Proximal Causality in the Adoption of Ethical God Beliefs
Draft@Iterations (c) Douglas R. White, Ren Feng, Giorgio Gosti, Elliott Wagner, B. Tolga Oztan and John Snarey 2011-2012*
- Printable Moral Gods Model & Data (Instructional page for Xi'an Jiaotung University course on Networks and Causality: Ren Feng, Haifeng Du and D. R. White)
Proximal Causality in the Adoption of Ethical God Beliefs
Former title: Inequality, Ethics and Altruism in the Evolution of Moral Gods
Introduction
In The Biology of Moral Systems, Richard Alexander (1987:4) noted that only five authors other than himself had "made extensive attempts to apply recent evolutionary theory to the study of ethical questions" although valiant attempts had been made by much earlier authors to relate evolution and ethics to Darwin (1871) and Huxley (1896). Themes explored by Alexander about religion, ethics and altruism spawned a research literature about the sources of human cooperation and altruism, including stimulated game-theoretic models of indirect reciprocity, the free-rider and punishment problems and also a host of divergent theories about the adaptive benefits of religious beliefs and beliefs about supernatural punishment for prosociality within and between groups, reproductive fitness, group-level adaptation, costly signaling and group commitment. One of his least cited themes, however, is the evolution of ethical religions as contributing to development of concepts of justice and law in response to the development of inequality.
The findings we report here draw on previous cross-cultural studies that use the data of the Standard Cross-Cultural Sample (SCCS, Murdock and White 1969) for the study of diffusion of what are called in that literature moral or ethical gods. For comparison with other studies, we trace the diffusion of those forms of religion distinguished in Swanson's (1960) "high gods" variable in his book Birth of the Gods. Murdock (1967:52. 61-125) coded the societies of his 1200+ Ethnographic Atlas societies using Swanson's categories and the codes on 168 of these societies were added as a variable of the Standard Cross-Cultural Sample of 186 societies. Snarey (1996) independently coded all 186 SCCS societies using the same categories as Murdock and Swanson. Many authors use the term moral or moralizing god beliefs to emphasize the relation of religious morality to group norms, e.g., increasing group solidarity, as in religious wars. We, like Snarey (1996), often refer to them as ethical god beliefs. In doing so we emphasize one of the main themes of Alexander: that interpersonal principles of justice are often present in monotheistic religions that envision a god’s involvement not only in moral prescriptions and sanctions, and intragroup benefits of indirect reciprocity, but ethical behavior towards others more generally.
Our focus is on issues of inequalities and interpersonal justice largely neglected in the cross-cultural literature as potential causes of beliefs in moral/ethical gods, whichever they are called. These issues involve how religion affects altruism, cooperation, and sanctions for morality in contexts of scarcity, inequality, and/or perceived injustice. This theme for the study of High Gods – which is also Alexander's – is also supported by Swanson's finding (1960:174) that "supernatural sanctions for interpersonal relations are most likely to appear in societies in which there are interpersonal differences according to wealth.” Swanson's tables (p. 166-167) also show supernatural sanctions correlated (p < .05) with considerable debt, property ownership, primogeniture, grain crops for sustenance, and (p < .0005) social class distinctions. Support for Swanson’s views reappear in Snarey's (1996) review of Swanson's work as potential issues of fairness or ethics: “supernatural support of human morality was more common in societies that also had parallel explicit social rules legitimating the morality of ‘interpersonal differences according to wealth.’"
More recent literature interpreting and suggesting hypotheses about the “moralizing” gods variable of the 1270-society Ethnographic Atlas (Murdock 1967) and SCCS focuses on “moral solidarity” issues such as the value of religious unity in warfare (Roes 1995, Johnson 2008). Roes and Raymond (2003) “speculatively picture the historical chain of events giving rise to a belief in moralizing gods” as follows:
- “human social groups became large as a result of between-group competition over preferred habitats and resources, but although larger social groups are more successful in competition, they also experience more pressures to fission. Morality unites a society by limiting infringements upon the rights of other society members, so if larger societies are indeed more likely to split, then those that remain intact may be expected to have more effective inviolable moral rules, such as those imposed by moralizing gods. Cross-cultural analyses support this line of thought: more competition between societies is found in environments rich in resources and larger societies tend to occupy these environments; large societies engage in external conflicts at higher rates and are more often characterized by beliefs in moralizing gods.”
The implication that moral gods coevolved with competitive success in warfare is a spurious one, as are the purported cross-cultural correlations between moralizing gods, large societies, and external conflicts. Brown and Eff (2010) refute these speculations using SCCS data, missing data imputation and controls for nonindependence of case and show that moral gods occur in the context of external peace, given diverse independent variables such as dependence on animal husbandry, agricultural potential, food scarcity and caste stratification. Eff and Routon (2012) show that moralizing gods, net of other factors, have an effect counter to external war.
Religion and inequality
Studies such as Roes (1995) or Johnson (2008) fail to treat the theme of ethical redress raised by Alexander: that religion may pose justifications for inequality that stabilize ranked or class societies, but only up to a point. That critical point is where inequality becomes so massive as to lack legitimacy and is judged to require new agreements on principles of ethics to redress the perceived wrongs of extreme inequalities. Pre-WW I Britain, as the world's leading economy, for example, went through two of these swings, often called "secular cycles" (Turchin and Nefedov 2009) in world systems (e.g., Fletcher 2010), demographic-structural historical dynamics theory (Goldstone 1991, 2002, Turchin 2003) and "long cycles" models of world politics (Modelski 1987, Modelski and Thompson 1995, White 2009, White, Tambayong, Kejžar 2008). The secular cycles in world systems model of Fletcher et al. (2010) also show reduction of external warfare following Malthusian crises of overpopulation. Swings and cycles of varying lengths are salient elements of historical dynamics. If and when Malthusian crises occur at a period of climax of a secular cycle, inequality may amplify to massive proportions because the normal rebalancing of exchange may not occur. Inequalities may amplify as between owners of productive property that is increasingly valuable in times of scarcity and increasingly numerous workers undervalued in times of overpopulation. Extremes of inequality may come to be viewed as unjust and require ethical rebalancing. In the last century the US went through a major swing from massive inequality (pre-1920) to the Roosevelt consensus (post-1935) to shift again to massive inequality (2004-2012), with implications for conflict, religious changes, and for new policy issues concerning inequality. Going back to the origins of Islam, Turchin (2006:97-102) explains eruptions of conflict and extreme inequalities in the secular cycles of the Arabic peninsula and trading center of Medina, for example, as "the triggering event [that] occurred during the decades around A.D. 600" for "people exposed to years and decades of chaos" who "begin to yearn for stability. Any message of hope ... becomes attractive. Monotheistic religions effectively address ... that need." Islam emerged among the many competing religious movements of that period.
Shifts or adoptions of major world religions, including those involving the diffusion of beliefs in "moralizing" or ethical gods, as in Judaism, Christianity and Islamin the SCCS, as well as 9-10 non world-religion cases, provide the contexts and questions explored in our study. Given the many cross-cultural studies on religion stimulated by the work of Alexander and Swanson, it is surprising that very few have focused on how the diffusion of "moralizing" or ethical gods is related to societal conditions of inequality. Focusing on this lapse in the literature, our research is oriented toward evaluating whether evidence for these missing connections could be found in the comparative ethnographic case-study data of the SCCS with a focus on processes that generate either scarcity -- as a motivation for cooperation -- or extreme inequality as a motivation for rebalancing the social contract.
The map in Figure 1 shows that most of the SCCS societies with "moralizing gods" are in Eurasia or North Africa and few in indigenous North America or Australia and the Pacific. The SCCS is composed of the most culturally diverse and earliest societies best-described in the ethnographic literature for each cultural region of the world, and include only three pre-1500 societies and six and eight others prior to 1700 and 1800. Eric Wolf's (1982) analysis of the myriad of ethnographically studied cases affected by Europe or European colonialism after 1500, then, would apply to the Old and New World societies in Figure 1, without overrepresenting their worldwide spread per se in the choice of SCCS societies. East Asian societies, as represented in this map, has many moral /ethnical religions (Hinduism, varieties of Buddhism, and many others), but few high gods in the sense of Swanson's (1960:210) definition of a monotheistic high god as a single spiritual being "who is believed to have created all reality and/or to be its ultimate governor." The map includes syncretisms in the New World that only partially fulfill this definition such as father or mother creator gods with a non-creator mate. Swanson's coding of types of high gods distinguishes four categories (n=number of cases for each as coded by Snarey 1996): (1) absent or not reported in substantial descriptions of religious beliefs (n=69), contrasted with remaining monotheistic cases; (2) present but inactive and unconcerned with human affairs (n=51), (3) present and in human affairs but not offering positive support to human morality (n=23), and (4) active and specifically supportive of human morality referred to as a moral or moralizing god (n=43).
