Preexisting social ties among Auschwitz prisoners support Holocaust survival

Edited by Dora L. Costa, University of California, Los Angeles, CA; received December 21, 2022; accepted June 12, 2023 by Editorial Board Member Mark Granovetter
July 11, 2023
120 (29) e2221654120

Significance

Do humans cooperate when facing extremely low survival chances? Are preexisting social ties increasing one’s chances of survival in life-and-death situations? Testimonies of surviving prisoners of deadly internment camps imply a critical role of the ability to form small mutual-support groups, but survivor testimonies are fundamentally selective. We provide evidence on the importance of preexisting social linkages for victims of the Holocaust, relying on a unique setting in which we observe individual histories and several types of social linkages of a large group of Jewish prisoners entering the Auschwitz-Birkenau concentration camp. We find that the availability of potential friends among fellow prisoners improved a prisoner’s chance of surviving the Holocaust.

Abstract

Survivor testimonies link survival in deadly POW camps, Gulags, and Nazi concentration camps to the formation of close friendships with other prisoners. To provide evidence free of survival bias on the importance of social ties for surviving the Holocaust, we study individual histories of 30 thousand Jewish prisoners who entered the Auschwitz-Birkenau concentration camp on transports from the Theresienstadt ghetto. We ask whether the availability of potential friends among fellow prisoners on a transport influenced the chances of surviving the Holocaust. Relying on multiple proxies of preexisting social networks and varying social-linkage composition of transports, we uncover a significant survival advantage to entering Auschwitz with a larger group of potential friends.
Experiencing violent conflict has been shown to support within-group cooperation (1, 2), but it is not clear whether humans cooperate when survival chances are extremely low. Social networks have been shown to mediate the health effects of stress (3) and to matter in high-stakes contexts (e.g., for soldiers in a war as in refs. 4 and 5), but it is not clear whether preexisting social ties in broad populations are transferable to life-and-death situations.
Survival in deadly internment camps, including POW camps, Soviet Gulags, and Nazi concentration camps has been linked to the ability of prisoners to form small mutual-support groups (69), which points to the importance of social networks in extreme circumstances. However, much of the existing literature (for Holocaust research, see refs. 1014) is based on survivor testimonies, which are fundamentally selective, particularly given low survival rates. It is plausible that those who did not survive also formed mutual-support groups; therefore, statistical analysis based on all prisoners is needed to provide a new type of evidence and to complement qualitative research. Such evidence is also needed to assess whether preexisting social ties are valuable in extremity within demographically diverse populations (i.e., among victims of the Holocaust) or whether only soldier camaraderie established on battlefield is valuable in deadly internment (as in ref. 5).
In this paper, we examine the importance of social networks (linkages, potential friends) for the Holocaust survival of Theresienstadt ghetto prisoners entering Auschwitz-Birkenau. Theresienstadt was an in-transit ghetto, while Auschwitz was a complex of labor and extermination camps. Our analysis is based on the near-complete database of a well-defined group of prisoners and thus avoids survival biases by incorporating information on those who did not survive the Holocaust. Preexisting friendships and social and family ties may be particularly valuable in the extreme environment of a Nazi concentration camp, where there are few market substitutes for social resources. For each Theresienstadt prisoner on a transport to Auschwitz, we construct a variety of proxies for the availability of potential friends on their transport. We then ask how preexisting social linkages affect Holocaust survival. We study several types of social ties that did not require strong social skills to be formed and exploit variation in the availability of potential friends that was outside of a prisoner’s control.
We find a survival advantage conferred by entering Auschwitz with several socially linked fellow prisoners based on measures of family links, camaraderie among prisoners, as well as based on social linkage proxies corresponding to predeportation administrative and residential ties.
We build on the historical research devoted to Theresienstadt (1518) and on the few statistical analyses of deadly internment camps and ghettos (5, 14, 19). Our results confirm the findings of qualitative work based on selective survival testimonies that being socially isolated was particularly costly during the Holocaust. Our findings from a situation of extremity also fit well into the literature highlighting the importance of social links in high-stakes (but not deadly) contexts (2023).

Theresienstadt and Its Records

Theresienstadt (Terezín) was a ghetto established by the SS in 1941 in the garrison city of the same name in German-occupied Czechoslovakia. The ghetto held mainly Czech, German, and Austrian Jews, and most of the ghetto’s population was eventually deported to extermination camps in occupied Poland. The ultimate destination of transports to the East was not disclosed to the ghetto self-administration. Auschwitz, the focus of our study, was the most frequent destination. We do not study prisoners deported to Treblinka and Maly Trostinec, the other two chief destinations of transports from Theresienstadt, because virtually none of these survived the Holocaust. Unlike Auschwitz, Treblinka and Maly Trostinec had no labor camp, they were solely extermination camps.
Our analysis is based on the near-complete database of individual histories of Theresienstadt prisoners ((24)) compiled by the Terezín Initiative Institute (TII), a nonprofit organization founded by an international association of surviving prisoners of the ghetto. The database covers information on 139,769 incoming prisoners, for 99.8% of whom we have information available on their Holocaust survival. We also have data on all transports out of Theresienstadt covering 88,059 of prisoners. The TII data cover prisoners’ names, gender, age, and academic titles including medical doctor titles. We amended the database by measures of social linkages among prisoners collected in various archives.

