoretical and empirical analysis elsewhere3 and will try to summarize and simplify its most relevant arguments here.

Perhaps I can best illustrate the ideas involved by way of two very simplified pure types. For one, consider a pure exchange society with a small, libertarian government. First, its general structures of expectations and their central status quo largely evolve from the private relationships among individuals, groups, and classes; government has little to do with their balancing of interests and expectations. What relevance government does have is by way of general laws that apply equally to all, guaranteeing basic civil liberties and political rights and trying only to minimize conflict between people pursuing their own goals. These laws dictate no particular outcome and benefit few particular interests. Such are highway laws, for example, which ease one's driving among other moving vehicles but favor no particular driver or destination or purpose (except emergency vehicles).

Second, those formal laws that the government does enforce are a small part of the overall structures of expectations and cover only a small number of behaviors.

Third, since every individual or group is free to establish its own equilibrium (and this is not an authoritative society with pervasive, compelling norms and customs), the prevailing social power is bargaining: the expectations that individuals and groups have of the outcome of their behavior is usually based on mutual exchange and adjustment. The free market operates socially and economically; a monetary and social price system allocates economic and social costs and benefits. The society constitutes a social field.

And fourth, the society is pluralistic. There is a diversity of interests, groups, and classes. Different interests are pushed and pulled in different directions by different obligations, benefits, and opportunities, the satisfaction of any one interest usually requiring the compromising or giving up of some others. One may be top dog in one group, in another the underdog. That is, one's class membership (as to whether in the command or obey class) may well differ from group to group and from organization to organization.

The outcome of all this is to cross-pressure interests and wills. As a participant in multiple structures of expectations, of multiple groups, and of multiple classes, individuals really are brokers among their different interests, deciding which to satisfy, which to ignore. It is then difficult to get excited about pushing any particular interest, for then one may lose out on some others. In other words, "win some, lose some." This perspective is the hallmark of an exchange society and its libertarian government. And it is this perspective generated by cross-pressures and pluralism in an exchange society that should minimize social violence. With diverse, overlapping groups, differing class memberships, and cross-pressured interests, a conflict front cannot easily form across the larger society. Without such a front, what violence that does occur is either localized and contained or is diverted into numerous channels and drained off before it involves many people, groups, or interests. Thus, spontaneous and largely nonviolent turmoil--a dimension of conflict4--is associated with exchange societies.

Next consider a totally different type of society, a coercive one in which a totalitarian government directs most activities. The political elite maintain their power mainly by coercion or force; all public news is censored and controlled; and there are no free or open elections. The society is organized by an elite acting through government in order to achieve some central goal or purpose, such as national glory, empire, economic development, or socialism. That is, the elite have turned the society into an organization (what I call an antifield); the government now encompasses, and is at the origin of, most structures of expectations, although some room will exist beyond the most absolute government control for individual interests and expectations (even in prison private friendships and relations form). Most groups are completely under government control and operation, and government dictates whether one is in the command or obey class within each group. Thus, authoritative and coercive powers in the society are centralized; the society is an organization.

These aspects of a coercive society have this significance for social violence. In a coercive society, most of what one does is controlled by government, by the same "they". There is one supreme elite officially commanding all others, regardless of the organization or group involved (compare this to the United States where not even the president can tell the editor of the Washington Post what to print). Although more or less effective in particular instances, the command is formally society-wide, over churches, farms, factories, schools; provinces, cities, and towns; families and individuals. The obey class, the mass of "outs", are always one-down.

The critical effect of this is to polarize class membership and interests. One central status quo, like a hierarchical organization chart that establishes who is subordinate to whom, determines one's overall rights and benefits: what one owns, eats, earns; and where one lives and works. This splits society, creating a latent conflict front cutting across groups and organizations and unifying classes in separate groups into one societal-wide, two-class division. Without numerous cross-pressured interests, one can no longer compromise and balance among them. For the polarization of society into one class system means that an individual's most important interests, even his life, is vitally affected by which side of the class front he is on. For this reason any development of leadership in the obey class is a most serious threat to the governing elite and may be harshly dealt with. And obey-class political sensitivity and consciousness that might dangerously weaken the legitimacy of the elite are avoided through thought control, brainwashing, and propaganda; selective isolation or elimination of dissidents; and jingoistic, nationalist campaigns that call for political unity and support against a foreign evil.

