Imagine that an STI which leads to infertility occurs in the female population with probability k. Then, an uninfected man and his wives can avoid infertility if and only if none of his wives have the infection. This occurs with probability: n . If we assume an empirically plausible rate for this STI, like 0.07 , then a monogamous man would be paired with fertile woman 0.93 of the time, and a man in 2-polygyny would be paired with fertile women 0.86 of the time. An ultra-wealthy man in 10-polygyny, however, would be more likely than not to have infertile wives—a prospect that could lead to diminishing fitness returns to, and hence, demand for, additional wives. A similar argument holds even if marriage is considered sequentially—as in serial monogamy—though the effect would be smaller. Finally, impediments to cooperation or even outright conflict among co-wives can be greater as the number of wives increases. Interference competition among co-wives could impose significant fitness costs in settings where effective child rearing benefits from cooperation. It could well be that incumbent wives resist incorporation of additional wives to the wealth sharing pool, perhaps with greater effectiveness as their numbers grow. Empirically exploring these and other possible explanations for the unexpectedly substantial diminishing fitness returns to additional wives that are not explainable by the division of rival wealth among wives would be a valuable next step, but one that would take us beyond the formal modelling, database and comparative statistical methods that we have presented.Recent anthropological research on the causes of war falls roughly into two schools : one which concludes that sociopolitical factors are the primary determinants of war ,large plastic pots and another which argues that environmental and technological factors are primary . Both “schools” have developed testable hypotheses about the conditions making a society more likely to go to war .
In this study we retest two hypotheses initially tested by Nolan : first, more productive subsistence technology leads to more war; and second, higher population density leads to more war. Both hypotheses come from ecological-evolutionary theory, which asserts that subsistence technology is the single most important factor affecting how societies are organized and how they interact with one another . Ecological evolutionary theory predicts warfare is more frequent in societies with more productive subsistence technology for three main reasons. First, armies need to eat; a society with unproductive subsistence technology would not have the food stock to sustain soldiers in a prolonged conflict. Second, food stores make for attractive targets; hence, societies with more productive subsistence technology are more likely to be attacked. And third, fixed investments in fields and structures give a society with more productive subsistence technology a strong incentive to defend its territory from attackers.Some have claimed that, compared to other societies, hunter-gatherers are relatively peaceful while others disagree . Ecological-evolutionary theorists tend towards the ‘relatively peaceful’ side of the debate. Nolan , for example, argues that hunter-gatherers usually lack the resources to sustain them during long periods of warfare. Likewise, they have little to be plundered and find it feasible to walk away from a confrontation and move to a new area. Ecological-evolutionary theorists expect the frequency of warfare to be higher among horticulturalists than among hunter-gatherers. Horticulturalists own more resources than hunting-gathering societies, and are more attractive targets. While hunter gatherers can with relative ease walk away from an attack and move to a new location , horticulturalists are more likely to defend their lands and structures . And since metal tools make horticulturalists more productive, and metal weapons make warriors more deadly, horticulturalists are expected to have even higher rates of warfare if they have learned to use metal .
Slavery has a similar effect in increasing the frequency of war: armies can capture slaves, which can then be used to produce food, which can be used to feed armies, which can then capture more slaves – thus creating a positive feedback warfare-slavery system . Ecological-evolutionary theorists expect plow agriculturalists to have even higher frequencies of external warfare than horticulturalists , since they have more productive subsistence technology, which compounds the warfare incentives affecting horticulturalists. Agrarians are even more reliant on their land than horticulturalists and therefore even less likely to walk away from a confrontation; they can produce even larger and more diverse food stores – further incentivizing plunderers; their larger food stores can feed larger armies; and slavery becomes even more profitable as more productive technology increases the returns to labor. For these reasons, most previous studies of the ecological-evolutionary theory of warfare have supported the idea that as subsistence technology becomes more productive, the frequency of warfare increases. The hypothesis of a positive correlation between population density and warfare is based upon the notion that, within a given subsistence technology type, increases in population density will increase the pressure on a society’s resources, thus motivating a society to plunder the resources and conqueror the arable lands of its neighbors. Hence, some studies maintain that the frequency of warfare will increase as population density increases. Other studies argue for a cubic relationship – claiming that warfare increases, decreases, and finally increases again across categories of increasing population density. In his study of precontact Polynesia, Younger finds a negative relationship between population density and violence. Thus, there is by no means a consensus within the literature about this relationship. In a 2003 paper published in Sociological Theory, Patrick D. Nolan sets out to test whether certain modes of subsistence are “structurally conducive” to warfare . Using variables from the Standard Cross-Cultural Sample , Nolan produces a number of contingency tables, examining whether the frequency of warfare varies across four subsistence modes and by population density.
