This measure has respondents place themselves on a tenstep ladder with the top rung labeled as the best possible life and the bottom rung as the worst possible life. When defined in this way, this scale measures discrepancies between people’s attainments and aspirations. But it does not measure discrepancies between their current status and expectations as to what they want and deserve and how they feel about it . Thus, this measure emphasizes RD’s cognitive component at the expense of its crucially important affective component. Second, Stouffer offered a concept, not a testable theory. This may have enhanced its popularity, but it, too, has restrained development of the idea. Only recently have full-fledged theories emerged that allow direct testing and falsification. In the 1980s, Heather Smith and I decided that what was needed was a meta-analysis of the far-flung research literature that employed the concept . As with most extensive meta-analyses, it took a while to complete the Herculean task—twenty five years of an off-and-on effort to be exact. Our first task was to clear the underbrush that had sprung up due to the lack of a precise theoretical and measurement model. Using inclusion criteria that ensured that RD as advanced by Stouffer was indeed being tested, a huge 76 percent drop-off occurred. While we initially secured 860 studies that purported to study some aspect of RD,blueberry pot size only 210 met our modest criteria and entered the meta-analysis.
Failing to exclude these marginal studies has been a major problem for previous qualitative reviews that did not employ strict inclusion rules. Consequently, to a great extent the criticism of RD involves this enormous research literature that does not actually involve RD. Like the Gurr study, much of this work is interesting and uncovers important findings; but these studies are testing other phenomena—not RD. Typically they ask if macro-indices of societal deprivation and frustration lead to collective protest. Such a hypothesis contrasts with that of RD; indeed, it is largely a rival hypothesis. It is important to note that the average effect sizes from this macro-level work are in general as large or larger than those found for RD studies; thus, their exclusion does not artificially inflate our meta-analytic RD effects.2 Our second task was to ascertain the mean effect sizes for the entire RD literature as of January 2010. This long-term effort amassed 210 separate studies, composed of 293 independent samples, 421 non-independent tests and 186,073 respondents. Three different tests provided evidence that our meta-analytic data were not altered by a publication bias that favored positive results. The mean effect sizes that emerged were highly statistically significant but small—.106 for studies, .144 for samples, and .134 for tests Why are the RD effects so small? We next tested three hypotheses for an explanation. First, our affect hypothesis predicts that those RD measures that explicitly included affect would yield significantly larger effects. That is, when people are clearly angry and resentful over their perceived disadvantage, the full RD effects would emerge.
As Martin and Murray insisted years ago, the affect dimension is crucial. One can detect a personal or group disadvantage but believe that it is fully justified—as system justification research has repeatedly shown . Indeed, experiments in New Zealand show that system-justifying beliefs act as a moderator for both IRD with well-being and GRD with political mobilization. <Query, Author: Remind audience of what these acronyms mean.> As expected, subjects high in these beliefs show smaller RD effects . Hence, we view feelings of anger, resentment, entitlement and deservingness as basic to the RD formulation. Our second proposition involves the fit hypothesis. This idea entails both conceptual and methodological considerations. We predicted that RD effects would be larger when the levels of analysis between RD and the outcome variable are the same . Put differently, we contend that RD effects are reduced when IRD is used to predict group-level phenomena and when GRD is used to predict individual-level phenomena. Surprisingly many studies mix the RD and dependent variable levels of analysis.Our third test is purely methodological. The research quality hypothesis holds that the more rigorous studies will yield larger effects. The inventor of meta-analysis, Gene Glass , broke from traditional reviewing practice by including poorly conducted studies as well as well-conducted ones in order to detect the effects of research quality. If, for example, the major effects of RD are found among the poorest conducted studies—as with the effects of psychotherapy for adult depression , we would strongly question RD’s predictive power. But we predicted the opposite—that the most rigorous RD studies would reveal the largest effects. We defined quality in terms of the reliability of both the RD and dependent variables. The meta-analytic results. Figure 1 provides the overall results by showing the percentages of variance accounted for by various subsets of the tests. For bar A, the worst conducted RD tests have none of our three desirable characteristics and yield only an r of .079. The next bar B in the histogram shows the mean r of .134 for all 421 tests. Bar C shows a mean r of .165 for those tests that did tap affect but had neither reliable measures nor a fit between the levels of analysis of RD and the outcome variable.