Many previous studies have used the SCCS data on "moralizing gods" but few had available ratings by Korotayev (2003) on Islam (n=19) and Christianity (n=6) among the moral god societies. In one case, Javanese, the national religion is coded as Islamic while the village has syncretic Hinduism, thus not fulfilling the high god definition. He distinguishes other societies superficially influenced by these religions in their classical forms (n=7, 24 respectively).Most of the Eurasian cases on map 1 are Islamic or Christian (19+6=25, 24 moral gods), or superficially so (7+24=31), of which only nine are judged as moral god societies. Other independent cases of full moral gods number only nine to ten or 21-23% of the total sample. The Yahgan case is controversial given the ethnographer's bias toward Egyptologist Father Schmidt's theory of a universal "primitive monotheism." To study what affects the presence or absence of moral gods for societies in the SCCS, which were mostly well-described ethnographically after 1800, requires recognition that the dependent variable is the diffusion or adoption of a moral god religion, as well as difference in origin for Christianity or Islam that derive from the first several centuries A.D. or the 7th-8th centuries. Because Snarey (1996) coded the ethnographic present of each society as to Missionization or unmissionized, this also becomes a predictor of the High God variable.
Testing hypotheses about the effects of scarcity and the crisis periods in secular cycles
Secular cycles are known to occur in all types of empires and many types of states, most of which are located in the Old World and few in pre-European Americas. One of the features of these swings is that they often recur in irregular time-spans of a century (thus "secular") or more. A "crisis period" for secular swings in imperial dynamics occurs when population-overrun of jobs or resources makes the job holders or seekers disadvantaged relative to owners of productive property. Productive property becomes more valuable as it becomes more scarce or more narrowly held. Swings leading to periods of scarcity and heightened inequalities can lead to social or physical conflict, eventually followed by decline, then a period of renewal; and eventually -- without interventions such as external wars that tip balances of power and resources depending on outcomes -- decline may be followed by the next major swing, experienced as a period of recovery and innovation with a rise in population that in the long run may eventually recreate the conditions for scarcity, inequality and the recurrence of conflict.
Although they found predictors of moral god beliefs in the variables of the SCCS, using far more sophisticated methods than Roes and Raymond (2003), who relied on a concatenation of correlations without proper statistical testing, Brown and Eff (2010) undertook to test Swanson's theory that beliefs in the hierarchical structuring of gods, culminating in monotheism and a single moral god, reflected deeply held beliefs and practices in the ideal constitutional order of political hierarchies. They did not test hypotheses derived from Alexander, other than in the speculative hypotheses of Roes and Raymond (2003), and they did not test Swanson's ideas about the association between supernatural sanctions and beliefs related to rebalancing social stresses and inequalities related to debt, property ownership, primogeniture, grain crops for sustenance, and social class distinctions and the issues of fairness or ethics in contexts and environments of scarcity, inequality, and/or perceived injustice that concerned Alexander. They did find that food scarcity predicted moral gods, similar to Snarey's (1996) finding about water scarcity. While their work refuted an argument based on the Roes' use of spurious correlations from the Atlas of 1270 societies (thereby inflating significance tests for highly clustered cases), they did not find direct support for Swanson's theory that religious beliefs reflected, cross-culturally, the quasi-political constitutional structure of hierarchical relations in society. It must be said on Swanson's behalf, however, that his "constitutional order" theory did work beautifully for his study of the Protestant reformation as against Medieval Catholicism. In that context, however, much of his explanation depended on conflicts within the economic order as to independent merchant-traders in politically independent cities or those struggling for independence versus the political empires dominated by monarchies and empires. CITE SOME OF THE STUDIES THAT FIND THE ECONOMIC DIMENSION TO EQUAL THAT OF THE POLITICAL.
Building on theoretical predictions about the effects of the crisis stage of Malthusian cycles at the limit of rising population relative to resources, our research focused on finding indicators of exchange systems in which the inequalities of resource scarcity would be amplified by the contrast between property owners and labor. In pastoral societies this would correspond to one between herd owners and herders of the animals of other clans or lineages. In agricultural societies the contrast would apply to that between landowners and laborers. In a political context of rulers and ruled it would apply not to a patrimonial government based on redistribution of goods but one with collection of taxes, and especially monetary taxes. We defined three such variables and found that all were significant, improving on Brown and Eff's regression model of moral gods. Using Snarey's fully coded HiGod4 measure further improved the model, as did combining his measure of water scarcity with an SCCS dry zone ecological variable. The effects of these improvements, especially in the independent variables, was remarkable, as shown in the section below. The definitions of all variables used in the model are shown in Table 1 (http://intersci.ss.uci.edu/wiki/index.php?title=Draft). The full SCCS codebook is at http://eclectic.ss.uci.edu/~drwhite/courses/SCCCodes.htm.
Table 1: SCCS and Snarey Variables and their Sample Sizes (n)
Eff and Dow 2009 pdf quick download Brown and Eff 2010 pdf quick download # Variable_____________.n= sccs$v________________________Description _1 Malthus event________. 14 Unmeasured in SCCS Data from Turchin for n=14 states and empires _2 HiGod4_______________.186 Snarey:HiGod4 v208==1)*1 High god (moral/ethical god, more fully coded than sccs$v238 by Murdock) _3 AnimXbwealth 18^28___.186 v206*(v208==1)*1 Bridewealth times dependence on animals _4 FxCmtyWages__11^12___.112 v61*Wages Fixed community and union of (wages,labor) variables (see 37:MsgFxCmtySize) _5 SuperjhWriting 31^32_.184 v237*v149 Jurisdictional hierarchy and Writing _6 No_rain_Dry__________.186 Snarey:No_rain*Dry Scarcity of rainfall _7 Socialclass__________.186 v270 Degree of social class stratification _8 Evileye______________.186 v1189 Belief in evil eye _9 Islamic______________.186 (v2002==2)*2+(v2002==4)*1 Islamic 10 Christian____________.186 (v2002==3)*2+(v2002==5)*1 Christian 11 FxCmty_______________.186 v61 Fixity of community 12 Wages________________.112 f(v1009, v1732) Wages for labor 13 Milk_________________.186 (v245>1)*1 Dummy: milk consumed 14 Logdate______________.186 log(v838)-6 outliers reduced Log of ethnographic study date 15 Missions_____________.186 Snarey:Missions Presence of missions 16 Socplex______________.093 v711 Societal complexity Guttman scale 17 Plow_________________.186 (v243>1)*1 Plow 18 Anim_________________.186 v206 Percentage subsistence: Animal husbandry 19 Ecorich______________.186 (sccs$v857==3|sccs$v857==4)*1 Richness of ecology +(sccs$v857==5)*2 20 PCAP_________________.186 v921, v928 1st PC: Agricultural potential high 21 Foodscarc____________.144 v1685 Chronic resource scarcity high 22 Popdens______________.186 v857 Population density 23 PCsize_______________.186 v63, v237 1st PC: Community size large 24 PCsize.2_____________.186 PCsize^2 PC size squared 25 Caststrat ___________.186 v270 Degree of caste stratificatiobn 26 Eextwar______________.186 v1650 Frequency of external war high 27 Frqintwar____________.160 v891 Frequency of internal war high 28 Bwealth______________.186 (v208==1)*1 Bridewealth 29 Fratgrpstr___________.082 v570 Fraternal interest group strength high (see 38:MsgFratgrpstr) 30 Aglateboy____________.148 v300 Aggression of late boys high 31 Writing______________.186 v149 Writing 32 Superjh______________.184 v237 Jurisdictional hierarchy above local community 33 Commsize_____________.186 v63 Community size large 34 AP1__________________.186 v921 Agricultural potential high: sum of scales 35 AP2__________________.186 v928 Agricultural potential high: minimum of scales 36 Dry__________________.186 (v855>4)*1 Dry ecozone 37 MsgFxCmtySize________.186 --- Dummy: Missing data for 4:FxCmtySize 38 MsgFratgrpstr________.186 --- Dummy: Missing data for 29:Fratgrpstr 40 sccs$labor___________.052+ sccs$labor=(3-(sccs$v1009<=3)*1)+(-1+(sccs$v1009==5)*1) v1009 wage labor 41 sccs$Labor___________.052+ sccs$Labor=replace(sccs$labor,is.na( sccs$labor),0) 39 sccs$wages___________.089+ sccs$wages=replace(sccs$v1732,is.na(sccs$v1732),0) v1732 wage labor 42 Wages+_______________.089+ Wages=apply(sccs[,c("wages","Labor")],1,max) 43 Wages________________.112 Wages[which(Wages==0)] = NA #computed as: v1009, v1732 wage labor sccs$labor=(3-(sccs$v1009<=3)*1)+(-1+(sccs$v1009==5)*1) #define 1st variable sccs$Labor=replace( sccs$labor,is.na( sccs$labor),0) #replace NA with 0, 1st variable sccs$wages=replace(sccs$v1732,is.na(sccs$v1732),0) #replace NA with 0, 2nd variable Wages=apply(sccs[,c("wages","Labor")],1,max) #take maximum of 1st, 2nd variables Wages[which(Wages==0)] = NA #replace 0 with NA for combined variable, 44 Distance_____________.186x185 Distance-weighted predictions from independent to dependent variables 45 Common Ancestry______.186x185 Distance-weighted predictions from independent to dependent variables