Social Linkages and Estimation Strategy

The literature on the coping strategies of concentration camp prisoners includes only a few statistical analyses that investigate what characteristics or strategies helped prisoners to survive. It is important that such analysis is multivariate—in order to compare otherwise comparable prisoners—and that it explores specific mechanisms that underpinned survival. Our focus on a mechanism based on social-linkage resources is motivated by the testimonies of survivors (e.g., ref. 6), which also guide our focus on several dimensions of social linkages. Finding similar effects for multiple measures would be suggestive of systematic forces. SI Appendix provides details on the Theresienstadt data, sources of the archival data on social ties, and examples of relevant survivor testimonies.
A prime social-resource group is that of one’s family and so it forms the basis of our first social-linkage measure. Second, prisoners from the same predeportation place of residence can form a natural mutual-support group. We define such groups based on predeportation street addresses of the Theresienstadt prisoners deported from Prague (there are 1,917 Prague street addresses available in the TII data and 14,791 prisoners with this information on transports to Auschwitz). For other prisoners, we do not have street address data, only the city from which their transport to Theresienstadt came. Third, we form a measure of administrative ties based on the self-organization of national Jewish communities, i.e., based on membership of the official predeportation Jewish self-administrations (Jüdische Kultusgemeinde in Prague, Israelitische Kultusgemeinde in Vienna, Berlin; henceforth referred to as JKG/IKG). We obtained lists of the members of the three organizations in 1941 and merged them with the TII data. The majority of the 2,680 members of JKG Prague, 677 members of IKG Vienna, and 371 members of IKG Berlin who entered Theresienstadt ended up on transports to Auschwitz. The three types of social linkages described above were formed prior to internment.
Next, we consider social linkages formed during internment. A measure of social linkages corresponding to camaraderie is based on the case of young Czech Theresienstadt prisoners, who, according to postwar testimonies discussed in SI Appendix, had often formed strong friendships (based, e.g., on sharing food sent from home) during their earlier internment in a low-security all-male agricultural labor camp, which was located in Lípa in a rural area of today’s Czech Republic. The Lípa camp is an example of the several thousand small labor camps, in which European Jews were interned before being deported to large ghettos and concentration camps (25). We merged records of the Theresienstadt ghetto with lists of Lípa prisoners. A total of 1,351 Czech Jews were interned in the Lípa camp, of whom 961 entered Theresienstadt.
Our final social-linkage measures are based on Theresienstadt networks. First, we observe members of a chain-mail community (104 women and 126 men, most ended up in Auschwitz) formed within Theresienstadt to share a copy of an underground satiric weekly (“Shalom for Friday,” henceforth referred to using the Czech abbreviation “SNAP”). Second, we consider prisoners who came to Theresienstadt on the same in-transport to be potentially socially linked. In-transports often combined residents from a set of predeportation neighborhoods; further, within Theresienstadt, prisoners from the same in-transport often shared similar conditions and housing. Hence, it is possible that they formed relevant social ties.
To identify the effect of social-linkage resources on Holocaust survival in Auschwitz-Birkenau, we use information on the number of potential friends available to prisoners across transports out of Theresienstadt, taking the composition of these transports as a setting in which the social mix of prisoners varied quasirandomly due to the demographic pressure of transport orders given by the SS. (In the next section, we provide evidence supporting this notion.) Consider the 601 former Lípa prisoners who ended up on 23 distinct transports from Theresienstadt to Auschwitz. To identify the effect of social-network resources on survival, we ask whether Lípa prisoners travelling to Auschwitz with a different number of fellow Lípa prisoners display different survival outcomes. We thus use variation in the number of Lípa prisoners across transports and ask whether arriving in Auschwitz with more potential friends improves survival prospects. Our analysis conditions on the average survival chances of all prisoners on a given transport to Auschwitz (by transport fixed effects), which is given by SS decisions in Auschwitz outside of prisoners’ influence. The effects of social linkages thus correspond to the within-transport gaps in survival chances between a typical prisoner and a “Lípa” prisoner, where this gap is contrasted across transports with a varying number of Lípa prisoners. We similarly condition on the number of potential friends on a transport to Auschwitz based on all of our measures of social linkages. (While the number of Lípa prisoners, JKG/IKG members, and SNAP prisoners travelling together varies only across transports to Auschwitz, there is within-transport variation in the size of an individual’s social network based on family size, predeportation place of residence, and on groups of prisoners who came to Theresienstadt on the same in-transport.) We measure social resources by gender given that the camp was segregated by gender. The maximum size of the set of potential friends on a transport varies from 4 (for family networks for both genders) to 295 (for men who arrived in Theresienstadt on the same transport).