In general, the elite control and manage the ever present threat from the obey-class through violence. Executions, forced resettlement, mass imprisonment in concentration camps, work and internal travel controls, and an extensive spy system and secret police are weapons to maintain class order. Fear of such personal violence and sanctions is the lot of the obey-class.

But violence also exists within the command-class as well: possible contenders for the top are eliminated or purged, elite who are possible or actual threats to policies are jailed or killed, and attempts to grab top power by some elite are more or less successful, and usually bloody.

Even aside from this actual violence of the government against the people or itself (both of which can make coercive societies the most violent), there is the ever present potential for class war. The sharp command-obey class division across society is like a geological fault line across the earth. Pressure builds up on both sides, until the stress is such that, when the fault slips in one place, its whole length may be unlocked, producing a severe earthquake over a large area. Similarly, when violence occurs in a polarized society over one issue and in one place, it can trigger violence along the entire class front. Thus, the social earthquake: mass violence, revolution, and class war.

By way of summary, Table 1 contrasts the two pure types of societies. Put simply, in a free society individuals have different statuses, belong to different and independent organizations, are in one or the other class from one group to the next, and largely ignore government. Interests are pluralistic and cross-pressured. And if social violence occurs, it is usually isolated and drained off.

However, in coercive societies, who and what one is depends largely on one's class relationship to the government--to the elite. It depends on political power. This polarizes interests and makes issues dividing ins and outs critical to the lives of each. Any social violence is most likely a question of elite power and policies and can thus escalate into a class war. However, even were a class war prevented or suppressed, the elite must use violence and terror to maintain their near total control.

Of course, these two types of societies are pure types, although there are states, such as the U.S.S.R., China, North Korea, and Cambodia, which come close enough to deserve the coercive label, and some western states, such as the United States and Switzerland, which are still, in spite of the dominance of central government, in many ways (as in freedom of speech and civil liberties) exchange societies with libertarian governments. Most states, however, lie somewhere between these pure types or between them and the authoritative or traditional type. Regardless, however, of type or mixture of types, the less individual freedom and the more the satisfaction of interests become a government matter, the more a single division of ins and outs is created in society, the more authority or coercion determines balances of interests, and the more a pervasive status quo defines who gets what. The less freedom, therefore, the more absolutely vital the status quo and the more likely and extreme the social violence in protecting or changing it.

This is not at all to deny that other factors condition and affect violence. Freedom has a dual role. It sets an upper limit on violence, and it influences the intensity and scope of violence. But the actual kind and level of violence beneath that upper limit also depends on other aspects of society, such as whether it is dominantly coercive or authoritarian (traditional), whether it is undergoing rapid economic development and has reached mid-range transitional status, whether it is communally heterogeneous (different regional, ethnic, racial, religious, linguistic divisions), and whether development is unbalanced in different sectors, especially where educational development outpaces the economic.5 Elsewhere6 I surveyed the empirical literature to test these factors and showed that research on internal conflict generally supports their role, but in conjunction with individual freedom and within the limits that this freedom established.7 And it is this violence bounding and influencing function of freedom--libertarianism--that I intend to test here.

To be clear about what will be tested, precisely, and what therefore requires definition, the following proposition is stated

Freedom/Domestic Violence Proposition: Freedom inhibits domestic violence (that is, the more libertarian a state, the less internally violent it can and tends to become).

There are two concepts and one verbal phrase involved here whose meaning must be made explicit: "libertarian," "violence," and "can and tends to become." I will deal with these in order.


By theory, "libertarian" will have two meanings, one a limited version of the other. The first is political freedom, involving civil liberties and political rights, what we usually mean by a liberal democratic, open system. Second, there is freedom in a more expansive sense, which not only includes political freedom but also the freedom of groups and individuals to pursue their socioeconomic interests free from government coercion. The latter reflects the classical liberal idea of limited and minimal government. While political freedom is consistent with a large, democratic socialist government, as in Sweden or Denmark, I argue8 that such centralized, semisocialist governments introduce a considerable measure of coercion that contributes to internal violence. We should find, therefore, that those states with less freedom, in comparison to less political freedom, should have a greater tendency to violence.

Freedom is best measured by the degree to which governmental power is decentralized and limited and society is based on exchange. Here it suffices to measure the level of economic freedom (from governmental ownership, control, and regulation). Thus, the two definitions:

Political freedom = civil liberties + political rights;

Freedom = political freedom + economic freedom.