Results
Our modeling results in Graphic 1 use the same regression methods as Brown and Eff (2010), with instruments for measuring autocorrelation effects (not shown in the graph) but with five major differences. One is that we ran regression models for links in a full network of graphs, one for each of the predictors of moral gods as well as for moral gods as the only dependent variable. Two, we used Brown and Eff's methods of missing data imputation on all missing data, for dependent and independent variables, and not just for the data missing for cases coded on the dependent variable. Three, we created dummy missing/nonmissing data quality control variables for those with >15% missing data. Four, we used the dy/dx first derivative of the effect on the dependent variable y of each independent variable x, a measure that is comparable for each variable and represented in Graphic 1 proportionally to line thickness, the thinnest for dy/dx=.10. Five, we assembled all the findings into the unified model in Graphic 1. All variables are significant at p<0.10, but the widths of lines reflect absolute values of comparable dy/dx effects between -1 and +1.
What is unusual about the structure of the results shown in Graphic 1 is that all of Brown and Eff's (2010) predictors of moral gods move to a position in the graph no longer directly predictive of the moral god variable; instead, six of their seven variables are mediated by our three new "Malthus event" variables. Their remaining variable, Food scarcity, is indirectly linked through a common effect, shared with HiGod4, on Islam, with its association to animal husbandry in dry environments. The blue dashed lines identify the three new independent variables expected to predict HiGod4 in the aftermath of a Malthusian crisis. Three of the core predictors of moral gods, that is, are those constructed from prior expectations about inequality of access to survival and wealth-resources as discussed by Turchin, Alexander, Swanson, Forrester et al., and others who consider social evolution from the perspectives of biology, sociology or empirical and quantitative history. These results help to justify our assumptions about the effects of Malthusan crises. The fourth of our predictors is water scarcity (No_rain_Dry, a weighted combination of Snarey's variable "No_rain" measuring scarcity of rain and an SCCS variable for dry ecozones). Also shown are the Islamic and Christian moral/ethical god religions, defined on a three-point scale of adoption, from none to partial to full, having the moral god variable as a common predictor (also true for Judaism and local religions of at least nine other societies). Other predictors of these specific religions are not shown. Finally, in addition to Snarey's variable for presence of Missions by the time of the ethnographic description, the Logdate variable (log of the date of description) is a strong predictor of moral gods for the rather obvious reason that the Islamic and Christian world religions tend to spread over time, and are subject to missionary or political efforts at conversion. The Missionization variable also serves as a control for covariance with other variables. Not shown in the graphic are the autocorrelation instruments such as spatial proximity, linguistic ancestry or having similar religions and what they add to total R2 which, like Logdate for predominantly world religions, is usually very high.
Conclusions
- ...[T]here is no greater irony than the fact that modern technological societies, whatever their degree of egalitianism or approach to the philosophical idea of morality ..., are teetering on the brink of world disaster as a result [of] their interactions with one another. The problem has become one of inducing beween and among societies the same processes of moralizing pressure and democratization that have developed so intricately within them" (Alexander 1987:105-106).
Also ironic is that, six years after 1987, correlations were used in a spurious fashion to create a fanciful theory attributed to Alexander using data from the SCCS and Ethnographic Atlas, namely that the evolution of moral religions was due to their role in the warfare of large societies in arms races. This "theory" was cited as late as 2008, before more careful study by Brown and Eff (2010) showed these conclusions to be false. Alexander's major concern, in fact, was how humans might come to learn a more generalized ethic that extends to other groups. Roes (1995:73), however, asked and concluded wrongly, arguing from weak correlational evidence and misguidedly citing Alexander, concluded as follows: "Why do moral rules exist? In Alexander’s (1987) view, 'moral systems have been designed to assist group members and explicitly not to assist the members of other competing groups', and morality emerged as a result of intense competition between human societies." There is no empirical evidence from comparative research to support these views, attributed to Alexander, who did not actually state either generalization in the first place.
Why do questions of causality matter? One obvious reason is to avoid spurious correlations. Another example of correlations that might be used to reach spurious conclusions is that between the SCCS moral gods variable of Swanson and the coding of evil eye beliefs by John M. Roberts (1976), of which David Levinson (1996: 65) has said:
- "Wherever it began, the [evil eye] belief goes back at least 5,000 [sic: actually, more like 4,000] years and was incorporated into the three major religions that developed in the Middle East--Judaism, Christianity, and Islam, although among many followers of the religions today it is regarded as a superstition."
- "Cross-culturally and throughout history, the evil eye belief is found mainly in cultures that produce social and physical goods that can be envied." "...following the development of agriculture and settled communities... Wealth distinctions were accompanied by status distinctions ... --royalty, upper classes, high castes, chiefs---who could be envied. Another characteristic of evil eye cultures is a strong belief in a high god active in human affairs. This belief, of course, is a central element of Judaism, Christianity, and Islam."
Part of the rationale for a diagram such as Graphic 1 is to help determine whether the correlation between MORAL GODS and EVIL EYE (R2=0.26) is spurious, due to common the coding of evil eye beliefs, rather than directly causal, in either direction. A major reason for multiple regressions with uniformly sized samples for variables with fully imputed missing data, and data quality dummy variable controls for variables with considerable missing data, all of which can be examined in a single graph, is to test whether correlations between variables are due to common the coding of evil eye beliefs. In Graphic 1, for example, the variables surrounding the variable of central interest (moral gods) include first and second order predictors (predictors of independent variables) and variables that imply by definition the central variables (Islamic and Christian societies). Analyzing how these might create spurious correlation is a first step toward eliminating what might be taken for causal effects. If the theory we have provided about secular cycles were correct, namely that ethical religions often tend to be adopted in the recovery period after secular Malthusian periods have passed their peak of conflict and inequality, then it might be plausible that another effect of "Malthusian events" is the creation of fatalistic beliefs in the possible recurrence of painful inequality blamed on the rich or prosperous. This would be a side effect, not "caused" by beliefs about religious sanctions against inequitable inequality.
The mechanics of establishing evidence for rejecting spuriousness among pairs of variables, both empirically coded and unmeasured, along with empirical correlations or regression dy/dx coefficients and correlations that are definitional or likely but unmeasured, was developed by the 1920s by Sewall Wright (1920, 1921, 1938) under the names of path analysis and structural equation modeling. These complex techniques, that require partial correlations or solutions of simultaneous equations, have remained beyond the capacities of most comparative researchers and their professional audiences, as in many fields of social science.
