We believe that the positive association between population density and war found in previous studies is a result of omitted variable bias. That is, failure to include all relevant control variables will lead to biased coefficient estimates. Obviously, any study relying on bivariate methods fails to include the relevant controls. We show in the appendix how omitted variable bias could, in our particular case, lead to a wide range of values for the association between population density and war. Leaving out important variables that are both highly correlated with population density and positively associated with war—variables such as the use of metal or the existence of writing and record-keeping—causes our measure for population density to capture variation that belongs to these other variables, and gives the false result that population density is positively associated with war. The exact source of the negative relationship between population density and war is unclear. But we speculate that it may be due to the reluctance of others to attack high population density communities or to the reluctance of high population density communities to attack others. The former is plausible since high population density communities would have improved land and structures worth defending, and would have the potential to field large, well-equipped forces for effective defense. The latter because relatively complex high population density communities may find war a costly disruption, both to their economy and to the ambitions of their elites, who are likely to have non-war strategies for obtaining and keeping status. We had speculated that communities with little political autonomy would be less likely to go to war,plastic garden pots since any antagonism toward an external community would require the approval of the larger polity before actual war could develop. The estimated coefficients reveal a more complex pattern, as shown in Figure 4: both totally dependent and fully autonomous communities have a higher propensity to go to war than do communities that are semi-autonomous.
It seems likely that the semi-autonomous are constrained by the larger polity from freely engaging in war, but are also too loosely controlled for the larger polity to draw them into irrelevant wars. On the other hand, totally dependent communities can be forced into wars by the larger polity, while the fully autonomous can enter any war that they choose. Most warlike of all are communities which have equal status to other communities in a pluralistic society. Such equal status may be the consequence of an often-exercised willingness to fight other communities. The theoretical minimum for the number of levels is two ; v236 considers up to a maximum of four levels. As Figure 5 shows, communities with this maximum number of levels are much less likely to engage in external war than are other communities. Decisions to engage in external war may be more difficult when larger numbers of political actors must agree, as would be the case in larger, more complex, communities. As an example, one might consider the Mae Enga, of the New Guinea highlands, who deliberate the war decision in large meetings, where all fighters have the chance to speak . From the above, one can see that features of communities with a low propensity to engage in war include: high population densities ; relatively complex community-level political structures ; and ties to other communities that constrain the free exercise of war . These features reflect a more complex social and material order. Chronic war is the enemy of order, since its object is to destroy the crops, structures, institutions, and lives of a people. One would expect a community with a long history of peace to have evolved a complex social and material life, able to sustain high population densities. Thus, the features identified by our model may be a cause of low levels of war, as we hypothesize, but can also be a consequence. Our results provide some insight into causality, in the form of the endogeneity tests. Since these show that no independent variables are endogenous, our estimated model in Table 5 can therefore be interpreted as representing solely the causes of external war, not consequences.
Three of the 12 major habitat dummies survived to the final model . Relative to all other habitat types, societies found within temperate coniferous forests or boreal forests/taigas have lower incidences of external warfare. Conversely, societies who make their homes in temperate broad leaf and mixed forests experience higher incidences of warfare. These results confirm Nolan’s suggestion that there are features of biophysical environments that affect the frequency of war. Though the exact paths of that effect are not clear, they are independent of the confounding effects of subsistence and cultural transmission, which are controlled for in our model. The general thesis of the ecological-evolutionary theorists is that ecology, subsistence type, and population density are the dominant determinants of the frequency with which a society goes to war. Table 7 shows that ecology, subsistence type, and population density together account for only about 17 percent of the variation in the frequency of external war. If one broadens the set of variables to include technology facilitating war , the broader set accounts for about 27 percent of the variation in the dependent variable. Thus, only when quite broadly defined does the general thesis of the ecological-evolutionary theorists find strong support in our results. Otterbein has argued that sociopolitical variables have much more influence than economic or ecological variables in determining the frequency and nature of war. Variables reflecting political organization account for about 11 percent of the variation in the frequency of external war; variables reflecting the strategies by which elites gain status account for another 10 percent. Add to this the five percent accounted for by the degree to which the supernatural supports morality, and the resulting 26 percent of variation in frequency of external war accounted for by sociopolitical factors is about the same as the 27 percent accounted for by the broadly defined ecological evolutionary variables. Some believe that the frequency of war may simply be a function of who a society happens to have as neighbors: Keely suggests that hostile neighbors may be the most important determinant of whether a society is warlike, and Younger finds that more isolated societies are more peaceful.