The Freedom House publication Freedom at Issue annually presents a seven-point scale of all states on their political rights and civil liberties.9. To paraphrase Freedom at Issue10 political rights are defined by an open, competitive electoral process through which leaders are clearly elected. States rated 1 have such; those rated the highest, 7, have none, in other words, those at the top believe they have the right to govern without challenge and are not constrained by public opinion or tradition. Civil liberties comprise the freedom of the press and the independence of the major media from government dictation; the protection of the individual by the courts; the freedom to express individual opinions without fear of imprisonment; the respect for private rights and desires in religion, occupation, residence, education, and the like; and the individual's ability to engage in rational political activities without fear for his life. States rated 1 have all these civil liberties; those rated 7, the highest, have complete censorship, political prisoners, no right of assembly, restricted travel, residence, and occupation, pervading fear tied in to a police-state environment, and swift and sure execution. I will operationalize political freedom, then, as the sum of these two scales, where the higher the combined scale value, the less the political freedom. Thus, for 1980, the U.S. is 1 + 1 = 2; Indonesia is 5 + 5 = 10; Kuwait is 6 + 4 = 10; and Vietnam is 7 + 7 = 14.11

Freedom at Issue also classifies states by their economic and political system. According to my estimate of the economic freedom under each subclassification, I distributed scale values 1-14 as shown in Table 2.

Freedom is then measured as the scale values for political freedom plus those for economic freedom, which gives equal weight to political and economic freedom. Thus, freedom for Belgium in 1977 equals 2 + 1 = 3; for Sweden, 2 + 9 = 11; for Senegal, 8 + 9 = 17; for Poland, 11 + 9 = 20; and for East Germany, 14 + 14 = 28.

I also need to group states and dyads by their degree of libertarianism and to keep distinct the two definitions of a libertarian state. The most objective way of doing this is to divide the political freedom scale into politically free, 2-4; partially politically free, 5-11; non-politically free, 12-14; and to divide the freedom scale into free, 3-7; partially free, 7-21; and non-free, 22-28. The resulting types are shown in Table 3.

These scales and types are now basic data on libertarianism for testing the proposition.


The second concept of concern is collective violence. Theoretically, collective violence is the wounding or killing of individuals or the destruction of property as a direct and immediate outcome or manifestation of group (social) or political conflict, or of government policies or actions. This would include, for example, food or race riots, the assassination of a religious leader, government purges involving executions, widespread government executions associated with the use of terror and repression, antigovernment terrorism, guerrilla war, and revolts. Criminal violence is excluded. The basic idea is that internal collective violence theoretically involves blood, death, or destruction along four conceptual axes:

Although there are some differences in the literature, factor analyses of internal conflict event data for states generally has delineated three empirical dimensions: turmoil, revolution, and subversion. Using these results and my own experience in coding and analyzing such event data, I first selected a set of indicators of violence that spans these three dimensions and thus captures the major empirical variance in internal violence across states, and second I added indicators where necessary to be sure that the four conceptual axes were also covered. The result was thirteen indicators: assassinations, political kidnapings, riots, political clashes, nonpolitical clashes, terrorism, bombings, private wars, guerrilla wars, revolts, rebellions, civil wars, and number killed.

In application to event data each indicator, except killings, is scaled from 0 (no violence) to 8. This scale is a conceptual product of two latent components: intensity and the goal/effect on government of the violence. These are defined in the upper part of Table 4; the lower part shows the actual coding scale values of event data on an indicator, scaled as a product of the two components.13

The various indicators, scaled as above, should provide a good measure of whether libertarianism lessens the scope and intensity of sociopolitical violence and in the aggregate will test the internal freedom proposition. But how can these indicators be aggregated to clearly perform the test? There are different ways of doing this through various scaling procedures, such as Guttman scaling or factor analysis (by summing factor scores across the empirical dimensions of the indicators--on factor analysis, see "Understanding Factor Analysis"). They all suffer from one critical defect, and that is they weight and sum by covariance or coincidence. As a result, riots may be given as much or more weight than rebellion or civil war, or many political kidnapings as much weight as half-a-million killed. While such aggregate scaling may give an aura of objectivity, it can confound what it is that should be defined: the scope and intensity of violence.