Pearl's (2000) formalizations of what is required for a series of directional effects among a set of variables to form a coherent graph that maintains the transitivity (or bidirectionality among variables at the base of a directed asymmetric graph) that is expected among variables that are causally related has made it possible to read a graphic format for intervariate relations as in Graphic 1 in ways that would be decidable in path analysis as to the relative strengths of direct and indirect relationships knowing which pairs of variables are independent, or conditionally independent given that certain other variables are held constant.
Examining graphs in this manner would help to
- Avoid fanciful "scenarios" built out of sequences of correlations (e.g., Roes 1995).
- Find mediating variables (e.g., the results of Brown and Eff 2010 when combined with our results)
- Test for residual direct effects x -> z given a mediating x -> y -> z (as in Graphic 1)
- Estimate likely effects fixing the values of control variables such as missing/nonmissing data for certain variables
In addition, as shown in Graphic 1, one of the advantages of comparing different empirical models for phenomena where adequate theory and measurable variables and controls exist is to abandon the idea that "better models," say in regression analysis, may wipe out "poor models." Instead, we may conceived that "mediators" may be found as our models advance, which is a crucial step in understanding posible paths of causality. Mediating processes may displace what are previously thought to be "direct effects" that become indirect as the result of the mediator. This acts to increase our ability to discover likely "causal alignments" of variables. In the case of a comparison of our findings about moral gods with those of Brown and Eff (2010), it is not the case that one is right and the other necessarily wrong. What we have learned concerns the possibility that one or more hidden or as yet unmeasured variables coming out of Malthusian events and dynamical secular cycles processes affect the social, economic and political relationships that affect the evolution and adoption of moral/ethical gods. They may do so in a more direct and processual manner -- as in the concept of an exchange system with a variable dynamic -- than static variables such as those chosen by Brown and Eff in their study. Graphic 1 suggests that the more dynamical exchange-relationship variables of our model may be the more proximal variables affecting the evolution of ethical religions, but they are affected in turn by the more static variables that we see in the Brown and Eff model.
Pearl might call our Graphic 1 a "causal graph" but causal graphs are not necessarily "causal" in terms of what might ordinarily be expected. They are a means of testing for or against causal relationship in a limited network of variables. The longer the chain of potential causalities, the more that multiplicative effects varying on a dy/dx scale between -1 and +1 will diminish in magnitude. Empirical studies of presumably linear effects, then, will be limited in their extent, with proximal and second order indirect effects stronger and easier to detect.
Relating regression techniques to causal graphs can be challenging. In our Graphic 1, for example, our "Malthus event" variables give a comparable model to Brown and Eff (2010) in terms of R2 but the dy/dx ratios of these ratios are rather small although siignificance. This may follow from the fact that these variables are rather specifically defined. As Routon and Eff (2011) note, the richer the statistical measurements, the better the capacities of multiple regression rather than categorical analyses or dichotomous measurement in the regression context.
References
- Fletcher,Jesse B., Jacob Apkarian, Robert A. Hanneman, Hiroko Inoue, Kirk Lawrence, Christopher Chase-Dunn. 2011. Demographic Regulators in Small-Scale World-Systems. Structure and Dynamics 5#1. pdf.
- Johnson, Dominic D. P. 2008. Gods of War: The Adaptive Logic of Religious Conflict. Pp. 111-118, In: The Evolution of Religion: Studies, Theories, and Critiques, eds., J Bulbulia, R Sosis, C Genet, R Genet, E Harris, and K Wyman (eds), Collins Foundation Press.
- Korotayev, Andrey. 2003. World Religions and the Social Evolution of the Old World Oikumene Civilizations. Lewiston, New York: Edwin Mellen.
- Levinson, David. 1996. Religion: A Cross-Cultural Dictionary. Oxford: Oxford University Press.
- Modelski, G. 1987. Long Cycles in World Politics. Seattle: University of Washington Press.
- Modelski, G., and William R. Thompson 1995. Leading Sectors and World Powers: The Coevolution of Global Economics and Politics. Columbia: Columbia University Press.
- Murdock, George P. 1967. Ethnographic Atlas. Pittsburgh: Pittsburgh University Press.
- Pearl, Judea. 2000 (2nd edition 2009). Causality: Models, Reasoning, and Inference. Cambridge: Cambridge University Press.
- John M. Roberts. 1976. Belief in the evil eye in world perspective. In The Evil Eye, ed. C. Maloney, pp. 223-78. New York: Columbia University Press.
- Roes, Frans L. 1995. The Size of Societies, Stratification, and Belief in High Gods Supportive of Human Morality. Politics and the Life Sciences 14(1):73-77.
- Abstract. Two hypotheses about belief in high gods supportive of human morality were tested with data from the Ethnographic Atlas and the Standard Cross-Cultural Sample. A significant positive relation between the size of societies and such a belief is demonstrated, and this relation appears to be independent of both regional differences and differences in stratification of the societies. On the other hand, stratification itself is also significantly related with the belief in high gods supportive of human morality, but this relation could not be shown to be independent of regional differences or differences in size.
- DRW: One hypothesis was that high gods are correlated in the Atlas with larger societies, which failed to replicate in two of the six world regions (Africa and Circum-Mediterranean, n=82 & 63), and in two other regions were only significant at p<.01 (East Asia and South America, n=68 & 52). "The Alexander hypothesis results from the idea that (a) societies increase in size as a result of intersocietal conflicts and competition, (b) conflicts of interests between members of a society are more likely in larger societies, (c) members of large societies have a common interest in preventing a disolution of the [sic] own society, (d) high gods supportive of human morality serve this common interest in discouraging intrasocietal conflicts, and (e) the belief in such gods would therefore be socially valued in larger societies. The testable hypothesis is that one should find a positive relation between society size and belief in high gods supportive of human morality. This hypothesis found clear support in this study."
- For the other hypothesis of Irons, Stratification and High Gods failed to replication in five of the six regions.
- with data from the SCCS. If all regions are taken together, there is a p < .001 positive relation between "society size" and "high gods," and p < .01 positive relation between "stratification" and "high gods." Within the stratified societies, there is a p < .01 positive significant relation between "society size" and "high gods." No further significant relations were found, which is possibly due to the small numbers of cases. These findings point in the same direction as the analysis of the Ethnographic Atlas data:
- Roes, Frans L., and Michel Raymond. 2003. Belief in Moralizing gods. Evolution and Human Behavior 24(2):126-135. pdf Raymond: Institute of Evolutionary Sciences (CNRS-UMR 5554)
- Abstract: According to Alexander's [Alexander, R. D. (1987). The biology of moral systems. New York: Aldine de Gruyter] theory of morality, human social groups became large as a result of between-group competition over preferred habitats and resources, but although larger social groups are more successful in competition, they also experience more pressures to fission. Morality unites a society by limiting infringements upon the rights of other society members, so if larger societies are indeed more likely to split, then those that remain intact may be expected to have more effective inviolable moral rules, such as those imposed by moralizing gods. Cross-cultural analyses support this line of thought: more competition between societies is found in environments rich in resources and larger societies tend to occupy these environments; large societies engage in external conflicts at higher rates and are more often characterized by beliefs in moralizing gods. An additional explanation is briefly discussed, and we speculatively picture the historical chain of events giving rise to a belief in moralizing gods.
- Snarey, John R. 1996. The natural environment's impact upon religious ethics: a cross-cultural study. Journal for the Scientific Study of Religion 35(2): 85-96.
- Turchin, Peter. 2005 War and Peace and War: The Rise and Fall of Empires. Pi Press. (paperback 2007)
- Turchin, Peter. 2005 Dynamical Feedbacks between Population Growth and Sociopolitical Instability in Agrarian States. Structure and Dynamics 1(1): 49-69.
- Turchin, Peter. 2008. Arise Cliodynamics. Nature 454: 34-35
- Turchin, Peter. 2009a. "Long Term Population Cycles in Human Societies." in This Year in Ecology and Evolutionary Biology 2009. Edited by R. S. Ostfeld and W. H. Schlesinger (Annals of the New York Academy of Sciences).
- Turchin, Peter, and Sergey A. Nefedov. 2009. Secular Cycles. Princeton: Princeton University Press.
- Turchin, Peter. 2010. Warfare and the Evolution of Social Complexity: A Multilevel-Selection Approach Structure and Dynamics 4#3.
