14 September 2005, 01:33
Social Marketing in Ryazan: Costs of the War in Chechnya
Theodore P. Gerber and Sarah E. Mendelson
This paper addresses the results of the June 2004 regional survey assessing the impact of the "skol'ko" campaign on residents of Ryazan. While there are other measures of the impact, we address here the quantitative results from the resurvey in Ryazan and the control group Kaluga. The results from this survey suggest that social marketing in Russia can have an impact even in a highly restricted media market and even on an issue that is as politically sensitive as the war in Chechnya when activists rely on public opinion to shape messages. Even with highly resonant messages however, we do not find indirect effects. In other words, these findings suggest that social marketing is effective at shaping specific attitudes, but activists should not assume or expect the public to link in their minds issues that a campaign does not explicitly address.
Exposure and Reaction to the Campaign in Ryazan (Table 1)
Finding: Our survey results suggest the Ryazan campaign reached a considerable part of the population and communicated the message clearly and compellingly.
Specifically, the "skol'ko" campaign appears to have reached a considerable proportion of the population in the city of Ryazan. Of the 220 respondents in Ryazan city, 31% said they had seen the materials at least once (Panel A). Thus, by our best estimate, about 3 out of 10 residents had some exposure to the campaign. Outside of Ryazan city only a handful of Ryazan oblast residents said they had seen the campaign materials. This enhances our confidence in the data: we would not expect many people outside of Ryazan city to have seen the campaign, yet obviously a few might have if they visited the capital while the campaign was underway.
In Kaluga oblast, we were initially surprised to see that 17% reported seeing the "skol'ko" posters or newspapers. This could raise questions about the validity of the survey, since it is unlikely that anything approaching that large a percentage of the Kaluga population would have traveled to Ryazan city. However, when we look further we see that the majority (60%) of Kaluga respondents who say the saw such materials also associate them with an election campaign, and most of the rest (22%) cannot say what they referred to. Only 3% associate them with Chechnya. (In contrast, 73% of Ryazan oblast residents outside the capital who say they saw the posters or newspapers correctly associate them with Chechnya.) It is entirely possible that campaign posters that somehow used the word "skol'ko" appeared in Kaluga. We would be much more concerned about the large number of Kalugans who report seeing the materials if most of them associate the slogan with Chechnya. Since the Kaluga "skol'ko" materials obviously are not part of the Chechnya campaign organized in Ryazan, there is no point in analyzing how Kaluga respondents reacted to the materials.
If a substantial number of Ryazan residents saw the materials, few actually discussed them: only 6% report discussing them on one or more occasions (Panel B).
As for the Ryazan respondents who saw the materials, the majority (57%) indeed associate the "skol'ko" campaign with Chechnya (Panel C). Clearly, the message got through to most who saw the posters, newspaper, and exhibit. Still, 29% associate "skol'ko" with elections and 10% cannot say. (There are too few who saw the materials in Ryazan outside of the capital to reach any definitive conclusions, though the 73% lends credence to our interpretation that these respondents did travel to Ryazan city and did see the materials.)
How did Ryazan city residents react to the "skol'ko" materials? About two thirds (67%) say that when they saw the materials they wanted the government to tell the truth about the number of casualties (Panel D). Three-quarters (75%) say they react by wishing the government would tell the truth about the financial costs of the war (Panel E). Three-fifths (60%) say their response is to wish the government would end the war as soon as possible. These results all suggest that the campaign materials generally elicited the desired response on the part of those who were exposed to them. Also, very few respondents had the opposite reaction to that intended.
Who saw the materials and reacted as expected? (Table 2)
Finding: Women were more likely to respond to this campaign than men, but they were less likely to have seen the campaign.
To see if exposure and reaction to the "skol'ko" materials varied by demographic characteristics, we ran logistic regressions to test for differences by cohort, gender, and education. We first examined whether these factors influence the probability of being exposed to the materials among Ryazan city residents (Panel A). We found all three factors have statistically significant effects: younger residents, males, and better-educated residents were all more likely to report seeing the "skol'ko" campaign collateral materials. We performed a similar analysis to test for variation in the probability of discussing the "skol'ko" campaign, and we found that only college education has any significant effect: it increases the probability (Panel B). Thus, the college educated in the city were both more likely than those without college education to see (and register cognitively) the "skol'ko" materials and to discuss them.