Accordingly, in order to best (from a theoretical and testing perspective) aggregate the indicators, they will be collapsed onto one scale according to the scaling transformation chart shown in Figure 1. The scale values of each indicator were aligned from left to right in the chart according to the underlying intensity and scope of violence it reflects, in comparison to the scale values of the other indicators. Thus, it can be seen that the scale values of the assassination indicator usually lie to the left of those for the other indicators, and for the killed indicator scale values of 8 and 9 (over 140,626 killed) surpass the highest scale value of any other indicator. The alignment of scale values is judgmental, reflecting my assessment of which measures more or less violence in a state, where violence is as understood in the theory and defined here.

With this chart, then, one converts from the scale values for an indicator to the final aggregate internal violence scale as follows.

  1. Determine the profile of scale values across the thirteen indicators for a state for a year. (For example, Argentina's nonzero profile for 1976 is: assassinations = 8, political kidnapping = 2, terrorism = 4, bombing = 2, guerrilla war = 4, killed = 5.)

  2. Of all scale values for all indicators, isolate the one furthest to the right on the transformation chart.

  3. Define for this one value the corresponding (vertical) Internal Conflict Scale value. This is then the measure of internal violence for a state for a year. (For example, consider the above scale profile of Argentina. The indicator for Argentina whose scale value 5 is furthest to the right on the chart of Figure 1 is killed, and the 5 is vertically aligned with 21 on the internal violence scale. Number 21 then becomes the aggregate measure of Argentina's internal violence in 1976.) These scale values are the aggregate violence data I will use to test the proposition.

Before moving on to data sources, there are a number of characteristics of the indicators and their scaling to note.

  • The indicators are of violent collective conflict only. Those that would reflect only nonviolent conflict, such as demonstrations, government crises, nonviolent coups, plots, and general strikes, are excluded. The reason is that the proposition refers to violence, and the occurrence of violence crosses a theoretical threshold: libertarian states actually may have more internal nonviolent conflict than other types of states (nonviolent conflict is the noise of a diverse, exchange society as individuals and groups spontaneously pursue their interests) while their collective violence should be less.

  • The scaling for each indicator recognizes that not all riots, or terrorism, or revolts, etc., are alike, and each should not be counted simply as one and summed, as is often done in event data research. Rather, the scaling clearly discriminates between, say, a minor riot in a provincial town and interconnected riots in the capital involving 10,000 people and threatening the downfall of a government, or between a minor peasant revolt easily contained and one that could cause a revolutionary change in the political system.

  • The transformation chart produces one ordinal scale measuring the scope and intensity of violence that clearly discriminates between states with minor, moderate, and extreme sociopolitical violence, as those adjectives are usually understood. The Internal Conflict Scale should avoid the distortion that would have been created by adding weighted or unweighted event data counts across the indicators, where many small riots could give a state a higher score than, say, a severe guerrilla war.

  • The Internal Conflict Scale helps to overcome data bias and unavailability. Data are usually most available for developed, libertarian states and for those of particular interest to the West. Thus, low level violence may easily be missed for non-Western, remote or nonlibertarian, or "unimportant" states. As violence becomes more intense, however, it is more likely to be reported, even for an Uganda, Cambodia, or North Yemen. For example, the mass killing of hundreds of thousands and perhaps millions of Cambodians by the Khmer Rouge was eventually reported, while, were several riots, bombings, or small revolts to have occurred, they probably would have gone unknown outside the country. To sum weighted or unweighted counts or scale values across indicators therefore biases the resulting scale downward for states that are closed (such as North Korea), remote (such as Togo), and "uninteresting" (such as Mali). The violence scale helps avoid this, since the more intense and greater in scope the violence, the more likely it will be reported; and one large rebellion, say in Afghanistan, is sufficient to place Afghanistan higher on the scale than states with a number of lower level events.


Finally, what does it mean to state "can and tends to become" in the domestic violence proposition? Consider Figure 2. One fundamental idea is that the degree of libertarianism in a state sets an upper bound on the scope and intensity that violence can reach, as shown by possible bounds A, B, or C. Let B, for example, be the upper bound. Then, violence may be less than this; it may be zero. But when violence does occur, the less libertarian the state, the greater in scope and more intense it can be.