- Swanson, Guy L. 1960. Birth of the gods. Ann Arbor, Mich., University of Michigan Press.
- White, Douglas R. 2009. Innovation in the Context of Networks, Hierarchies, and Cohesion. Pp. 153-193 in, Complexity Perspectives in Innovation and Social Change. D.Lane, D.Pumain, S. van der Leeuw and G.West (eds). Berlin: Springer (Methodos series).
- White, Douglas R., L. Tambayong, and N. Kejžar)2008 Oscillatory dynamics of city-size distributions in world historical systems. In, G. Modelski, T. Devezas and W. Thompson, eds. Globalization as Evolutionary Process: Modeling, Simulating, and Forecasting Global Change. pp. 190-225. London: Routledge. http://intersci.ss.uci.edu/wiki/pw/ModelskiCh9WTK.pdf
- Wolf, Eric. 1982. Europe and the “People without History. New York: Barnes & Noble.
- [[Wikipedia:Sewall Wright|Wright, Sewall. 1921. Correlation and causation. J. Agricultural Research 20: 557–585.
- Wright, Sewall. 1923. The theory of path coefficients: A reply to Niles’ criticism. Genetics 8: 239–255.
- Wright, Sewall. 1934. The method of path coefficients. Annals of Mathematical Statistics 5: 161–215.
Temporary notes and Afterthoughts
Evil eye as a confirmatory indicator of inequality and past crises for moral god religions
See: Tables for Moral Gods and Evil Eye
Afterthoughts: Effects of and Controls for Autocorrelation
Distance autocorrelation has a major effect itself and on other variables. Language (common ancestors) has little or no effect. Adjacency of coreligionists has the main effect ok eliminating any effect of AnimXbwealth and FxcmtyWages, which are associated with Islam and Christianity, respectively.
Afterthoughts: Measuring autocorrelation effects on different independent variables
get results of o <- lm(WY~WX) results for indep variables where WY is the depvar
> fit_instrument <- function(y,X,W,idxs,post_select=FALSE) {
+
+ Y <- as.matrix(y)
+ if (!is.null(idxs) && !post_select) {
+ X = X[idxs,]; Y = Y[idxs]; W = W[idxs,idxs]
+ }
+
+ WX <- W%*%X
+ WY <- W%*%Y
+
+ if (!is.null(idxs) && post_select) {
+ X = X[idxs,]; Y = Y[idxs]; W = W[idxs,idxs]; WY = WY[idxs]; WX = WX[idxs,]
+ }
+
+ o <- lm(WY~WX)
+
+ #--the fitted value is our instrumental variable--
+ y_hat <- fitted(o)
+
+ #--keep R2 from this regression--
+ r2 <- summary(o)$r.squared
+
+ return(list(y_hat=y_hat,r2=r2,wy=WY))
+ }
> o <- lm(WY~WX)
Error in eval(expr, envir, enclos) : object 'WY' not found
Afterthoughts: Create a temporal autocorrelation W matrix that can be dot*multiplied by any other W?
logdate=log(sccs$v838)-6
Wtime=matrix(data = 0, nrow = 186, ncol = 186)
Ld= my_sccs$logdate
for(i in 1:186) {
for(j in 1:186) {
d= Ld[i]-Ld[j]
Wtime[i,j]=abs(d)
}}
Wtime=Wtime/rowSums(Wtime) # check that rows sum to 1: Wtime[1:3,1:3] Wtime[184:186,184:186] sum(Wtime[55,1:186])
#after dot product with another W matrix, renormalize row sums to 1 #could do time and space #could do time and religion #could do time and Wlink distance (Murdock-White alignment) # does time and language make and sense??
Appendix: Nine EduR Models and Malthusian events
Of the variables in Brown and Eff (2010): when the HIDDEN VARIABLES model adds AnimXbwealth: Caststrat, Anim, PCsize, Pcsize2, and PCAP are NO LONGER SIGNIFICANT (AnimXbwealth is the mediator; for PCAP one of two mediators, the other being FxCmtyWages). Same for Eextwar, for which SuperjhWriting is the mediator. Foodscarc also loses significance.
What we discover here from the enlarged causal graph, which we would not know from separate regression models, is that the Brown and Eff (2010) model is not "wrong" because outperformed by the "hidden variable" (effects of Malthusian events) model, with new "proximal variables" measuring potentials for perceived injustice in amplified inequalites), is that the Brown-Eff variables are "distal variables" mediated by more proximal ones, and having indirect transitive effects. One of the purposes of "causal graph" constructions is precisely this: to show that some effects are not direct but mediated.
HiGod4
MOVE BELOW : (adding money v17 >3 or >4 to the AnimXbwealth and FxCmtyWages predictors of moral gods improves the model and reduces the significance of Missions to p=0.12 in the first case but restores the significance of Missions to p=0.09. Adding money to the SuperjhWriting dependent variable improves the R2 by .03). See paragraph in the "Testing hypotheses" section above.
See also: EduR-2 Ethical gods: no control for Logdate or missions EduR-1.5 HiGod=moralgod=sccs$v238 source("examples/create/create_EduR_1/create_EduR_1.5DistB_EffHiddenVariables7.R") depvar=HiGod4 coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded (Intercept) -2.094 NA NA NA 11.502 27259.941 0.00069607 NA NA distance 0.639 2.000 1.278 0.426 15.007 32700.369 0.00010732 1.838 0.426 SUM http://bit.ly/uyMItW evileye 0.367 1.000 0.367 0.122 4.467 82458.637 0.03456847 1.568 0.122 1.000 SuperjhWriting 0.057 9.000 0.515 0.172 4.170 28300.941 0.04116461 1.373 0.172 1.000 Protection of written laws? Exposure to excessive taxes? No_rain_Dry 0.213 3.000 0.639 0.213 7.575 160336.427 0.00591998 1.398 0.213 1.000 AnimXbwealth 0.058 10.000 0.584 0.195 3.032 4720907.873 0.08162571 1.378 0.195 1.000 FxCmtyWages 0.310 1.000 0.310 0.103 3.799 2243.182 0.05141511 1.304 0.103 0.602 (THIS AND ALL MOST OTHER VARIABLES ARE FULLY IMPUTED, including miss cases for depvar) FxCmtyWages 0.292 1.000 0.292 0.097 3.067 484.689 0.08052533 1.340 0.097 0.602 CONDITIONAL IMPUTATION HAS A LESSER EFFECT. dy/dx =0.08 shows a small missing data effect Missions 0.247 1.000 0.247 0.082 2.845 185088.429 0.09166207 1.161 0.082 1.000 SUM 0.887 logdate 0.824 2.883 2.376 0.792 7.106 29742.505 0.00768691 1.122 0.792 1.000 Note: releW wipes out AnimXbwealth andFxCmtyWages 0.00104675 1.501 0.562 Train R2:final model Train R2:IV_distance 0.4343500 0.9798926 Fstat df pvalue RESET 2.031 20565.464 0.154 Wald.on.restrs 0.991 91.431 0.322 NCV 0.020 73085.562 0.888 SW.normal 0.673 35792.521 0.412 lag..distance 2.484 183856.031 0.115 MISSING DATA LITTLE EFFECT BUT FULL IMPUTATION RAISES VALUES OF FxCmtyWages coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded (Intercept) -2.118 NA NA NA 11.402 11505.805 0.00073608 NA NA 0.000 distance 0.640 2.000 1.280 0.427 14.667 24534.978 0.00012859 1.859 0.427 0.000 FxCmtyW_mis 0.018 1.000 0.018 0.006 0.016 39747.129 0.89955443 1.041 0.006 1.000 SuperjhWriting 0.058 9.000 0.526 0.175 4.078 7247.518 0.04348189 1.428 0.175 1.000 No_rain_Dry 0.209 3.000 0.626 0.209 7.186 113746.192 0.00734755 1.397 0.209 1.000 AnimXbwealth 0.059 10.000 0.589 0.196 3.041 5027793.977 0.08119173 1.385 