Survival after Entering Auschwitz

We model differences in surviving the Holocaust for prisoners entering Auschwitz. We do not observe place or date of Holocaust death for the Theresienstadt prisoners who perish after entering Auschwitz. It is possible that some of these prisoners left Auschwitz for other concentration camps or ended up in one of the death marches at the end of the war. Our estimates thus speak to the extreme experience of a typical prisoner entering Auschwitz, not only to imprisonment in Auschwitz-Birkenau.
Of the 27 transports from Theresienstadt to Auschwitz, seven had survival rates of under 2%. Three transports (transport Ds in 1943 and Ek and Em in 1944) had survival rates of about 20%, as survival rates in Auschwitz improved toward the end of the war. We condition on the transport-specific survival rates, which we consider externally given by the prevailing conditions in Auschwitz, and so we study differences in survival relative to the transport-wide average survival rate. We exploit within-transport as well as across-transport variation in multiple types of social linkages, which we view as quasirandom. The SS specified the demographic composition of transports out of Theresienstadt, but the selection of individual prisoners was under the influence of the ghetto’s Jewish self-administration for most transports. In SI Appendix, we provide a description of the transport selection process in Theresienstadt and show that there is no systematic relationship between transport survival rates and transport averages of social linkages. For eight transports, the selection of prisoners in Theresienstadt was controlled directly by the SS, not by the self-administration; these eight are omitted from our analysis since the selection process may have SS-specific goals (SI Appendix provides additional discussion). We thus study the Holocaust survival of the 14,546 male prisoners and 16,200 female prisoners of Czech, Austrian, and German origin on 19 transports.
There is no evidence in the historical literature suggesting that selection into transports would consider social linkages beyond family ties. Further, the ghetto’s self-administration, which compiled out-transports of one to two thousand prisoners at a time under significant time pressure, did not have data available on many of the social ties we measure here, with the benefit of hindsight. Nevertheless, it is important to consider the possibility that Holocaust survival of prisoners entering Auschwitz was related to their unobservable prosocial traits reflected in the Theresienstadt out-transport selection. Below, we therefore assess the sensitivity of our baseline findings to unobservables by estimating sample selection models.
We test whether the improved ability to form close friendships by prisoners with access to preexisting social linkages on their transport to Auschwitz improves chances of Holocaust survival (binary indicator Sit). For each prisoner i on transport t belonging to social networks of type j, we condition on the number of prisoners from his/her social network traveling on the same transport, denoted Nij (e.g., for the number of Lípa prisoners—“N Lípa”). We also condition on transport indicators αt capturing transport-wide survival levels. Finally, we condition on a set of prisoners’ characteristics Xi consisting of prisoners’ age, length of Theresienstadt imprisonment prior to out-transport to Auschwitz, a Prague-deportation indicator, nationality indicators, and indicators for being a member of the JKG/IKG organizations, having been in the Lípa camp, and for having family members on transport. We thus estimate the following OLS binary-outcome regression:
Sit=αt+jβjNij+Xiγ+εit,
[1]
and its Probit equivalent. We cluster standard errors by transports out of Theresienstadt. Wild bootstrap inference (26) confirms traditional asymptotic inference.
In Table 1, we show results for both male and female prisoners; specifically, we present average marginal effects (AMEs) from Probit and OLS models of Holocaust survival. For our measures of available social linkages, the AMEs represent the effects on survival chances of one additional linked fellow prisoner on a transport (see SI Appendix for details on the calculation of the AMEs). Several types of available social linkages we observe imply that arriving in Auschwitz with a larger group of male potential friends supports survival in extreme circumstances. The estimates in the first two columns of Table 1 suggest that having been imprisoned together earlier, having resided together, and arriving in Theresienstadt together (thus sharing a network in Theresienstadt) generates social ties that confer a survival advantage in a deadly concentration camp. One of the measures of predeportation administrative social ties is similarly helpful. The advantage grows with the size of the group of potential friends, as this increases the chances that prisoners with social links stay together. The estimates for women in the right panel of Table 1 imply that all social linkage measures, including family ties and the SNAP linkages as well as administrative ties, are increasing survival chances. The estimated effect magnitude is consistently larger for women than for men.
Table 1.
Average marginal effects from survival models
 MalesFemales
 (1) Probit(2) OLS(3) Probit(4) OLS(5) Probit(6) OLS(7) Probit(8) OLS
N Lipa0.00101**
(0.000416)
0.00330***
(0.000322)
0.000598**
(0.000266)
0.00325***
(0.000343)
N Family0.00280
(0.00439)
0.00672
(0.00695)
0.00170
(0.00251)
0.00694
(0.00693)
0.0102**
(0.00469)
0.0349**
(0.0139)
0.00428**
(0.00208)
0.0373**
(0.0133)
N Same street0.0160***
(0.00587)
0.0327***
(0.00921)
0.00948***
(0.00286)
0.0328***
(0.00904)
0.0240***
(0.00337)
0.0978***
(0.00896)
0.0111**
(0.00470)
0.0954***
(0.00818)
N SNAP0.00171*
(0.000911)
0.00227
(0.00165)
0.00101**
(0.000473)
0.00216
(0.00163)
0.0101***
(0.00262)
0.0712***
(0.0103)
0.00494**
(0.00212)
0.0692***
(0.00986)
N in-transport0.000195***
(0.0000547)
0.0000920
(0.0000584)
0.000111***
(0.0000309)
0.0000761
(0.0000619)
0.000131
(0.0000923)
0.000511
(0.000345)
0.0000984*
(0.0000506)
0.000403
(0.000306)
N JKG Prague0.0000941
(0.000106)
0.0000232
(0.000104)
0.0000558
(0.0000648)
0.0000327
(0.000100)
0.000862***
(0.000302)
0.00782***
(0.00143)
0.000412**
(0.000201)
0.00773***
(0.00130)
N IKG Vienna-0.00107
(0.00104)
-0.00119
(0.00126)
-0.000666
(0.000634)
-0.00116
(0.00127)
0.00382***
(0.000744)
-0.000281
(0.00151)
0.00200***
(0.000474)
-0.000496
(0.00192)