Moreover, the larger the range of violence--the higher the upper bound in Figure 2--, the more intense the actual violence in states is likely to be. By theory we would not expect that when violence occurs it would always jump to the maximum possible level; violence could either escalate to that level or be resolved in some fashion at lower levels. All those authoritarian states that could be torn apart by revolution may have only minor revolts, terrorism, or guerrilla war. This is to say that empirical cases of violence should be spread throughout the space of violence shown in Figure 2, but tend to be higher, the less libertarian, as illustrated by lines a, b, and c in Figure 2. (The reason alternative lines are shown here and for the upper bound is because the theory specifies only an increasing function, but not the type of function, which for example might be logarithmic, exponential, or linear). However, because of the greater range of violence as libertarianism decreases, the empirical correlation between violence and libertarianism should not be very high (which has been the case in the empirical literature--on the nature and meaning of statistical correlation, see Understanding Correlation); but because of the theoretical tendency shown by lines a, b, and c, the correlation must be significantly positive between lack of libertarianism and violence. All this is to say that with decreasing freedom the central tendency of sociopolitical violence is upward.

Incidently, if Figure 2 does describe the actual empirical relationship between libertarianism and violence (and that it does will be shown subsequently), it makes clear why many empirical studies have missed the underlying relationship between freedom and violence. A correlation measures a necessary and sufficient relationship between two variables (see Understanding Correlation). That is, a perfect linear correlation would be along line a in Figure 2, and cases off the line, such as those shown in the space of violence, subtract from that correlation. However, by theory as libertarianism decreases, violence will fan out from such a line. Indeed, the best theoretical plot of libertarianism and violence should be a right triangle and not a regression line. Therefore, correlational analysis involving measures of domestic violence and liberal democracy have turned up only low or, at best, moderate correlations. And without theory to illuminate the empirical results, the essential relationship specified in the Domestic Violence Proposition had been missed, at least by me.15


In 1976 1 began to collect daily all the data used here and in my related report.16 As I read the newspapers, news and regional magazines, journals, and other sources, I clipped and filed relevant news and conflict items and articles. Then in early 1981 I went through these files to code the thirteen indicators of violence.

The data base is diverse, involving the local newspapers, The Wall Street Journal, and The New York Times, as well as the weekly U.S. News and World Report, Time, and Newsweek. Moreover, I used throughout the daily Department of Defense morning and afternoon editions of "Current News," a collection of current clippings and articles on defense and relevant conflicts17 from all the major news media, magazines, and journals.

Data were coded for all legally sovereign and independent states for each of the five years from January 1, 1976, to December 31, 1980. These data were then reduced to the scaled indicators and finally to the Internal Conflict Scale. The sample size is 261 (states times years). This will constitute the Full Sample for the tests.

Now, some states, like Angola, Cambodia, Laos, and Lebanon, have much violence for all five years and thus would weight--perhaps bias--the results. To avoid this, a second sample will consist only of that year of maximum violence for each state, 1976-1980, in order to gauge the relationship between a state's libertarianism and its highest violence reached during the five years. Moreover, to make the sample more conservative, if a state's violence scale values are tied for a number of years, that year with the lowest--more libertarian--political freedom and freedom score will be selected (this will bias against the proposition). This sample I will call the Max Violence sample. It consists of ninety-four states.


By nature of the domestic violence proposition, there are two hypotheses to be tested, both of which must be accepted before the proposition itself can be empirically endorsed. The first hypothesis, along with its null hypothesis (that which will actually be tested), is:

Hypothesis 1: The more libertarian a state, the less internally violent it tends to become.

Null Hypothesis 1: The more libertarian a state, the less internally violent it does not tend to become (i.e., violence is either positively related to or independent of libertarianism).

Now for the tests.

Null Hypothesis 1 does not specify whether the relationship between libertarianism and violence is linear or curvilinear, only that it is positively contingent or independent. A contingency test is therefore appropriate, as shown in Table 5 for the previously defined political freedom and freedom types and the Full Sample.

Were Null Hypothesis 1 correct, then the contingencies should either be randomly spread throughout the Table 5 or tend to be much greater than the expected values in the upper right portion. Actually, as can be seen, the contingencies for violence tend to move toward the lower right. Of course, this could be due to chance, but the chi-square (2) test for both the political freedom and freedom contingencies shows that it is highly unlikely--over ten-billion to one--to get accidently this kind of distribution were Null Hypothesis 1 correct. Therefore, on these data Null Hypothesis 1 should be rejected and Hypothesis 1 accepted for both political freedom and freedom that the more libertarian a state, the less violent it tends to become.

The significant results, however, may be due to overloading the data with states that are high in violence for several years but that just happen to be nonlibertarian. To test this possibility, Table 5 provides the contingencies and chi-square results for the Max Sample. Here also the contingencies significantly depart from Null Hypothesis 1 (although barely so for freedom) and Hypothesis 1 should be accepted.