0.196 1.000 FxCmtyWages 0.273 1.000 0.273 0.091 2.590 337.266 0.10849776 1.365 0.091 0.602 evileye 0.376 1.000 0.376 0.125 4.680 94676.089 0.03052514 1.561 0.125 1.000 Missions 0.252 1.000 0.252 0.084 2.903 103456.602 0.08843098 1.167 0.084 1.000 logdate 0.831 2.883 2.395 0.798 6.935 5563.895 0.00847495 1.138 0.798 1.000 THE POSITIVE CORRELATION R=.08 IS LOW BUT MEANS IMPUTED VALUES HIGHER, CODED VARIABLES LOWER THAN IMPUTED 0 0.125 0.25 0.375 0.5 0.625 0.75 0.875 1 0 72 0 0 0 0 0 0 0 40 N=112 Coded variable 1 24 7 3 7 3 5 4 5 16 N=074 Missing data estimated F 96 7 3 7 3 5 4 5 56 N=186 table(output$FxCmtyWages) > cor(output$FxCmtyW_mis,output$FxCmtyWages) [1] 0.08518986 Presence of Missing data CORRELATED with FxCmtyWages real and estimated > my_sccs$FxCmtyWages [1] 0 0 0 NA NA 0 NA 0 NA NA 1 0 0 NA NA 0 NA NA 0 NA 0 0 0 NA 0 0 NA 0 1 1 [31] NA NA NA 0 NA 0 1 NA NA NA 0 1 1 NA NA NA 1 NA NA 1 1 NA NA 1 NA NA 1 0 1 NA [61] 0 1 NA NA 0 0 NA 1 NA NA 1 0 NA NA 1 NA 0 1 0 0 0 NA NA 1 0 0 NA NA NA 0 [91] 0 0 1 1 1 NA NA 0 1 1 NA NA 0 0 NA NA NA 1 1 1 1 NA NA 1 NA 1 1 0 0 NA [121] 0 0 0 0 0 NA 0 NA 0 0 NA 0 0 NA 0 0 0 0 0 0 0 0 NA 0 0 NA 0 NA 0 0 [151] 0 1 1 NA NA 1 NA 1 0 1 NA 0 0 NA 1 1 1 0 0 NA NA 1 0 NA 0 NA NA NA 0 0 [181] NA NA 0 1 NA 0 > my_sccs$FxCmtyW_mis [1] 0 0 0 1 1 0 1 0 1 1 0 0 0 1 1 0 1 1 0 1 0 0 0 1 0 0 1 0 0 0 1 1 1 0 1 0 0 1 1 1 0 0 0 1 1 1 0 1 1 0 [51] 0 1 1 0 1 1 0 0 0 1 0 0 1 1 0 0 1 0 1 1 0 0 1 1 0 1 0 0 0 0 0 1 1 0 0 0 1 1 1 0 0 0 0 0 0 1 1 0 0 0 [101] 1 1 0 0 1 1 1 0 0 0 0 1 1 0 1 0 0 0 0 1 0 0 0 0 0 1 0 1 0 0 1 0 0 1 0 0 0 0 0 0 0 0 1 0 0 1 0 1 0 0 [151] 0 0 0 1 1 0 1 0 0 0 1 0 0 1 0 0 0 0 0 1 1 0 0 1 0 1 1 1 0 0 1 1 0 0 1 0
Evileye
EduR-3 source("examples/create/create_EduR_1/create_EduR_3HiddenEvileye1.R") depvar=Evileye coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded (Intercept) -0.226 NA NA NA 1.545 2189609.35 0.21393330 NA NA 0.000 distance 0.776 2.000 1.551 1.551 30.036 1680416.35 0.00000004 1.740 1.551 0.000 SUM http://bit.ly/rYU2uG Higod4 0.070 3.000 0.210 0.210 5.885 47891915.69 0.01527369 1.551 0.210 1.000 SuperjhWriting -0.029 9.000 -0.263 -0.263 5.195 345444.37 0.02265261 1.718 0.263 1.000 No protection of written laws? Exposure to excessive taxes? PCsize.2 0.016 28.847 0.464 0.464 5.356 20286533.61 0.02065154 1.368 0.464 1.000 caststrat 0.101 3.000 0.303 0.303 5.165 55558.57 0.02304814 1.202 0.303 0.973 anim 0.035 9.000 0.317 0.317 4.546 114572.43 0.03299123 1.649 0.317 1.000 ~ p-0.14 AnimXbwealth Missions 0.118 1.000 0.118 0.118 4.114 1155271.09 0.04251789 1.087 0.118 1.000 SUM 1.675 Train R2:final model Train R2:IV_distance 0.4298996 0.9832074 Fstat df pvalue RESET 7.992 73352.59 0.005 log anim? Wald.on.restrs 0.048 35961.58 0.827 NCV 0.891 605237.75 0.345 SW.normal 28.965 16361025.50 0.000 lag..distance 1.220 249474.05 0.269 table(sccs$v1188,sccs$v238) #evil eye, moral god code of Murdock 1 2 3 4 1 1 0 0 0 2 22 15 3 2 3 24 10 1 3 4 8 11 1 4 5 3 3 1 1 6 5 2 1 4 7 3 0 4 6 8 2 6 2 20 R2~0.28=correlation with Snarey HiGod4
Malthusian event
This is the only unmeasured variable, shown by the fifth node up from the bottem of the long row of color yellow that radiates dashed blue lines. Societies in the SCCS are often located in a larger polity or empire that goes through Turchin and Nefedov (2007) "secular cycles" which have periods of overpopulation relative to resources. These period usually lead to exaggerated inequalities because those who own productive property have the extra value-advantages of controlling scarce resources and paying excessively lower wages to labor due to surplus labor. In societies affected causally by a Malthusian event, a period of recovery from inequalities perceived as "unethical" or beyond acceptable limits may be followed by acceptance of a moralizing high god religion, while also leaving a lasting fatalism about the dangers of inequalities expressed in evil eye beliefs. What is not shown in the graph is the high correlation between Moralizing god beliefs and evil eye beliefs. These are thought not to be causally related except through common causal variables. Causal graph analysis may show that this correlation disappears when controlling for the variables that mediate the effects of Malthusian event. Those mediating variables are AnimXbwealth, FxCmtyWages and SuperjhWriting, each involving a greater potentiality for inequalities to develop in Malthusian periods. Because we have not yet measured which societies have experienced Malthusian events we refer to them and to the mediating conditions for inequality as "hidden variables."
AnimXbwealth
EduR-11 AnimXbwealth=((sccs$v208==1)*1)*sccs$v206 CONDITIONAL dep_var=log(1+((sccs$v208==1)*1)*sccs$v206) depvar=AnimX coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded t.cor (Intercept) -0.699 NA NA NA 23.624 295.163 0.00000191 NA NA Wpath 0.428 2.000 0.857 0.357 12.041 335.824 0.00058829 1.752 0.357 0.000 0.565 SUM http://bit.ly/uFafTC no_rain_dry 0.112 3.000 0.335 0.140 6.646 1542.228 0.01002830 1.174 0.140 1.000 0.348 fratgrpstr 0.254 4.000 1.016 0.424 35.132 143.800 0.00000002 1.774 0.424 0.441 0.617 PCsize -0.050 8.061 -0.405 -0.169 3.664 932.198 0.05589641 1.410 0.169 1.000 0.060 PCsize.2 -0.018 28.847 -0.524 -0.218 3.126 309.153 0.07805015 1.240 0.218 1.000 -0.150 SUM 0.951 Train R2:final model Train R2:IV_Wpath 0.4852531 0.8356619 Fstat df pvalue RESET 3.370 48.435 0.073 Wald.on.restrs 0.402 2327.335 0.526 NCV 13.957 34.109 0.001 heteroscedasdicity SW.normal 5.095 574.594 0.024 lag..Wpath 0.148 872.563 0.700
The "Malthusian event" posited here is a function of the fact that the highest levels of dependence on pastoralism, which almost always involves horses, camels, donkeys, mules, and often sheep or goats, when combined with Bridewealth (including animals), creates a system of socio-economic exchange in which some clans or lineages can accumulate larger herds and engage in long distance trade. Both the herds at home and the riches brought in by trade are forms of productive wealth accumulation. While alliances through bridewealth or brides are often reciprocally balanced, there remains a strong possibility that wealth will support population growth and lead to a Malhusian event amplified by advantages accruing to the richest clans or lineages, and subordinating others to herders or warriors in the service of the wealthier. This can produce a crisis of perceived excessive inequality, social conflict, collapse of the growth phase, and support, in the following period of recovery, openness to conversion to an "ethical high god" religion as in the example of Islam.