In order to test for demographic variations in reactions to the campaign, we analyzed the sample of respondents from Ryazan city and the rest of Ryazan oblast who reported seeing the "skol'ko" materials. (We added the non-capital residents from Ryazan in order to increase statistical power; we are justified in doing so based on our conclusion, discussed above, that non-capital residents who report seeing the materials did actually see them.) We found no significant demographic variations in the probability of correctly identifying the theme of the campaign as Chechnya (Panel C). Of course, the sample size is small, so all null findings must be considered tentative. However, we don't have strong theoretical reasons to expect demographic variation in the interpretation of the campaign, so this null finding does not surprise us.
We do find significant effects of gender on reaction to the campaign: women who saw the materials are significantly more likely than men to react by wanting to know the truth about Russian casualties and about costs (Panels D and E.) In fact, these effects are very strong: the odds of women having the desired reactions are, respectively, 5.3 and 2.8 times higher than the odds for men (to determine the odds ratio, take the exponential of the logistic regression coefficient). This is important when we consider that women are less likely to have seen the campaign materials.
Taken together, these findings imply that the impact of the campaign could be considerably enhanced if it were to reach more women: they are more likely to have the desired reaction, but less likely to be exposed to the campaign. We also find evidence that the college-educated are more likely to respond as hoped regarding the economic costs of the war (Panel E). There are no demographic variations in the probability of reacting by wanting the war to end immediately, and no effects of cohort on any of the three reactions.
Chechnya Dynamics (Table 3)
Finding: Residents of Ryazan city expressed a dramatically increased anxiety about the casualties associated with the war which we interpret as a result of the campaign.
The campaign focused on the casualties and economic cost of the war, so we first examine our most direct measures of perceptions of these aspects. We have data from the 2002, 2003, and 2004 regional surveys on the main feelings respondents experience when they hear reports about the war (Panel A). In all years, the most common feeling in all regions was "alarm over losses of Russian troops." However, in Ryazan city the percentage of respondents who identify this feeling jumped from 53% to 81% between February 2003 and July 2004. It jumped from 53% to 73% in non-capital Ryazan. In Kaluga, it remained relatively stable during this period, inching from 60% to 64%. The sharp jump in Ryazan city (where the campaign took place) compared to Kaluga (where there was not campaign) is consistent with one objective of the campaign: to increase public concerns about Russian casualties. The findings are as good as might be hoped for, in terms of demonstrating a larger impact of the campaign in Ryazan on public opinion.
Similarly, we find evidence consistent with the campaign achieving its second goal: raising public concerns about the economic costs of the war. In Ryazan city, the relevant percentage increased from 19% to 32% following the campaign; in non-capital Ryazan, from 17% to 30%. In Kaluga oblast, the figure actually fell from 23% to 18%. Again, we can never rule out alternative explanations - though we are hard pressed to come up with any. The most likely explanation for this pattern is that the campaign had the intended impact.
We note that anger at the Russian government declined in Ryazan city (from 17% to 7%), which may seem counter to the goals of the campaign. However, since respondents were limited to choosing two feelings, this is probably an artifact of the increasing salience of casualties and economic costs in Ryazan city. Finally, the percentage reporting that they never hear reports about Chechnya fell dramatically in Ryazan, which may also have something to do with the campaign.
We find further evidence that the campaign had the desired impact in responses to the question about how important concerns about costs of the war are in shaping how respondents think about the war (Panel B). Here we see a sharp jump in the percentage of Ryazan city residents - from 77% to 92% -- following the campaign. In Kaluga the percentage remains essentially stable and in non-capital Ryazan it actually declines. Here too we believe the burden is on skeptics to come up with an alternative explanation that is more plausible than the argument that the campaign had an impact.
Finding: The campaign did not influence Russians' views on the preferred policy course on the war in Chechnya.
Although the most direct goals of the campaign were to increase concerns about casualties and costs, we hypothesized that these concerns might translate into evidence of a broad desire to end the war as soon as possible. We do not find such results. (Panel C).
While the percentages favoring intensification of military action drop, they actually drop somewhat less sharply in Ryazan city than in Kaluga or non-capital Ryazan. We cannot, therefore, attribute the drop in support for escalation to the campaign. It is interesting to note, though, that support for starting negotiations (without declaring a ceasefire) increased substantially in Ryazan city, but not elsewhere. Support for a withdrawal of troops followed by negotiations decreased correspondingly in Ryazan city. Finally, the greatest surge of all is in the percentage of those with no opinion in Ryazan city.
On the one hand, the answer of what should be done in Chechnya stumps most policy experts the world over, and the campaign's goal never included a specific policy option such as negotiation rather than intensification. On the other hand, the campaign message did not provoke a coherent opposition to the war.