By theory, freedom should do significantly better than political freedom in rejecting Null Hypothesis 1. But, while in fact also significant, freedom does no better. This will be discussed subsequently.

A different but appropriate way of testing Null Hypothesis 1 is by finding the best fitting curve to the data (the theory specifies no specific type of curve). If Null Hypothesis 1 is correct, the curve of violence as a function of political freedom or freedom should slope downward with decreasing libertarianism or have a nonsignificant coefficient of determination.

For the Full Sample for political freedom and freedom, the best fitting equations (of power, logarithmic, exponential, and linear types) are in fact linear.

y = 5.3 + 1.1x1
R2 = .17 p <.2 x 10-12

y = 8 + .5x2

R2 = .1 p <.5 x 10-7

y = Internal Violence Scale Value
x1 = political freedom score
x2 = freedom score
R2 = coefficient of determination17
N = 261

The small coefficients of determination for the equations are to be expected, since by theory much variance is likely beneath the upper bound established by libertarianism, and this variance should increase with decreasing libertarianism. The question concerns only the direction (slope) of the equations, which actually is inconsistent with Null Hypothesis 1, and their tendency (coefficient of determination), which for both is highly significant. Accordingly, for these tests also, we should accept Hypothesis 1 that the more libertarian a state, the less violent it tends to become.

For the Max Sample, the best fitting (of power, logarithmic, exponential, and linear) equations are:

y = 5.7 + x1
R2 = .15 p <.000003

y = - 1.2 + 6.2nx2
R2 = .09 p <.002

n = natural logarithm
N = 94

Again, only the significance of these equations and their direction is pertinent, and, as can be seen, they also strongly favor accepting Hypothesis 1.

Figure 3a shows the Full Sample plots of the political freedom and freedom equations; Figure 3b does this for the Max Sample. (The freedom scores were transformed, as will be described in the next section, to enable their equation estimates to be plotted in the same space with political freedom; the equations, however, are for the Full Sample.)

In sum, the two kinds of tests (contingencies and curve-fitting) agree that the probability of Null Hypothesis 1 being true is well below all commonly used thresholds of statistical significance. Therefore, Hypothesis 1 should be accepted for these data and measurements.


The proposition asserts two things: one is about the tendency of violence, which is Hypothesis 1; the other is about violence having an upper bound. The second hypothesis therefore concerns this upper bound and, along with its null hypothesis, is as follows.

Hypothesis 2: The more libertarian states are, the less their peak violence.

Null Hypothesis 2: The more libertarian states are, the less their peak violence is not (i.e., peak violence is either positively related to or independent of libertarianism).

The tests follow.

Possible upper bounds are illustrated in Figure 2. In order to test for such upper bounds here, two procedures were followed. First, the largest Internal Violence Scale value was selected from the Full Sample for each political freedom score. To best compare the results of analyzing these peak values to similar results for freedom, transformed scale values18 were used for the latter, since the scale is twice as long as for political freedom.

Second, power, logarithmic, exponential, and linear equations were calculated for these peak violence scores. Among these, the best fitting equations are:

y = 11.3 + 1.9x1
R2 = .85 p <.000004

y = 8.4x26
R2 = .73 p <.0001

y = peak internal violence scale values
x1political freedom scores
x2 = transformed freedom scores
R2 = coefficient of determination
N = 13

Two aspects of these equations should be noted. First, they both are highly significant. And second, as should be the case here since we are testing for an upper bound, the variance accounted for by each equation is high: 85 and 73 percent, respectively. Therefore, we should accept Hypothesis 2, that the more libertarian states are, the less their peak violence.

Figure 3c plots together the peak violence equations for political freedom and freedom. As can be seen, except for those states that are quite libertarian, the lack of economic freedom means a higher peak of violence over what would be expected from civil liberties and political rights alone. The exception are those states that are most libertarian, where the loss of some economic freedom may actually mean lowering the peak of violence. This aspect of the plots is contrary to theory, and it remains for further research to show whether it is an idiosyncrasy of these data. Most of the range over both plots, however, does conform to expectations, surely enough to say that the loss of economic freedom increases the peak violence in a state over that resulting from the denial of civil liberties and political rights alone.

Tests of both Hypothesis 1 and Hypothesis 2 have been successful, more than enough to conclude that the internal violence proposition is empirically supported.


