Commemorating local victims of past atrocities and far-right support over time

In a panel analysis, we find evidence that in the election after a Stolperstein is placed in a polling station (on average 1,312 eligible voters in 0.48 km²) fewer people vote for the AfD in that area. This finding suggests that AfD vote shares decrease in local neighborhoods where countermemorials are placed.

Formally, we run three models that each predict AfD vote shares^** with Stolpersteine as a continuous variable of the number of Stolpersteine (ln), as a dummy variable (at least one Stolperstein vs. no Stolpersteine), and as a continuous variable of the number of distinct locations with Stolpersteine (ln) (please see Materials and Methods and the Fig. 3 A and B). We include polling station and election fixed effects in the models to account for fixed unit characteristics as well as temporal trends. By including polling station fixed effects, we rule out the effect of constant variables (time-invariant confounders) such as how many Jewish residents lived in the polling station before WWII, whether there was a settlement at all during that time and whether the polling station is in a traditionally liberal or conservative neighborhood of Berlin. By including election fixed effects, we rule out temporal trends such as the number of Stolpersteine and AfD vote shares independently increasing over time. Our dataset includes information from seven elections over 9 y.

The results of the panel analysis are reported in Table 2 and indicate that a greater number of Stolpersteine in polling stations before an election is associated with a lower AfD vote share in the following election.^†† The coefficient size is large: polling stations that received one or more Stolpersteine for the first time prior to an election record 0.96% points fewer AfD votes compared to those that received none.^‡‡ Given that the AfD vote share is around 10%, a 0.96% points decrease is substantively significant. The results also suggest that more Stolpersteine (models 1 and 4 in Table 2) and more unique locations with Stolpersteine (models 3 and 6 in Table 2) in a place correspond to further decreases in AfD vote shares.^§§ These results are robust to changes in the cutoff of between-elections overlap (from 75% to 99%; SI Appendix, section F). Moreover, as would be expected because people frequently cross the arbitrarily demarcated polling stations as they walk around their neighborhoods, we observe that AfD vote shares in one unit are also predicted by preelection Stolpersteine in adjacent units (SI Appendix, section G).^¶¶

We still observe a negative relationship between Stolpersteine and AfD vote share when we use the estimator suggested in ref. 22 to correct for problems stemming from negative weights and heterogeneous treatment effects with two-way fixed-effects in panels (SI Appendix, Table S7). For further discussion, please see SI Appendix, section H. Overall, the panel analysis provides strong evidence for a strong negative relationship between Stolpersteine and AfD vote shares.

In a discontinuity analysis, we find that fewer people vote for the AfD in polling stations that receive at least one Stolperstein before the elections (i.e., treatment group) than in polling stations that receive at least one Stolperstein after the elections but not before the election (i.e., control group). This finding suggests that Stolpersteine cause a reduction in AfD votes.

The discontinuity analysis leverages noise around an arbitrary cutoff value to create treatment and control conditions (25). In our case, election times are arbitrary cutoffs that we assume to be orthogonal to the timing of Stolpersteine placement. The placement of new Stolpersteine takes place sporadically when the artist comes to Berlin. New Stolpersteine are not prioritized based on neighborhood characteristics. Instead, priority is given to elderly descendants of victims/survivors who are still alive and able to come to the placement ceremony. The timing is not related to elections—time to elections could not predict Stolpersteine placements in additional analyses we ran. Moreover, we assume that whether a Stolpersteine is placed shortly before or shortly after this arbitrary cutoff date is noisy because there is varying (up to 4 y) lag between commissioning and placing a Stolperstein. Thus, in the time window shortly before and after an election, Stolpersteine in treatment and control condition have likely already been commissioned but—for reasons that have nothing to do with the outcome (i.e., AfD vote shares)—have not been placed yet. Therefore, variables that might motivate people to both commission a Stolperstein and also against vote for the AfD, such as concerns about fascism, exist in both conditions (recall that such concerns are already minimal because most people who commission Stolpersteine live outside of Berlin). In other words, because of the arbitrariness of the time lag between commissioning and placing Stolpersteine, polling stations that receive a Stolperstein soon after elections can serve as counterfactuals to polling stations receiving a Stolperstein just before elections.