N IKG Berlin0.0338***
(0.00162)
0.00108
(0.00149)
0.0253***
(0.00423)
0.00116
(0.00146)
0.0388
(0.0238)
0.0395
(0.0438)
0.0184*
(0.0102)
0.0375
(0.0437)
Ac. title-0.0166-0.0248-0.00928-0.02290.02890.1200.01140.125
 (nonmedical)(0.0177)(0.0223)(0.0105)(0.0231)(0.0198)(0.0953)(0.00757)(0.0961)
Doctor0.0529**
(0.0234)
0.0400**
(0.0142)
0.0330***
(0.0121)
0.0423**
(0.0155)
0.0505**
(0.0256)
0.128
(0.0870)
0.0190
(0.0138)
0.139
(0.0821)
Age (in years)-0.00185***
(0.0000828)
-0.00177***
(0.000492)
-0.000709***
(0.000118)
-0.00188***
(0.000452)
-0.00129***
(0.0000558)
-0.00153**
(0.000540)
-0.000586**
(0.000240)
-0.00162***
(0.000520)
pval ρ = 00.6470.1680.0740.028
Clusters1919191919191919
No. of observations14,54614,54614,54614,54616,20016,20016,20016,200
Notes: SEs clustered by transports in parentheses; significance codes: *P <  0.1, **P <  0.05, ***P <  0.01. ρ= correlation coefficient between residuals in the selection and survival equations. Transport fixed effects, effects of time spent in Theresienstadt and of age in years and its square and cube, group membership fixed effects including the Austrian/Czech/German nationality indicators, Prague residency fixed effect, and selection equations not shown (SI Appendix).
In richer specifications presented in our SI Appendix, we confirm that the survival effects of social linkages are linear in the size of social networks, and we find that they are stronger for younger prisoners, who may form mutual-support communes more readily. Alternatively, this finding may mechanically correspond to the fact that younger prisoners had generally higher survival chances, so that there was more scope for social networks to support survival.
We also find in Table 1 that medical doctors were more likely to survive, possibly thanks to their valuable skills, or thanks to doctors arriving at Auschwitz in better-than-average health. Other academic titles were not helpful. The effect of age on survival is nonlinear (SI Appendix), which cannot be conveyed by the age AME. Conditional on transport-wide survival rates and other controls, prisoners in their mid-twenties were most likely to survive among both men and women. Survival chances then decline steeply with age, such that prisoners aged 45 were about 15 percentage points less likely to survive compared to those aged 25.
We rely on variation in social linkages across transports that is driven by the transport selection pressure in Theresienstadt. We thus minimize potential omitted variable bias where prisoners who are more prosocial tend to have more friends and are more likely to survive. However, prisoners in poorer health may have found it more difficult to join social networks in Theresienstadt, such as the network corresponding to the satirical magazine distribution chain, and their poorer heath may have also reduced their survival chances in Auschwitz. We therefore assess the sensitivity of our baseline findings to unobservables. First, we estimate specifications additionally controlling for prisoners’ ability to evade transport selections out of Theresienstadt before eventually ending up on a transport to Auschwitz. This control likely captures unobservable social capital and thus allows us to explore the sensitivity of our baseline findings to transport selection on unobservables. In SI Appendix, we indeed find that Theresienstadt prisoners who were better able to evade transports (conditional on observable controls) were more likely to survive after arriving at Auschwitz, which suggests the importance of some unobservable, including potentially social linkages, for both avoiding selection and survival. Importantly, our main coefficients of interest are not materially affected by whether we control for the evaded-transport-risk regressor. Second, we estimate sample selection models linking unobservables across the transport selection in Theresienstadt and the Auschwitz survival equations. We fit the Probit model with sample selection proposed in ref. 27. In this model, residuals from the transport selection equation are allowed to correlate with residuals in the survival equation, allowing the model to account for the presence of unobserved heterogeneity. Mindful of the fact that this Probit model assumes joint normality of the residuals, which may be violated, we also fit a distribution-free alternative by OLS. Specifically, in the OLS model of survival in Auschwitz, we include a step function with 10 dummy variables controlling for each decile of the selection probability estimated by a first-stage OLS. To identify these models, we rely on exclusion restrictions corresponding to the transport selection pressure in Theresienstadt that arose due to the combination of demographic-type SS transport orders with the size of prisoner groups of a given type in the ghetto at the time. When the selection pressure was higher, prisoners of a given type arriving in the ghetto on recent in-transports were often selected for the next out-transport. This resulted in a prisoner’s deportation origin (the city where the in-transport originated) predicting a prisoner’s out-transport selection chances. Selection pressure also led to prisoner ordering within in-transports (transport numbers assigned to each prisoner on a transport corresponded to their position on transport lists) predicting out-transport selection: Prisoners at the top of an in-transport list were assigned to the next out-transport. In our SI Appendix, we thus estimate transport selection specifications where we control for quartile fixed effects in prisoner order on in-transports and for in-transport origins. We exclude these effects from the Auschwitz survival equation based on our assumption that the grouping of prisoners entering Theresienstadt, which affected their probability of being selected for an out-transport due to demographic-level selection pressure, has no predictive power for survival in Auschwitz after prisoners were regrouped by the out-transport selection process in Theresienstadt. We thus assume that in-transport ordering of prisoners and the availability of different prisoner types for out-transports are independent of prisoner survival chances after entering Auschwitz.