FxCmtyWages.R
see EduR-9 FxCmtyWages for definition of the FxCmtyWages variable and see User_talk:Qingzi_Huang#FxCmtyWages latest Results by_Niu for proposed variabes - add:
popdens, plow, Neg-v733 Sexc2Agri (FemContrib2Agric) dvar=FxCmtyW coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded t.cor (Intercept) 0.003 NA NA NA 0.000 461.332 0.99441588 NA NA EduR-9 distance 1.099 2.00 2.197 1.099 27.312 562.339 0.00000024 1.005 1.099 0 0.388 PCAP -0.031 19.92 -0.618 -0.309 2.660 383.289 0.10371767 1.191 0.309 1 -0.051 Agri Potential Drop given Missing data on depvar ecorich -0.220 2.00 -0.441 -0.220 5.674 1000.706 0.01740264 1.193 0.220 1 -0.127 Train R2:final model Train R2:IV_distance 0.1268772 0.9356842 Fstat df pvalue RESET 0.365 34.748 0.550 Wald.on.restrs 0.431 35.633 0.516 NCV 0.275 1376.116 0.600 SW.normal 13.852 241.366 0.000 lag..distance 1.416 651.773 0.235 EFFECT OF MISSING DATA - ELIMINATE PCAP coef xrange dx dx/dy Fstat ddf value VIF |dx/dy| coded t.cor (Intercept) 0.071 NA NA NA 0.029 1758.113 0.86446421 NA NA 0 0.000 distance 1.072 2.00 2.143 1.072 30.617 2514.352 0.00000003 1.011 1.072 0 0.405 MisFxCmtyW 0.087 1.00 0.087 0.043 0.341 43.559 0.56227595 1.014 0.043 1 0.062 PCAP -0.033 19.92 -0.667 -0.333 2.229 38.686 0.14354883 1.199 0.333 1 -0.035 Drop given Missing data on depvar ecorich -0.260 2.00 -0.520 -0.260 5.628 43.121 0.02221035 1.198 0.260 1 -0.183
SuperjhWriting
User:Wei_Wang#results_of_round1 Xian EduR-10 depvar=SuperjhWriting=sccs$v237*(1+((sccs$v149>=3)*1)) gives the same result coef range effect ratio Fstat ddf pvalue VIF abs.ratio %coded tot.cor part.cor (Intercept) -0.408 NA NA NA 0.553 2667.310 0.45712200 NA NA 0.000 0.000 0.000 distance 0.041 2 0.083 0.009 0.075 532.097 0.78451073 1.626 0.009 0.000 0.451 0.437 caststrat 0.499 3 1.496 0.166 5.050 4430.731 0.02468206 1.166 0.166 0.973 0.359 0.358 eeextwar 0.054 16 0.857 0.095 6.023 767.097 0.01434110 1.036 0.095 0.828 0.217 0.218 socplex 0.335 5 1.674 0.186 11.546 108.593 0.00095015 1.843 0.186 0.500 0.590 0.586 plow 2.929 1 2.929 0.325 37.451 1225.800 0.00000000 1.693 0.325 1.000 0.654 0.650 popdens 0.293 4 1.171 0.130 6.693 302.254 0.01014357 1.460 0.130 1.000 0.496 0.495 Train R2:final model Train R2:IV_distance 0.5718715 0.9740887 Fstat df pvalue RESET 1.574 132.228 0.212 Wald.on.restrs 0.190 2933.401 0.663 NCV 26.117 466.649 0.000 SW.normal 23.865 672.428 0.000 lag..distance 4.124 10251.022 0.042
Islamic
EduR-5 Islam=((sccs$v2002==2)*2)+((sccs$v2002==4)*1) depvISLAMcoef xrange dx dy/dx Fstat ddf value VIF |dy/dx| coded t.cor p.cor (Intercept) -0.673 NA NA NA 34.492 113.805 0.00000004 NA NA 0.000 0.000 0.000 distance 0.602 2.000 1.204 0.602 18.604 5470.963 0.00001637 1.859 0.602 0.000 0.512 0.225 HiGod4 0.087 3.000 0.260 0.130 6.500 2445.160 0.01085060 1.429 0.130 1.000 0.481 0.387 SuperjhWriting 0.051 9.000 0.455 0.227 7.436 2097.158 0.00644552 2.383 0.227 1.000 0.346 0.252 PCsize2 -0.017 28.847 -0.497 -0.248 4.073 8023.796 0.04361172 1.393 0.248 1.000 -0.007 -0.075 plow -0.218 1.000 -0.218 -0.109 3.114 3490.681 0.07771609 2.014 0.109 1.000 0.204 0.073 AnimXbwealth 0.050 10.000 0.497 0.248 7.830 4927.431 0.00515748 1.545 0.248 1.000 0.498 0.398 caststrat 0.134 3.000 0.403 0.201 4.687 72.787 0.03367567 1.221 0.201 0.973 0.338 0.294 foodscarc 0.124 4.000 0.495 0.247 14.447 36.910 0.00052271 1.042 0.247 0.774 0.298 0.342 Train R2:final model Train R2:IV_distance 0.5129229 0.9669130 Fstat df pvalue RESET 31.719 76.411 0.000 Wald.on.restrs 2.450 148.974 0.120 NCV 83.315 114.724 0.000 SW.normal 18.534 245.153 0.000 lag..distance 0.178 4320.841 0.673
ISLAM OR XIAN - STRONG IslamXian=((sccs$v2002==2)*1)+((sccs$v2002==3)*1)
dep ISLAM XIAN coef xrange dx dy/dx Fstat ddf value VIF |dy/dx| coded t.cor p.cor
(Intercept) -0.561 NA NA NA 37.623 1674.999 0.00000000 NA NA 0.000 0.000 0.000
distance 0.738 2.000 1.475 0.738 38.034 6254.112 0.00000000 1.423 0.738 0.000 0.566 0.201
HiGod4 0.141 3.000 0.422 0.211 17.698 28851.930 0.00002596 1.392 0.211 1.000 0.536 0.416
SuperjhWriting 0.041 9.000 0.365 0.183 7.360 5399.573 0.00669172 1.553 0.183 1.000 0.378 0.260
PCsize2 -0.018 28.847 -0.518 -0.259 4.637 30963.463 0.03130616 1.322 0.259 1.000 0.050 -0.043
foodscarc 0.108 4.000 0.433 0.216 16.238 505.735 0.00006445 1.026 0.216 0.774 0.271 0.322
Train R2:final model Train R2:IV_distance
0.4912084 0.9660026
Fstat df pvalue
RESET 31.080 203.268 0.000
Wald.on.restrs 0.955 21.661 0.339
NCV 95.139 125.668 0.000
SW.normal 26.839 5696.623 0.000
lag..distance 0.158 3196.139 0.691
Christian
EduR-12 Xian=((sccs$v2002==3)*2)+((sccs$v2002==5)*1) dVar Xian coef xrange dx dy/dx Fstat ddf value VIF |dy/dx| coded t.cor p.cor (Intercept) -0.133 NA NA NA 2.598 15772.985 0.10703772 NA NA 0.000 0.000 0.000 distance 1.128 2 2.255 1.128 22.548 1437.587 0.00000225 1.054 1.128 0.000 0.394 0.079 HiGod4 0.085 3 0.255 0.128 8.570 13087.928 0.00342256 1.263 0.128 1.000 0.217 0.191 eeextwar -0.010 16 -0.165 -0.083 4.817 1085.259 0.02839148 1.010 0.083 0.828 -0.184 -0.184 AnimXbwealth -0.043 10 -0.429 -0.215 9.030 44606.163 0.00265744 1.206 0.215 1.000 -0.137 -0.156 FxCmtyWages 0.213 1 0.213 0.107 8.707 129.722 0.00376346 1.099 0.107 0.602 0.341 0.301 Train R2:final model Train R2:IV_distance 0.2627975 0.9433891 Fstat df pvalue RESET 37.063 177.159 0.000 Wald.on.restrs 5.923 88.640 0.017 NCV 93.816 2913.007 0.000 SW.normal 35.924 12062.049 0.000 lag..distance 1.003 22373.503 0.317
Fratgrpstr
EduR-6 fratgrpstr=sccs$v570 dv Fratgrpstr coef range effect ratio Fstat ddf pvalue VIF abs.ratio %coded tot.cor part.cor (Intercept) 1.304 NA NA NA 6.992 29801.445 0.00819052 NA NA 0.000 0.000 0.000 distance 0.365 16.080 5.861 1.465 10.794 4018883.741 0.00101805 1.932 1.465 1.000 0.679 0.507 PCAP 0.052 8.061 0.419 0.105 3.201 557150.715 0.07361080 1.639 0.105 1.000 0.020 -0.006 PCsize -0.227 3.000 -0.682 -0.170 12.480 1666351.723 0.00041134 1.798 0.170 0.939 -0.568 -0.557 caststrat -0.296 4.000 -1.185 -0.296 4.814 21236.442 0.02823830 1.190 0.296 1.000 0.164 0.113 popdens 0.096 9.000 0.866 0.217 2.088 370353.447 0.14847176 1.748 0.217 1.000 0.443 0.434 anim -0.133 10.000 -1.333 -0.333 2.775 264621.894 0.09576597 5.846 0.333 1.000 0.583 0.517 AnimXbwealth 0.363 1.000 0.363 0.091 36.212 70667.445 0.00000000 4.134 0.091 1.000 0.736 0.701 milk 0.417 4.000 1.670 0.417 2.443 682919.177 0.11804105 2.532 0.417 1.000 0.626 0.559 frqintwar 0.129 2.000 0.258 0.064 1.279 6161.768 0.25818940 1.082 0.064 0.817 0.443 0.434 ecorich 0.330 2.000 0.659 0.165 6.521 255559.124 0.01065910 1.495 0.165 1.000 -0.128 -0.111 Train R2:final model Train R2:IV_distance 0.7927921 0.9880978 Fstat df pvalue RESET 1.910 958828.73 0.167 Wald.on.restrs 0.020 45064.59 0.887 NCV 1.455 183178.54 0.228 SW.normal 1.226 101289.91 0.268 lag..distance 1.707 25126.34 0.191 DEP_VAR=MORALGODS