The results of the discontinuity analysis are reported in Fig. 4 and show that the effect of Stolpersteine on AfD vote share is significant and negative. Given our context, carrying out a conventional regression discontinuity design analysis with a running variable as suggested in ref. 26 is not possible. Thus, we vary the time window between three months and a year, run the regression analysis with this sample, and present the results (given that the artists places Stolpersteine every three to four months, when we further limit the time window, we do not have a large enough number of observations for a meaningful analysis). Across samples, the effect of Stolpersteine is quite substantive and varies between −1 and −0.5. Overall, the discontinuity analysis provides causal evidence for the impact of Stolpersteine on the AfD vote share.

A decrease in the AfD vote share could either be the result of fewer voters casting their votes for the AfD (i.e., a change in the numerator) or of higher turnout (i.e., a change in the denominator). To rule out that the observed effect is an artifact of higher turnout rates, we repeated the panel analysis with the count of total votes, turnout ratio, and the count of AfD votes as alternative outcomes. We find no change in turnout or total votes cast. We do find a decrease in votes cast for the AfD.

These additional analyses indicate that the decrease in the AfD vote share is indeed driven by a decrease in the number of votes cast for the AfD, thus confirming that what we observe is a change in voting preferences, not turnout. Center-right parties, such as the Christian Democrats (CDU) in Germany, play a crucial role in shaping the far-right vote share (27). As we observe a change in voting preference away from the AfD, we might therefore see an increase in the vote share of the center right. Indeed, when we repeat the panel analysis with the vote share of CDU, we find a positive association between Stolpersteine and CDU vote share. For more details on these additional analyses and further discussion, please see SI Appendix, section K.

Stolpersteine may influence residents in two ways: as material objects and as a social practice. As material objects, the brass plaques with information about victims may draw the attention of passersby to the history of the Holocaust and highlight how widespread and omnipresent the Nazi crimes were. For potential AfD voters, this reminder is in stark contrast to the AfD’s downplaying of and moral disregard for the victims and crimes of the NS past. Such incongruence in moral values is an important predictor of people’s vote choice (28) and could possibly make some people reconsider their potential vote for the AfD.

Results from two surveys about Stolpersteine are consistent with this argument. First, respondents agreed highly that Stolpersteine are reminders of how widespread NS crimes were in an online survey we conducted with a convenience sample (N = 210). This was followed by statements about feeling empathy for the victims. To illustrate, when asked to describe what they associate with Stolpersteine, one respondent wrote: “You realize immediately that the racial extermination program by the Nazis was executed in every part of Germany. The idea ‘but where we lived it wasn’t so bad’ clearly becomes absurd [when seeing the stones].” Another wrote: “I first always think about how these people must have felt back then when they were abducted, how afraid of death they may have been [...]”

Second, in a different survey by other authors (N = 739), people report strong emotional, negative reactions to Stolpersteine (29). Especially their high levels of anger and disgust suggest that Stolpersteine may indeed remind passersby of moral violations against the victims (30).

Instead of, or in addition to the material presence of the memorials, Berlin residents may respond to the social practices that accompany them. Social practices give memorials life and render them meaningful as people actively engage with them (31). As described in the introduction, the Stolpersteine memorial invites everyone in a neighborhood to participate and many do. For example, residents attend the placement ceremonies, receive or hand out printed biographies of the victims, and attend regular initiatives to clean the stones or place candles and flowers next to them, especially on the anniversaries of the pogroms (November 9th) and of liberation (May 8th).