Bar D shows a mean r of .203 when the tests boast both fit and an affect measure but unreliable measures. Finally, bar E records the results of the optimal tests. It reveals a mean r of . 230 when all three of our concerns are met—reliable measures that tap affect and have the same level of analysis between RD and the dependent variable. In addition, direct statistical tests of our three hypotheses all provide significance levels at the .05 level. But how are we to interpret this final mean r of .23 across many types of dependent variables? First, this effect size represents a Cohen’s d of .47, which is recognized as a solid medium-sized effect in social science research . Second, r = .23 is consistent with mean effect sizes of published meta-analyses generally. Thus, an extensive meta-analysis of inter group contact effects on prejudice yielded a mean r of -.21 . And a synthesis of more than 120 meta-analyses dealing with psychological assessment produced a mean r of .27 . Similarly,plant raspberry in container a synthesis of 22 meta-analyses concerning ability self evaluations and actual performance obtained a mean r of .29 . Larger effect sizes are rare in meta-analyses because they typically include a heterogeneous variety of research formats, contexts, and subjects. Accordingly, these results solidly support the importance of RD when it is tested appropriately. Universality of the RD phenomena. Our meta-analysis also addresses a question too seldom raised by social psychologists—the universality of the phenomenon. Positive results were recorded from 30 different nations across the globe with widely contrasting societies and cultures . Relative versus objective deprivation. One limiting possibility of RD effects is that they may simply reflect absolute deprivation. Relevant research rejects this possibility. Adler and her colleagues found that the subjective sense of social status was more strongly and consistently related to six health factors than objective social status. Smith and her colleagues uncovered a similar result. They located 26 studies that allow a direct comparison of the relative and absolute deprivation. All 26 studies used income as the objective measure of deprivation. The mean r for the RD measures proved to be .18 as compared with an r = .12 for the objective deprivation measures In terms of the percentage of variance explained, RD’s mean effects are more than twice that of absolute deprivation. These data supply yet another reason why macro-level measures of collective deprivation cannot be used to gauge the perceived RD of individuals. The range of dependent variables. The wide-ranging RD research literature includes four broad types of dependent variables . Two explore the links between IRD and internal states and individual behavior; two others explore the links between GRD and intergroup attitudes and collective behavior. The results from each of these four domains resemble the results for the total sample.3 1) The internal states dependent variables include such outcomes as psychological stress, depression, physical health and altered self-evaluations. Without our three moderators, the 188 tests of this domain yields a mean r of .173 . For the subset with two of our critical moderators , the mean effect size rises significantly to r = .271. 2) The individual behavior dependent variables encompass both normative and non-normative actions. Escape behaviors are also included. For this category, the basic mean r for 126 tests without any of our three moderators is . 118.
When affect and fit are also considered, the mean r becomes . 142. Interestingly, we found that when affect is measured and the respondents compare prejudice measures but variables tapping stereotypes, in group identification, and nationalism as well as in group bias and identification. For the 299 tests in this domain without any of our three moderators the mean effect is r = .115 For the subset boasting two of our critical moderators , the effect becomes strikingly larger—a mean r = .320. Using straightforward measures of GRD, I have had repeated success over the years predicting prejudice and related indicators. In the 1960s, Reeve Vanneman and I showed that GRD correlated positively with white voting for the racist presidential candidate, George Wallace, and against black mayoral candidates in Gary, Cleveland, Los Angeles, and Newark . Years later, with Dutch and German colleagues, we employed three different European surveys to show that GRD significantly enhanced both blatant and subtle forms of prejudice against immigrants in each of four nations—France, Germany, the Netherlands, and the United Kingdom . This study also found that IRD related positively to prejudice with its effect fully mediated by GRD. In other words, IRD plays its role in the prejudice process by increasing GRD which in turn increases prejudice against immigrants. In addition, these surveys socially located those with high levels of relative deprivation. For both IRD and GRD, the relatively deprived in Western Europe are poorer, pay less attention to politics,and are more politically alienated and feel less politically efficacious . But does GRD contribute to our understanding and prediction of inter group prejudice beyond that already provided by other, better-known predictors? A 2002 national probability phone survey of German citizens offers an answer by including sixteen major predictors of prejudice against resident foreigners . The largest predictors, as expected from the vast prejudice research literature, are social dominance orientation, positive contact , and authoritarianism. But following these “big three,” GRD ranks together with political inefficacy as the next most important and highly significant predictors of anti-immigrant prejudice in Germany. Thus, GRD adds significantly to the prediction of prejudice even when fifteen other predictors are included in the regression. 4) The collective behavior dependent variables range from self-reported rioting and intentionally sabotaging job performance to a readiness to approve of violent politics, sign petitions, and join strikes. Note that our range of collective behavior dependent variables is somewhat broader than that employed by some other investigators. Of greatest interest to sociologists and political scientists, this category of RD research has been the most criticized. The meta-analytic results for collective behavior are similar to those for the entire RD data set as shown in Figure 1. Before applying our three moderators , the mean r for nine tests is only .065. For the total sample, the mean r is .153. For the 23 tests that involve both affect and fit but less reliable measures, the mean r rises to .211. For the remaining 18 tests with all three of our critical moderators, the mean effect reaches r = .234.Criticism of RD work has centered on its use as a predictor of collective protest and violence. The 1960s witnessed many studies that claimed RD was a central component of such mass activities. This trend culminated in the 1970 publication of Gurr’s Why Men Rebel. Then the 1970s and 1980s witnessed a torrent of publications attacking this use of RD. Indeed, such papers became almost a cottage industry in political science as well as sociology. They attracted considerable attention in part because distinguished scholars—such as Edward Muller in political science and Charles Tilly , John McCarthy and Mayer Zald in sociology—led the way.