He finds that societies with advanced horticulture or agrarian subsistence engage in warfare much more frequently than those with foraging or simple horticulture. He also finds that high population densities are associated with more war in societies with foraging or agrarian subsistence, but not in those with horticulture. While we find much of value in Nolan’s theoretical discussion, there are serious problems with his empirical analysis. First, in adopting a contingency table approach, Nolan chooses a method that requires reducing the variation found in the original SCCS variables, in order to have relatively few cells in each table. Thus he takes a measure of frequency of warfare , available in the SCCS with 18 discrete values, and turns it into a dummy variable, with only two discrete values. Likewise, he employs only four general subsistence categories, a feat he manages by lumping hunting with gathering, and discarding 54 of the 186 SCCS societies that subsist as mounted hunters, fishers, pastoralists, or rely equally on two or more subsistence modes . Particularly problematic is the removal of relatively warlike pastoralists and mounted hunters from his sample. Table 1 compares Nolan’s categories with the categories given in SCCS v858. The four subsistence taxonomies presented in the SCCS provide much richer detail: variable v246 has seven categories; variable v833 has eight; variable v858 has 11; and combining variables v833and v834gives 28 categories. But even more variation can be found by using variables v814–v819, which provide actual ordinal measures of the percentage dependence on each category of agriculture, domestic animals, fishing, hunting, gathering, and trade. Variation is the great friend of any statistical analysis, and the SCCS contains variables that provide abundant variation on subsistence technology. That Nolan reformulates his variables to reduce variation is due simply to his choice of technique— contingency tables work well only when there are relatively few cells in the table. Contingency tables also limit the analysis to pairwise relationships between variables, and at best can be modified to fit a three-way relationship. When there are confounding variables, as there always are,plastic pots for plants the results from a pairwise analysis will be biased. It is on this count that multivariate models provide their greatest advantage: one can control for the effects of other variables and thus produce unbiased estimates. And because multivariate methods consider a large set of variables, one can gain a sense of how important a particular relationship is in the grander scheme of things. Galton’s problem—the confounding effect of cultural transmission—is a major methodological issue in empirical studies using cross-cultural survey data. There are no effective ways to control for Galton’s problem within a contingency table framework and Nolan wisely does not try. Dow has developed an effective way of modeling Galton’s problem within a multivariate model framework, which provides yet another advantage to multivariate methods over contingency tables. Finally, there is the problem of missing data. As shown by Dow and Eff , dropping observations for which data values are missing can lead to bias, even within a multivariate modeling context. Nolan dropped observations that were not even missing—removing 54 societies that did not fit neatly into his subsistence taxonomy categories .
The appropriate way to handle missing data is the technique of multiple imputation . Nolan’s empirical work does not do justice to his theoretical discussion. In what follows, we investigate the role of subsistence technology and population density in causing war, using contemporary best-practice statistical methods. Wealth accumulation provides the means for elites to enhance their status, and this status-enhancement strategy is both an alternative to the strategy of war and can be disrupted by war, making war less attractive to elites. SCCS variable v17 was included as a scale for the use of money and credit. Small communities are less likely to have powerful elites , and are therefore less likely to be driven by elites to war. In fact, Younger’s study of Polynesian war found that smaller communities were more peaceful. We therefore include a scale for community size . Elites will also be less able to drive a society to external war when their power is less secure; variable v575 was included as an “unstable political power index.” If local political structures are complex, elite power may be fragmented, and it may be difficult for all relevant actors to agree upon a course of action, which would make war less frequent. For this reason, v236, a measure of the number of levels in the jurisdictional hierarchy of the local community, is included. Similarly, we introduce variable v757, which measures whether political authority is simultaneously religious authority, reasoning that more heterogeneous elites are less likely to find common cause in external war. Military historians speak of the three “C’s” in the analysis of war: causes, conduct, and consequences . Our analysis focuses on the causes of war, but features of a society facilitating its conduct will also serve to make a society more likely to go to war, and in this way the feature can be seen as a cause. For example, when females contribute a great deal to subsistence, the opportunity cost of sending a man to war is lower, and the choice of war therefore is less costly. For this reason, we include variable v826, “average female contribution to subsistence.” Certain kinds of technology facilitate the conduct of war. For example, the development of metal weapons makes warfare more practicable and increases its incidence . Thus, we create the variable metal as the composite for four SCCS variables concerning the use of metal tools and weapons.4 Similarly, Nolan notes that limited communications and transportation technology “were the only real constraints” horticulturalists faced in using war to control more land and subject peoples. We employ here v149, which measures the degree to which a society utilizes writing and records. A prior study by Brown and Eff found the frequency of external warfare and the presence of moralizing gods to be inversely related; suggesting that morality reinforced by supernatural forces may serve to constrain warlike behavior. We investigate this by including SCCS variable v238 in our model. The bottom panel in Figure 3 plots the total effect on the ordinate and pdens on the abscissa. Since pdens is standardized, the total effect will be positive for societies with below-mean population density, and negative for societies with above-mean population density. The resulting plot is nearly linear—fitting a negatively sloped straight line with an R 2 of 0.796. Our model’s negative relationship between population density and the frequency of war is consistent with Younger’s findings but not those of Nolan or Keeley .