These social practices increase the visibility of the stones and the awareness of important historical dates, such as November 9th. This heightened awareness in a neighborhood may decrease voting for the AfD, the party that downplays the significance of the Holocaust, through moral value incongruence as explained above. In addition, Stolpersteine might especially via these social practices become a symbol or “cue” to others that this is a place where victims of violence and oppression are valued. Past research shows that such cues can be powerful levers of behavior change (32). Stolpersteine might signal a local social norm of commemorating and honoring the victims of the Holocaust to residents, which is incompatible with voting for the revisionist AfD. Of course, to prove the mechanisms we suggest here, more systematic experimental and survey research is needed.

If social norms of commemoration are the mechanism, Stolpersteine likely contribute to a local climate of commemorating victims of past atrocities and condemning revisionist actors in a dynamic process. With each additional Stolperstein in a neighborhood, they promote social norms of an active culture of remembrance, which may spawn additional Stolpersteine placements and more social rituals. This assumption is consistent with our results showing a) no evidence for a backlash–suggesting that a receptive local climate evolves together with Stolpersteine placement,^## and b) a robust trend where an increasing number of Stolpersteine further decreases AfD vote shares over time, although this effect saturates especially if more Stolpersteine are placed in the same location. Such a dynamic model is also consistent with theories on wise interventions that changes occur in an ongoing, mutually reinforcing transaction between people and their social environments (33).

First, without a randomized experiment, we can never fully rule out omitted variable bias. Randomizing Stolpersteine or other memorials to neighborhoods is impossible for practical and ethical reasons. Practically, commissioning Stolpersteine takes a long time and is expensive. Ethically, we believe that it would be wrong to commission or otherwise modify (e.g., clean) Stolpersteine that commemorate victims of the Nazi regime for experimental purposes as we would essentially instrumentalize the victims for our experiment.

In the absence of a randomized experiment, both Stolpersteine and AfD votes could be driven by an unobserved variable such as a positive local diversity climate. This is unlikely because our data are very granular (between 1,709 and 2,257 polling stations over 11 y). Moreover, 80% of people who commission Stolpersteine live outside of Berlin. While not likely, external commissioners might have social ties to residents in specific areas, and approximately 20% of Stolpersteine are placed by local organizations and residents. But even in that case, the people who might influence external commissioners or are part of the 20% that commission stones themselves, are “earlier adopters,” and most likely not the same people who switch their vote from the AfD to a different party.

Second, our data do not allow us to distinguish between the impact Stolpersteine may have as material objects versus via the social practices accompanying them, such as the installment ceremonies. While the former would primarily encourage the placement of additional memorials, the latter would suggest that the focus should lie on more active engagement with memorials, whether old or new. Future research should investigate the impact of memorials as material objects or as social practices.

Third, while Stolpersteine are an interesting and widespread case, they are just one case, and the data only cover Berlin, Germany. Countries that committed past atrocities fall on a spectrum from acceptance to denial (3), and Germany is far on the acceptance dimension. While we expect our results to generalize to other places that largely accept past atrocities, we do not expect them to generalize to places of denial, such as Turkey. Because Stolpersteine or a similar memorial would be in strong opposition to the platform of virtually all governmental actors there, not just the far right as in Germany, we might instead see backlash. In places marked by “mnemonic warfare” (3), meanwhile, such as the United States, polarization might moderate the effects of local memorials. Examining the interplay between the level of public acceptance and local reactions to public memorials would be an interesting question for future research.

As an increasing number of initiatives seeks to increase the public visibility of past atrocities and the fates of victims, we investigated how one such initiative affects support for a revisionist party. The robust negative relationship we find between the installment of memorials and the vote share of the far right within the same district, speaks to the power of making past injustice visible. While further research is needed on the causal direction of the effect and the underlying mechanisms, the results we present here suggest that public memorials have implications for contemporary political behavior.

O.T., R.D., and B.F. designed research; O.T., R.D., and B.F. performed research; O.T. and R.D. contributed new reagents/analytic tools; O.T. analyzed data; and O.T., R.D., and B.F. wrote the paper.

The authors declare no competing interest.