Our interpretation of such findings is that while the public may correctly understand the message and react to it in the intended fashion, it is hard to predict and shape the second-order (broader) reactions. For example, we could not exclude the possibility - one explicitly raised by some of the Ryazan activists at our first meeting - that a campaign around Russian casualties might have actually increased nationalism against Chechens. We find no such results. Thus, we are reassured that there was no increase in the percentage of respondents who feel "anger at Chechens" as a primary response to reports about the war. (Panel B).
Campaign Impact at the Individual Level (Table 5)
Finding: We have empirical results that suggest an individual who was both exposed to the campaign and correctly understood the campaign was affected by the message. The results are dramatic among older women and the young.
For those observers interested in quantitative representation of the campaign's impact at the individual level, we provide below a detailed discussion of a logistic regression model.
There are numerous ways a campaign can have an impact, but the most straightforward is that it can directly shape the views of those who are exposed to it. It is difficult to test for such an impact because exposure to the campaign may itself be remembered differently by those who are affected differently, impact and exposure may be jointly shaped by unobserved variables (producing a "selection" effect rather than a "treatment" effect), and the number of exposed may be too small to yield adequate statistical power for the appropriate statistical test. Nonetheless, we analyzed whether Ryazan city respondents who recalled seeing or discussing the campaign differ from those who did not with regard to the two key measures of the intended result: identifying alarm at Russian casualties and alarm at the war's costs as among the strongest feelings experienced in response to reports about the war.
Our models for the probability of feeling alarm at Russian casualties (Panel A) indicate that indeed the level of concern increased significantly in Ryazan city in 2004. This conclusion is based on the strong, significant positive effect of the Ryazan city*2004 interaction in all models, which more than offsets the negative (and non-significant) main effect of Ryazan city and adds to the significant, positive effect of the 2004 dummy variable. Based on these models, alarm over losses increased in all three cities in 2004 relative to 2003, but much more sharply in Ryazan city than elsewhere. We also note that our results imply younger respondents are generally less likely than older respondents to experience alarm at casualties, while women are more likely to do so than men.
When we enter the measures of direct exposure to the campaign, we find that mere exposure actually reduces alarm (the coefficient is -.527), but exposure and correct interpretation of the campaign as relating to Chechnya has a significant, positive effect of .736. Consider in tandem (as they must be), the two coefficients imply that those who saw the campaign but did not associate it with Chechnya are less likely to experience alarm at losses than those who never saw the campaign at all; however, those who saw the campaign and correctly identified it with Chechnya (as did the majority of those who saw the campaign in Ryazan, as we saw above) are more likely to experience alarm at Russian losses than those who never saw the campaign at all. In fact, their odds of experiencing alarm are roughly 23% (e.736-.527 = 1.23) higher. Discussing the campaign appears to have no impact. Including the three measures of campaign exposure significantly improves the overall fit of the model based on the likelihood ratio chi-square test (chi-square = 15.2, df = 3, p = .002). These results are clearly consistent with the hypothesis that the campaign had the intended effect on the views of those who saw it and correctly understood its theme. (To repeat: the results do not "prove" such an impact, for the reasons described above. They do, however, fail to disprove it.)
Moreover, when we add the interaction involving under-30 and correct interpretation of the campaign, the results are dramatic. This single variable produces a highly significant improvement in model fit based on the likelihood ration chi-square test. The strong, positive, and statistically significant interaction effect implies that the impact of the campaign is quite dramatic among younger Russians who saw the materials and correctly interpreted them. The effect is large enough in magnitude to more than offset the negative main effect of the under 30 dummy variable: while Russians under 30 who did not see (or did not correctly interpret) the campaign are substantially less likely to experience alarm over troop losses compared to those over 30 who did not see the materials, those under 30 who saw and understood the materials are much, much more likely than those over 30 (whether or not they saw the materials) to experience alarm over losses. In fact, once we enter the interaction effect, the main effect of correctly interpreting the "skol'ko" campaign becomes negative (but not significant). This implies that the entire positive effect of the campaign evident in Model 2 stems from its impact on Russians under 30. Given this finding, the earlier result that those under 30 were more likely to encounter the materials is encouraging.
A similar, though not identical, pattern emerges with respect to alarm at the money spent in Chechnya (Panel B). Most importantly, we again see a larger spike in Ryazan city in 2004 than elsewhere and also a strong positive interaction effect involving under 30 and correct interpretation of the campaign materials. Out multivariate results are in both cases consistent with the hypothesis that the campaign had the desired effect on those under 30. We can put it even more strongly: the results suggest that to the extent the campaign had any direct effect at all (and they suggest that it did), the effect was concentrated almost entirely among the youth.