In all tests conducted above and in the related "Libertarianism and International Violence," both freedom and political freedom were significant for all hypotheses and for all tests whether direct or indirect. However, the question remains whether freedom is the better definition, as theory would argue. Now, "better" has two interpretations. One is that freedom is significantly more significant than political freedom in the various tests of hypotheses and curvilinear regression and analyses. In no case here was this so, and in the other paper the results are mixed. Political freedom was most clearly and significantly better for the monadic and dyadic foreign violence chi-squares.20

The second interpretation of "better" is that freedom should account for more violence than does the comparable level of political freedom. That is, the further loss of economic freedom should increase the potential for violence over that for the loss of political freedom alone. For all the curvilinear or linear regression equations defined here and in the other report,"Libertarianism and International Violence," this was true over all or most of the range in political freedom or freedom scores.

Keeping in mind that freedom is the theoretically preferred definition, and that both political freedom and freedom were always significantly related to internal, foreign, and dyadic violence, that the results are mixed for freedom being significantly more significant, and that lack of freedom always or usually accounted for more violence, I believe we can prudently conclude that freedom should remain the primary definition of libertarianism. Libertarianism, then, should mean the existence of civil liberties, political rights, and economic freedom. And this tends to minimize violence within and between states.


The tests conducted above have shown that systematic evidence supports the Freedom/Domestic Violence Proposition: and, thus, the underlying theory. The more libertarian a state, the significantly less internal violence it has, and the significantly and predictably (in variance terms) lower its possible peak violence. Moreover, the tests conducted in the other, complementary report "Libertarianism and International Violence," show empirically that the more libertarian a state, the less it tends to be involved in foreign violence; the more libertarian two states, the less violence tends to occur between them; and libertarian states themselves do not have violence. The four empirically supported propositions are as follows.

Joint-Freedom Proposition: Libertarian systems mutually preclude violence (violence will occur between two states only if at least one is nonlibertarian).

Freedom/Dyadic Foreign Violence Proposition: Freedom inhibits international violence (the more libertarian two states, the less tendency toward mutual violence).

Freedom/Foreign Violence Proposition: Freedom inhibits foreign violence (the more libertarian a state, the less it tends to be involved in foreign violence).

Freedom/Domestic Violence Proposition: Freedom inhibits domestic violence (the more libertarian a state, the less internally violent it can and tends to become).

These propositions are also consistent with the general findings in the quantitative literature that I have surveyed and noted elsewhere22 and further support the conclusions of my Understanding Conflict and War. Moreover, the whole character of Understanding Conflict and War--including the theoretical and philosophical arguments--, the evidence and empirical analysis bearing on a variety of surrounding psychological, interpersonal, societal, and international propositions, and the integrating concept of a conflict helix endow the propositions tested in these reports with a credibility well beyond the very positive empirical results shown.

All this suggests that there is a fundamental principle involved, one that lies at the heart of the theory and underlies the propositions tested in these papers. Common to these propositions, as the operational political freedom and freedom definitions show, is the idea of freedom from arbitrary, repressive, and irresponsible government power. Positively, it is the freedom of religion, speech, and privacy, fair trial and security against arbitrary arrest and imprisonment, and the right to participate in open and free elections of leaders and votes on public issues. In other words, people are guaranteed rights over and above government, leaders are fully responsible to the people, and there is a culture of law and limited government. That is, government is relatively small, its size (power) restrained. All this is captured in the following principle already established in Vol. 4: War, Power, Peace:

The Polarity Principle: The more government, the more violence.24

By "more government" is meant more centralization of government power, more intervention in personal, social, and economic affairs and activities, more limits on political criticism and competition, and more narrowing of electoral choices. In other words, by "more government" is meant less freedom, less civil liberties, political rights, and economic freedom.

To understand this principle, then, is to comprehend at a fundamental level that the more government in society (and the principle is meant to apply not only to state-societies, but to societies at any level, including the global society of states), the less the diversity and mobility, the less the cross-pressures, the greater the polarization of interests, and the more likely that a conflict-front can traverse society and that violence can engulf all. Moreover, the more government, the more freedom the foreign policy elite have to engage the state in foreign violence and war.

While of course much more empirical and theoretical work on these ideas and assertions must be done, and no careful scientist can put to rest a social hypothesis as beyond prudent question, sufficient theoretical support and empirical findings have accumulated across all these studies to say with reasonable credibility that the classical liberal was likely correct: limiting government and increasing individual freedom promote peace.


* Scanned from Comparative Politics 16 (July 1984): 443-462. Typographical errors have been corrected, clarifications added, and style updated.