The estimated specifications presented in the last two columns of each panel of Table 1 are consistent with selection on unobservables having little effect. While the estimated correlation coefficient between residuals in the Probit model is insignificant, the joint test of significance of the control function dummies in the OLS model (reported under P-val: ρ = 0) suggests statistical significance for women, which may indicate a violation of the joint normality. Even with a significant contribution from the selection equation, however, the results are similar to our baseline estimates, with the Probit estimated effects being somewhat smaller. We conclude that both in terms of the historical literature on Theresienstadt, and in terms of our estimated models, there is no evidence that would undermine our interpretation of differences in prisoners’ social linkages across transports being quasirandom.
Our baseline findings are also robust to adding transports bound for camps/ghettos in Riga and Raasika to the analysis (see SI Appendix for results based on 31 transports).
The AMEs in Table 1 are particularly large for prisoners who resided at the same street address prior to deportation—a measure of social linkages that is both more precise and available for more prisoners than those based on other approaches. Further, the female AMEs in Table 1 are all larger than the corresponding male effects. In Fig. 1, we offer an alternative assessment of effect magnitudes across all social-network measures, one based not on adding one socially linked prisoner, but on adding one standard deviation in the size of a given network type. Fig. 1, which only visualizes the effects of social networks from Table 1 with P values under 0.1, suggests that the Lípa social ties for men and IKG Vienna ties for women were more helpful than other social linkages, when evaluated based on a standard deviation change. The effect of family linkages appears to be small, even though the family coefficient in Table 1 for women is sizeable; this is due to the fact that family groups travelling together were the smallest social networks we measure (with a maximum network size of 4). Fig. 1 implies that a typical effect of a one-standard-deviation increase in the size of a prisoner’s social network is to improve his/her survival chances by about 2 percentage points. A similar improvement in survival chances would result from a reduction in a prisoner’s age of about 3 y (within the 25 to 45 age bracket, where survival chances decline rapidly with age). The average of transport-wide survival rates across the 19 transports we study is 6%; hence, the 2% effect of social networks corresponds to increasing survival chances by a third of this base rate. These are sizeable effects, especially given that our estimates are likely lower bounds to the extent that our measures of social linkages contain measurement error, and because we do not measure all social ties between prisoners, the base-group prisoner is not fully isolated in the social space of the camp.
Fig. 1.
Expected survival advantage due to social networks larger by one SD. Notes: The graph plots the expected survival advantage (in percentage points, with 95% confidence intervals) based on survival effects reported in columns (1) and (5) of Table 1 and corresponding to a 1-standard-deviation increase in the number of linked prisoners around the mean size of each social network type. Effects that are not significant at the 10% level are not shown.
Overall, we interpret our estimates as implying that the availability of potential friends supports survival in the extreme conditions of a Nazi concentration camp and that groups of socially linked prisoners generate valuable opportunities to form small mutually supportive “communes,” Survival testimonies (in SI Appendix) speak of mutual-support groups of typically two to three friends; the networks we measure correspond to the pool from which such close friends can be recruited. The larger the pool, the (linearly) higher the chances of finding a friend based on preexisting social ties.
Which mechanisms could correspond to the survival effects we uncover? Survival testimonies do not imply that small communes would enforce prosocial behavior within groups. Testimonies of Auschwitz survivors (e.g., ref. 28) imply that it was crucial to get advice on the operations of the camp from more experienced prisoners. However, our measures of social ties capture linkages among prisoners who arrive in Auschwitz together and who are similarly inexperienced in the camps operations. Importantly, we find survival effects for essentially all social networks we measure, for men and women, and for youth and prime-aged adults, based on social networks corresponding to prisoner linkages and links based on predeportation ties. This pattern of our findings suggests that there is a common mechanism at play, one that is not based on a particular advantage such as physical strength, which would be more applicable to ties among young male Lípa prisoners than administrative or residential linkages among women. (Even though Table 1 implies that medical doctors were more likely to survive, in SI Appendix, we do not find any survival advantage of having a medical doctor in one’s social network, which is also consistent with the notion that particular advantages are not behind the pattern of our estimates.) Instead, our results are consistent with the widespread appearance of the Muselmann phenomenon in survivor testimonies (2830), where prisoners who gave up hope and the will to live quickly perished in the extreme conditions of the camps. Survival testimonies (we list in SI Appendix) imply that small groups of friends were formed based on preexisting social ties, where friends shared food and provided emotional support to each other in moments of despair, not only in moments of weakness and ill health, i.e., that such groups also helped to stimulate the will to continue fighting to survive.
The survival effects we estimate in Table 1 are larger for women, suggesting that women may collaborate more than men. Future work is needed to confirm gender differences in mutual-support cooperation under extreme stress and to shed light on the sources of such differences, including, potentially, evolutionary underpinnings. For example, Taylor et al. (31) argue that women are predisposed toward dyadic tend-and-befriend interdependence as stress (threat) levels increase, because selection pressures for caregiving in the face of threat have operated more strongly on women than on men. Our research design is not well suited to ask whether women are differentially providing material versus emotional support; this distinction is also absent from recent research that finds more supportive behavior among women during high-stress episodes (e.g., ref. 32).