We include two variables that provide some measure of the effect of neighbors, and together they account for about 12 percent of the variation in the frequency of external war. The first of these confirms Younger’s view that more isolated societies are more peaceful. The second— our network lag term—shows the effect of cultural transmission. The network lag term’s optimal composite weight matrix indicates that societies will tend to engage in war at much the same frequencies as their geographical neighbors and their co-religionists. Table 5 contains two pieces of evidence suggesting that a society will not be much influenced by the frequency with which their ancestors went to war: the near-zero value of the composite weight for linguistic phylogeny, and the high p-value for the LM test for spatial lag based on linguistic phylogeny. In other words, there is evidence here that vertical transmission does not account for the frequency of external war. We re-examine Patrick Nolan’s empirical work on the causes of war. We criticize his methods, which consist of bivariate or tri-variate tabular analyses, for sacrificing variation, for ignoring confounding variables, for failing to show the relative importance of the analyzed effects, for ignoring Galton’s problem, and for ignoring the problem of missing data. Our approach is to build a multivariate model,square pots which uses multiple imputation to handle the problem of missing data, and uses a network lag term to handle Galton’s problem. Our results reinforce Nolan’s conclusions on a few points, notably the positive association between metal technology and war. And while this relationship is important, it is hardly decisive—accounting for about 6.5 percent of the variation in the frequency of external war. When we evaluate the total importance of all factors related to ecology, subsistence, population density, and technology, we find that together they explain about 27 percent of the variation in the frequency of external war. This is comparable to the 26 percent explained by a broad set of sociopolitical factors. Thus our results suggest that those who argue for ecological-evolutionary theory, such as Nolan, are about as correct as those who argue that sociopolitical factors are the main determinants of war, such as Otterbein . This serves as an example of the superiority of multivariate methods: by including all of the most likely determinants of war, one can gain a sense of their relative importance. For the first of the two specific hypotheses advanced by Nolan—more productive subsistence leads to more frequent war—we find only qualified support. Taking the proportion of subsistence derived from agriculture as a measure of productivity, we find the relationship to be quadratic. As Nolan would predict, increases in agriculture’s importance leads to increases in the frequency of external war, but for non-agricultural societies only. For societies primarily relying on agriculture, we find a result opposite to that predicted by Nolan: increases in agriculture’s importance lower the frequency of external war.
Our results explicitly contradict Nolan’s second hypothesis: that higher population densities lead to higher frequency of war. We find a strictly negative relationship, in which high population densities discourage war. In the appendix we show that omitted variable bias is the probable reason that other studies failed to find a negative relationship. This highlights again the necessity of multivariate models in cross-cultural research—only by considering all important confounding factors can a model be free of omitted variable bias. Finally, we feel encouraged that our results support an optimistic view of peace among human societies. The propensity to engage in war is not vertically transmitted, is not a behavior that a society is locked into by the practices of its ancestors, but rather appears to be a product of current conditions. And many of the features of contemporary societies appear to be those which favor peace: high population densities; moderately restricted political autonomy; more complex political structures; widespread belief in moralizing gods; and the prevalence of capitalism as a means for elites to gain status. If peace is our goal, perhaps we are heading in the right direction.In alternate bearing theory, there are three competing hypothesis that attempt to explain how the fruit negatively influence vegetative growth. The competition hypothesis suggests that the demand between two sink tissues determines the flow of nutrients to each organ, the inhibitor hypothesis suggests that either the leaves or the fruits suppress flower development even when nutrient supplies are adequate, while the dominance hypothesis suggests that the fruits reverse the apical dominance mechanism, suppressing the apical meristem and subsequent vegetative growth. Although the competition hypothesis is favored by the growth trade-off observed in many alternate bearing species, there been few attempts to determine which of the mechanisms predominates in actual growing tissues. One way of doing this would be to observe the expression profile of actively growing meristems subjected to a heavy fruit load, as each of the hypothesis could be expected to produce a unique signature of up and down regulated genes. For example, a recent genetic study of alternate bearing apple trees was able to demonstrate several genes related to auxin and gibberellin hormone pathways were located in alternate bearing quantitative trait loci. The presence of auxin genes could be used to support the fruit dominance hypothesis, while gibberellin and floral induction genes might indicate the presence of an inhibitor pathway. An impressive set of microarray data from alternate bearing mandarin scions found that several glucan and trehalose sugar-related genes were activated during the ON year. The authors further argued that the FT paralogs CiFT1 and CiFT3 were involved in suppressing vegetative growth, but recommended more work to validate this idea. One aspect not captured by either study is the degree of fruit load, which varies continuously over the annual production cycle and may even produce a concentration gradient in the case of the inhibitor theory. However, even if reported in terms of the alternate bearing index or fruit biomass, the use of averaging and biological replicates would obscure the signal before dosage sensitive responses could be extracted from the data.