One other finding from these analyses is worth mentioning: net of the controls for direct exposure, the magnitude of the Ryazan city*2004 interaction effect declines, but only partly. Thus, the aggregate increase in the alarm about military casualties and economic costs is partly, but only partly, attributable to the direct impact of the campaign. (If it stemmed entirely from the direct impact of the campaign, the interaction term would become statistically indistinguishable from zero net of the variables entered in models 2 and 3. If it did not stem at all from the direct impact of the campaign, it would not change at all going from model 1 to models 2 and 3.) Of course, the campaign could also have fostered increasing alarm over casualties and losses in indirect ways - e.g., by provoking discussions of the campaign's topic that did not specifically refer to the campaign itself, by raising the issues of casualties and costs in a subconscious, subliminal manner, or by provoking various other reactions.
Views on NGOs (Table 6)
Finding: We find no impact from the campaign on the profile of NGOs. We conclude that if NGOs wish to raise their profile, campaigns must be designed specifically to "brand" organizations.
We posited that a possible collateral benefit of the campaign would be to increase the profile and public support of human rights and anti-war NGOs in the region. How have awareness of and views toward NGOs that oppose the war in Chechnya changed in these regions between February 2003 and July 2004? We found there was in fact no direct impact.
The campaign does not appear to have affected the profile of such groups either positively or negatively. Asked about non-specific NGOs that oppose the war in Chechnya (Panel 6), majorities expressed positive views in all three regions before and after the campaign; although the level of support increased in Ryazan city, the increase is moderate and comparable in magnitude to the increases in the other two regions. The percentage who never heard of any such groups did not decline appreciably in Ryazan city.
The campaigns were mounted by a coalition of organizations and did not advance the name of any one specific NGO. The familiarity with Memorial actually decreased somewhat in Ryazan city, while it increased in non-capital Ryazan and remained stable in Kaluga (Panel B). Moreover, among those who have heard of Memorial in Ryazan city, approval of its activities actually decreased following the campaign, while it increased in non-capital Ryazan and remained stable in Kaluga.
The familiarity with the Committees of Soldiers' Mothers improved in Ryazan city following the campaign, but it also did so in Kaluga (Panel C) so we think this was independent of the campaign. Among those familiar with the organization, approval of its activities remained stable in Kaluga and non-capital Ryazan, but appears to have declined somewhat in Ryazan city. There was not much change in any of the regions with respect to familiarity with Moscow-Helsinki Group (Panel D), and a rather confusing pattern of shifts in approval among the few who have heard of the organization (increasing approach in non-capital Ryazan, decreasing in Kaluga and Ryazan city.)
All told, the data suggest that the campaign did not raise the profile of NGOs that oppose the war in Chechnya and defend human rights. Future campaigns need to devote more explicit attention to the task of raising the profile of HR NGOs if that is a goal. Moreover, we have no evidence of any effects on other general attitudes not addressed by the campaign, such as abstract attitudes toward rights.
 To do this, we first estimated a logistic regression model on the combined samples from all three regional surveys (April 2002, February 2003, and July 2004) where we controlled for demographic variables, a rural/urban difference, an additive regional effect (contrasting Ryazan city and non-capital Ryazan to Kaluga), year effects (contrasting 2002 and 2004 to 2003), and interaction terms to test whether the measures did indeed increase in Ryazan city in 2004, net of the control variables. This specification comprises our first model for both outcomes. Next, we entered three measures of campaign exposure: a dummy variable indicating that the respondent reported saw the "skol'ko" campaign materials in 2004, a dummy variable indicating that the respondent both saw the materials and correctly identified them as dealing with Chechnya, and a dummy variable indicating that the respondent discussed the "skol'ko" materials on at least one occasion. If the campaign had a direct impact on people's level of alarm about casualties or costs, it should show up as the positive, statistically significant effects of these dummy variables, net of the controls. This model constitutes our second model for both outcomes. Finally, we incorporate an additional interaction term - the product of the dummy for correct interpretation of the campaign and the dummy variable denoting those under 30 - in order to test the possibility that the campaign had a greater impact on younger citizens.
 We do have on anomalous finding to report although we have no evidence that it is related to the campaign. We record significant and substantial increases in support for all three civil liberties in Ryazan city: the percentage saying that freedom of religion should be top priority grew from 40% to 58%, the percentage saying the same about freedom of expression from 44% to 73%, and, for freedom of association, from 24% to 47%. Support for freedom from torture and the property rights also increased. While it is tempting to attribute the increased demand for these rights to this campaign - and, indeed, we are certainly open to such an interpretation - the fact that we find similar increases in non-capital Ryazan (where very few saw the campaign) suggests that something else is going on. We do note that the increases in support for these rights are smaller or non-significant in Kaluga.