Conclusions

Deportation and killing of civilians were prevalent in Europe throughout the 20th century (33) and continues throughout the world today. Investigating the social structure of internment camps is thus relevant not only as a study of history. We assess the importance of social networks in an important Holocaust setting. In the absence of direct information on prisoner friendships, we employ social-linkage proxies based on various preexisting networks.
Our analysis generates complementary evidence to and a statistical check on the large part of the Holocaust literature based on fundamentally selective survival testimonies. It supports this literature in its emphasis on the importance of mutual-support groups as a key survival strategy of prisoners facing extreme survival pressure. Social ties corresponding to shared previous residence, earlier and current shared imprisonment, as well as predeportation ties all generated a significant survival advantage.
The evidence we provide extends the literature on the importance of social links in high-stakes contexts. While Costa and Kahn (5) study the effect of social bonds formed among soldiers in battle for their survival in a deadly POW camp, we study a demographically diverse civilian prisoner population (including women, for whom we find particularly strong effects). Further, we study the effects of prewar social ties as well as linkages formed within prisoner societies in underpinning survival in a deadly camp. Our findings imply that a variety of social ties outside of the close bonds of family or brothers-in-arms support survival and that life-supporting cooperation arises even when survival chances are extremely low.
Our evidence is also relevant to the literature studying parochial altruism—the notion that experience of violent conflict supports within-group cooperation among survivors (1, 2). An alternative mechanism highlighted here is that those more prone to cooperation (having larger social networks) are more likely to survive violent conflicts. Finally, our analysis contributes to the large literature on the importance of social networks for health outcomes (e.g., ref. 3) by providing evidence on the transferability of social linkages generated in normal social environments to the truly extreme conditions of deadly internment camps.