EduR models and their wiki pages
MisFxCmtyWages undersampling is studied in EduR-1.5 MisFratgrpstr study of undersampling is needed EduR-6 Two variables are symmetric predictors but can be discounted:
- For Evileye the R2s=.21, .122, with HiGod the slightly better depvar but both can be discounted although p=.015, .035
- For Fratgrpstr R2s=.091, .424, with AnimXwealth by far the better depvar, so discount depvar Fratgrpstr although its significance is p<.00000001.
- EduMod|| EduR | My sccs columns
- EduR-0 largely OBSOLETE models for Macs.
- EduR-1 Moral gods II with Pastoral Exchange
- EduR-1.0 R2=.11 Value of Children
- EduR-1.1 R2=.33-43) HiGods (Moral, Ethical) Brown and Eff (2010)
- EduR-1.2 LangOnly/DistanceOnly Brown and Eff 2010
- EduR-1.3 Brown and Eff distance AND language
- EduR-1.4 R2=.40 correct v235 as substitute for v237 Moral gods mixed model
- EduR-1.5 R2=.43 Hidden variables for Moral gods (foodscarc=sccs$v1685 p=.06 cond. p=0.13 full?)
- EduR-1.6 (incomplete) DistLangBrown
- EduR-1.7 R2=.41 Moral gods II - mix of Brown-Eff variables and Pastoral Exchange
- EduR-1.8 R2=.43 Islam alone (and Missions) reduce significance of FxCmty
- EduR-1.9 Missions, Time as control variable treatments
- EduR-2 R2=.43 Ethical gods (Hidden variables 1.5 dy/dx rises 33% for FxCmtyWages without Logdate, Missions
- EduR-3 R2=.43 Evil eye
- EduR-4 R2=.43 Money
- EduR-5 R2=.51 Islam model (graphic 9), prototype from*EduR-11
- EduR-6 R2=.80! Fratgrpstr (graphic 29) fn(AnimX, milk, ecorich, PCcap NEG: PCsize, caststrat, anim)
- EduR-7 General polygyny
- EduR-8 R2=.41 (1,2) Moral gods II with Pastoral Exchange (mixed like R-7)
- EduR-9 R2=.18 FxCmtyWages (graphic 4) (now graphic 1?) (Capitals=combined variables)
- EduR-10 R2=.57 SuperjhWriting (graphic 5)
- EduR-11 R2=.51 Pastoralexch (graphic 3) AnimXbwealth ReligWmatrix
- EduR-12R2=.26 Xian-Christian (graphic 10)
- EduR-13 socialclass (graphic 7)
- EduR-14 Caststrat (graphic 25)
- EduR-15 Foodscarc (graphic 21)
- EduR-16 PCsize (graphic 23) (not PCsize.2, graphic 24)
- EduR-17 Eextwar (graphic 26)
- EduR-18 No_rain_Dry (graphic 6)
- Eff_and_Dow_2009#Program2009_for_PC
- EduR-1.5#Comparison_of_Imputation_Methods.2C_partial_vs_FULL is Foodscarc defined correctly? YES, analysis redone
Misc
Moral god codes
Swanson (1960) found only five of his 39 primitive and ancient cultures (Cuna, Ancient Egyptians, Israelites, Nuer, Yahgan!?) had moral gods coded with reasonable certainty (13%, compared to 23% in the SCCS) societies, including more complex societies. His hypothesis was Durkheimian: the hierarchical aspect of monotheistic gods should reflect levels of hierarchy in the constitutional and secular orders of government and political structure. Our study finds correlation to be weak and only a marginal effect with many other predictors.
Political competition, government stability, and religion contribute to the causes and effects of secular cycles.
FOOTNOTE 1 Ethnographer Martin Gusinde, for example, a student of Father Wilhelm Schmidt (who taught that all societies had high gods), attributed an invisible, omnipotent, and omniscient spirit living in heaven, beyond the stars with astute powers of observation and strict disciplinary enforcement over his "Yahgan children" of Tierra del Fuego, who become extinct after long contacts with Europeans and their priests. In fairness to the Yahgan, we coded only this one SCCS case as "missing data" after reviewing Gusinde's writings and imputed this case as with other missing data.
Additional graphics
Scraps of things
Alexander, our first paragraph: (including Campbell) Campbell (1972) noted that "The behavioral dispositions which produce complex social interdependence and self-sacrificial altruism must instead be products of culturally evolved indoctrination, which has had to counter self-serving genetic tendencies."
Warning message: package 'foreign' was built under R version 2.13.2
Testing the red herring hypotheses (drop?)
"Alexander's hypothesis" is the name given by Roes (1995:73-74; 77) to a string of conjectures that he attributed to Alexander as a coherent theory: 1) 2) 3) 4).
- With the growth of a society, the number of conflicting interests increases <--- below
- which results in an increased potential for internal conflicts
- societies increase in size as a result of (successful) intersocietal conflicts and competition
- intrasocietal conflicts of interests are more likely in larger societies
- members of large societies have a common interest in preventing a dissolution of their own society
- high gods supportive of human morality serve this common interest in discouraging intrasocietal interests
- the belief in such gods would therefore be socially valued in larger societies
- One of his correlations: cor(sccs$v234,sccs$HiGod4) = 0.034 (comb.Rdata) v234=settlement patterns
- Roes and Raymond
- Embellished by Johnson
Unexecuted remnants of the original plan
The "hidden variable" in contrast to proximal and distant effects with respect to Ethical God beliefs is a "Malthus event" Turchin process/Banfield for nomads
Banfield: Ecological affordances: begins with the ecology 1) Central asian nomads - more empire formation 2) agree with agrarians to keep trade routes open did dummy variables for missing data in FxCmtyWage (little effect) and Fratgrastr (constructed but not yet tested) Run comb/HiGod4 to Fix Yahgan?
Proposal
Notes for Templeton Proposal Keyword: Temple
SVG
http://intersci.ss.uci.edu/wiki/svg/CausalGraphMoralGods.svg