1. Note omitted.

2. See my "Libertarianism and International Violence," and "Libertarian Propositions on Violence within and between Nations: "A Test Against Published Research Results"

3. Vol. 2: The Conflict Helix.

4. Ibid., Section 35.4.

5. I do not regard this as a complete specification of factors, and if I were developing a forecasting model of internal violence I would follow Ted Robert Gurr and Mark Irving Lichbach, "Forecasting Domestic Political Conflict," in J. David Singer and Michael Wallace, eds., To Auger Well: Early Warning Indicators in World Politics (Beverly Hills: Sage Publications, 1979), pp. 153-193, in including the organizational strength and external support for dissident groups and the regime, and the disposition to conflict of a people (independent of the factors mentioned in the text). Incidentally, the findings of Gurr and Lichbach on their conflict model (for eighty-six states for internal violence from 1961 to 1965) strongly support the theory: "democracy" increases the extent of protest, while significantly reducing the severity (ratio of killed to population) both of protests and rebellion.

6. See Chapter 35 in Vol. 2: The Conflict Helix.

7. Subsequently, in "Libertarian Propositions on Violence within and between Nations: "A Test Against Published Research Results" I did a resurvey of the quantitative and systematic cross-national literature and found that, out of twenty-two relevant published analyses, five strongly supported the proposition that libertarianism is inversely related to internal violence, fourteen provided moderate support, one was ambiguous, and two were negative.

8. For the associated theory and classification of governments see Vol. 2: The Conflict Helix, Part VIII.

9. For the years 1976-1980, see Freedom at Issue (January-February 1977), 9; (January-February 1978), 7; (January-February 1979), 4-5; (January-February 1980), 4-5; (January-February 1981), 4-5. Also, beginning for 1978, Freedom House has published a Freedom in the World Yearbook, that presents the same ratings along with a brief description of freedom in each state. For example, see Raymond D. Gastil, Freedom in the World 1981 (New York: Freedom House, 1981).

10. Freedom at Issue (January-February 1981), 6.

11. Freedom House's ratings are explicit and consistent across years, cover all states and territories, and, most important, discriminate in regard to what I mean by libertarianism. To check its discrimination consider how Freedom House rates states whose freedom or nonfreedom is marginal or controversial. Israel, for example, is rated 2 in 1978 for political tights and 2 for civil liberties, compared to 1, 1 for Austria, Denmark, and Iceland. Poland and Yugoslavia are 6, 5, compared to 7, 6 for the USSR; 6, 6 for China; and 7, 7 for Albania. Iran in 1977 is 6, 5, changed to 5, 5 in 1980 after the revolution.

12. Note omitted.

13. A reliability test of both the scaling of the foreign conflict events and their resulting Foreign Conflict Scale was done in my "Libertarianism and International Violence," and showed the scaling and transformation to be highly reliable. Similar reliability tests were done for the internal violence scales, with the same good results.

14. Note omitted.

15. 1 know this has been true of my earlier work, particularly for the descriptive, empirical correlations in my Dimensions of Nations (Beverly Hills: Sage Publications, 1972), and those underlying the results in my National Attributes and Behavior (Beverly Hills: Sage Publications, 1978). The empirical analyses for these volumes was completed much before the theory developed in Vol. 2: The Conflict Helix.

16. "Libertarianism and International Violence."

17. "Current News" is prepared by the Current News Branch, Department of Air Force, as executive agent for the Department of Defense.

17a. On the coefficient of determination, see "correlation squared" of Understanding Correlation.

18. The transformation was as follows. Freedom scores 3 and 4 were each given a new score of 2; 5 and 6 a new score of 3; 7 and 8 a new score of 4.... ; 27 and 28 a new score of 14. Then, the highest violence score for 3 and 4 becomes that for new score 2; the highest violence score for 5 and 6 becomes that for new score 3, etc.

19. Note omitted.

20. This is for Hypotheses 2 and Hypotheses 3 of the related report on "Libertarianism and International Violence,".

21. Note omitted.

22. See Chapter 35 of Vol. 2: The Conflict Helix; and "A Test of Libertarian Propositions on Violence."

23. Note omitted.

24. See also Chapter 18 of my The Conflict Helix: Principles and Practices...., where I devote a chapter to the principle and treat it as one in a connected sequence of psychological, interpersonal, and social principles.

You are the visitor since 11/27/02

Go to top of document