Materials and Methods

We estimate binary-outcome Probit and OLS (ordinary least squares) models of Holocaust survival of Auschwitz prisoners (we do not know their date of death and thus cannot estimate duration models), in which we leverage quasirandom differences in the social-linkage composition of transports to Auschwitz in order to contrast survival chances across prisoners depending on whether they entered the camp alongside a group of socially linked potential friends. We study the importance of predeportation social ties as well as that of prisoner–society social ties: We consider as “socially linked” prisoners who had family ties on the transport; shared a place of residence prior to deportation to Theresienstadt; were formerly interned together in an agricultural-labor camp; were linked by a distribution chain of an underground satirical weekly while in Theresienstadt; arrived in the ghetto on the same in-transport; or had predeportation social ties as members of Jewish self-administrations in Prague, Vienna, and Berlin.

Data, Materials, and Software Availability

The analysis is based on the database of Theresienstadt prisoners (24) provided by the Theresienstadt Initiative Institute (TII), which was founded by Holocaust survivors. The full data can be obtained from the institute. Information about individual Theresienstadt victims of the Holocaust can also be accessed through the online TII database (34). Auxiliary data sources are listed and referenced in SI Appendix.

Acknowledgments

This research was supported by GACR grant no. 19-05523S. We would like to thank the Terezín Initiative Institute for providing access to the Theresienstadt database, in particular Tereza Štěpková and Aneta Plzáková for their constant support. We are also grateful to Alena Jindrová for sharing data on Lípa-camp prisoners, Tomáš Fedorovič for social-linkage data help, and the Jewish Museum in Prague and the Terezín Memorial for access to their archives. We appreciate helpful comments from Michal Bauer, René Böheim, Tomáš Fedorovič, Michal Frankl, Benjamin Frommer, Anna Hájková, Éva Kovács, Tatjana Lichtenstein, and Christian Ochsner. All errors and interpretations are our own.

Author contributions

M.B., T.J., and Š.J. designed research; performed research; contributed new reagents/analytic tools; analyzed data; and wrote the paper.

Competing interests

The authors declare no competing interests.

Supporting Information

Appendix 01 (PDF)

References

1
R. L. Trivers, The evolution of reciprocal altruism. Q. Rev. Biol. 46, 35–57 (1971).
2
J. K. Choi, S. Bowles, The coevolution of parochial altruism and war. Science 318, 636–40 (2007).
3
J. S. House, K. R. Landis, D. Umberson, Social relationships and health. Science 241, 540–545 (1988).
4
E. A. Shils, M. Janowitz, Cohesion and disintegration in the Wehrmacht in World War II. Public Opin. Q. 12, 280–315 (1948).
5
D. L. Costa, M. E. Kahn, Surviving Andersonville: The benefits of social networks in POW camps. Am. Econ. Rev. 97, 1467–1487 (2007).
6
S. Davidson, The Nazi Concentration Camps (Yad Vashem, 1984), pp. 555–572.
7
J. McElroy, This was Andersonville (McDowell Obolensky, New York, NY, 1957).
8
P. Schmolling, Human reactions to the Nazi concentration camps: A summing up. J. Hum. Stress 10, 108–120 (1984).
9
A. Applebaum, Gulag: A History (Doubleday, New York, NY, 2003).
10
L. Eitinger, Concentration Camp Survivors in Norway and Israel (Universitetsforlaget, Oslo, Norway, 1964).
11
E. Luchterhand, Prisoner behavior and social system in the Nazi concentration camps. Int. J. Soc. Psyc. 13, 245–264 (1967).
12
J. E. Dimsdale, The coping behavior of Nazi concentration camp survivors. Am. J. Psychiatry 131, 792–797 (1974).
13
W. Sofsky, The Order of Terror: The Concentration Camp (Princeton University Press, 1999).
14
M. Suderland, Inside Concentration Camps (Polity Press, Cambridge, UK, 2013).
15
A. Hájková, The Last Ghetto. An Everyday History of Theresienstadt (Oxford University Press, 2020).
16
H. G. Adler, J. Adler, Theresienstadt 1941–1945: The Face of a Coerced Community (Cambridge University Press, Cambridge, UK, 2017).
17
M. Frankl, Teresienstädter Gedenkbuch Österreichische Jüdinnen und Juden in Teresienstadt 1942–1945 (Institut Theresienstädter Initiative, 2005), pp. 71–86.
18
K. Lagus, J. Polák, Město za Mřížemi [City behind Bars]. (Baset, Prague, ed. 2, 2006).
19
E. Finkel, Ordinary Jews (Princeton University Press, Princeton, NJ, 2017).
20
D. Battiston, “The persistent effects of brief interactions: Evidence from immigrant ships” (MPRA Paper, 2018).
21
R. Fisman, J. Shi, Y. Wang, R. Xu, Social ties and favoritism in Chinese science. J. Polit. Econ. 126, 1134–1171 (2018).
22
M. Kelly, C. O. Grada, Market contagion: Evidence from the panics of 1854 and 1857. Am. Econ. Rev. 90, 1110–1124 (2000).
23
B. A. Stuart, E. J. Taylor, The effect of social connectedness on crime: Evidence from the Great Migration. Rev. Econ. Stat. 103, 18–33 (2021).
24
Theresienstadt Initiative Institute, TII database of Theresienstadt prisoners (2020).
25
G. P. Megargee, Ed., The United States Holocaust Memorial Museum Encyclopedia of Camps and Ghettos 1933–1945 (Indiana University Press& US Memorial Holocaust Museum, Bloomington, IN, 2009).
26
C. A. Cameron, D. L. Miller, A practitioners guide to cluster–robust inference. J. Hum. Res. 50, 317–372 (2015).
27
W. P. M. M. Van de Ven, B. M. S. Van Praag, The demand for deductibles in private health insurance: A probit model with sample selection. J. Econ. 17, 229–252 (1981).
28
P. Levi, Se Questo un Uomo [If This Is a Man] (De Silva, Italy, 1947).
29
V. E. Frankl, Ein Psycholog erlebt das Konzentrationslager [Man’s Search for Meaning] (Verlag für Jugend und Volk, Vienna, 1946).
30
G. Agamben, Remnants of Auschwitz: The Witness and the Archive (Zone Books, New York, NY, 1999).
31
S. E. Taylor et al., Biobehavioral responses to stress in females: Tend-and-befriend, not fight-or-flight. Psychol. Rev. 107, 411–429 (2000).
32
E. Haller et al., To help or not to help? Prosocial behavior, its association with well-being, and predictors of prosocial behavior during the coronavirus disease pandemic. Front. Psychol. 12 (2022).
33
N. M. Naimark, Fires of Hatred: Ethnic Cleansing in Twentieth-Century Europe (Harvard University Press, Cambridge, MA, 2001).
34
Theresienstadt Initiative Institute, Database of victims. TII online database of Holocaust victims. https://www.holocaust.cz/en/database-of-victims/. Accessed 15 July 2020.

Information & Authors

Information

Published in

Go to Proceedings of the National Academy of Sciences
Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 120 | No. 29
July 18, 2023
PubMed: 37432991

Classifications

Data, Materials, and Software Availability

The analysis is based on the database of Theresienstadt prisoners (24) provided by the Theresienstadt Initiative Institute (TII), which was founded by Holocaust survivors. The full data can be obtained from the institute. Information about individual Theresienstadt victims of the Holocaust can also be accessed through the online TII database (34). Auxiliary data sources are listed and referenced in SI Appendix.

Submission history

Received: December 21, 2022
Accepted: June 12, 2023
Published online: July 11, 2023
Published in issue: July 18, 2023

Keywords

  1. social networks
  2. Holocaust survival
  3. Nazi concentration camp/ghetto

Acknowledgments

This research was supported by GACR grant no. 19-05523S. We would like to thank the Terezín Initiative Institute for providing access to the Theresienstadt database, in particular Tereza Štěpková and Aneta Plzáková for their constant support. We are also grateful to Alena Jindrová for sharing data on Lípa-camp prisoners, Tomáš Fedorovič for social-linkage data help, and the Jewish Museum in Prague and the Terezín Memorial for access to their archives. We appreciate helpful comments from Michal Bauer, René Böheim, Tomáš Fedorovič, Michal Frankl, Benjamin Frommer, Anna Hájková, Éva Kovács, Tatjana Lichtenstein, and Christian Ochsner. All errors and interpretations are our own.
Author Contributions
M.B., T.J., and Š.J. designed research; performed research; contributed new reagents/analytic tools; analyzed data; and wrote the paper.
Competing Interests
The authors declare no competing interests.

Notes

This article is a PNAS Direct Submission. D.L.C. is a guest editor invited by the Editorial Board.

Authors

Affiliations

Center for Economic Research and Graduate Education of Charles University and the Economics Institute of the Czech Academy of Sciences, Praha 1, 111 21, Czech Republic
Healthcare Innovation Institute, Moravian Business College Olomouc, Olomouc 779 00, Czech Republic
Center for Economic Research and Graduate Education of Charles University and the Economics Institute of the Czech Academy of Sciences, Praha 1, 111 21, Czech Republic

Notes

1
To whom correspondence may be addressed. Email: [email protected].

Metrics & Citations

Metrics

Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.


Citation statements




Altmetrics

Citations

If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

View Options

View options

PDF format

Download this article as a PDF file

DOWNLOAD PDF

Get Access

Login options

Check if you have access through your login credentials or your institution to get full access on this article.

Personal login Institutional Login

Recommend to a librarian

Recommend PNAS to a Librarian

Purchase options

Purchase this article to get full access to it.

Single Article Purchase

Preexisting social ties among Auschwitz prisoners support Holocaust survival
Proceedings of the National Academy of Sciences
  • Vol. 120
  • No. 29

Media

Figures

Tables

Other

Share

Share

Share article link

Share on social media