SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References

We use individual-level survey and county-level expenditure data to examine the extent to which Hanukkah celebrations among US Jews are driven by the presence of Christmas. We document that Jews with young children are more likely to celebrate Hanukkah, that this effect is greater for reform Jews and for strongly-identified Jews, and that Jewish-related expenditure on Hanukkah is higher in counties with lower shares of Jews. All these findings are consistent with the hypothesis that celebration of religious holidays is designed not only for worship and enjoyment but also to provide a counterbalance for children against competing cultural influences.

Is religious activity responsive to the presence and activity of other religions? How do religious minorities persist and keep their children from converting? We investigate these questions by examining the extent to which the celebration of Hanukkah, a Jewish holiday that is celebrated in December, is driven by the presence of Christmas. Hanukkah celebration in the US is especially suited to address these questions; US Jews are a minority who account for less than 2% of the population, and conversion and intermarriage, which is estimated at over 40% (United Jewish Communities, 2000), are key concerns among American Jews.

A key observation that motivated this work is that Hanukkah is a minor holiday in Judaism in general and in Israel in particular but it is one of the most celebrated Jewish holidays in the US. Hanukkah is often called the ‘Jewish Christmas’ because American Jewish parents give their children gifts, like their Christian neighbours. Surveys we conducted in both Israel and the US confirm that Hanukkah is perceived to be much less important in Israel. This stark difference in the importance of Hanukkah in Israel (where Jews are a majority) and in the US (where Jews are a minority) suggests that the extent of Hanukkah celebration in the US may be driven by the presence of Christmas. With so many other differences between Israel and the US, however, one should be cautious drawing any interpretation from this anecdotal fact. Our strategy is therefore to look within the US, by comparing the behaviour of different American Jewish households.

Our hypothesis is that Jews with children are more likely to be affected by the presence of Christmas, because Jewish parents might worry that their children would feel left out, intermarry, or convert. That is, Christmas, a fun holiday for children, induces Jewish parents to ‘compete’. Thus, if the presence of Christmas is important, we expect that Jewish parents will celebrate Hanukkah more intensively than Jews without children. To account for the alternative hypothesis that children induce more intensive celebration of all holidays regardless of Christmas, we use the intensity of Passover celebration as a control. To account for the alternative hypothesis that Hanukkah is simply a more fun holiday for children than Passover, we use a difference-in-differences approach whereby we identify groups (secular and reform Jews) that a priori seem more likely than other groups (orthodox Jews) to be responsive to the presence of Christmas; their children interact more with the non-Jewish population and thus may be at a higher ‘risk’ of intermarriage or conversion. Similarly, we identify a group of Jewish parents who are likely to view possible intermarriage or conversion more negatively than others and ask whether their response is stronger. This difference-in-differences strategy is valid under the assumption that whether an individual is reform, orthodox, or secular and whether an individual feels strongly or less strongly about their Jewish identity is an individual ‘type’, which does not change over the life cycle. Under this assumption, comparing individuals of the same type, with and without young children, is similar to comparing the same individual over different stages of their life cycle.

We employ two data sets to examine these effects. The first and primary source is individual-level survey data that contain information on the self-reported intensity of Hanukkah celebration. The second source of data is at the county level and contains information on expenditure on Jewish items during Hanukkah and during other parts of the year. If the presence of Christmas is important and residence location is primarily driven by non-religious factors, then Jews who live in mostly Christian locations are expected to celebrate Hanukkah (compared with other holidays) more intensively. Although the evidence from these data is, by its nature, less conclusive, it complements the survey by providing information on what Jews actually do rather than what they say.

We present four findings. First, Jews with children under 18 are more likely to celebrate Hanukkah than other Jewish holidays. Second, the correlation of having children at home with Hanukkah celebration is highest for reform Jews (who are most exposed to Christmas), followed by conservative Jews, and is lowest for orthodox Jews. Third, the correlation of having children at home with Hanukkah celebration is higher for strongly-identified Jews. In contrast, these differences in correlation are not present for other Jewish holidays. Fourth, ‘Jewish products’ have higher sales at Hanukkah in US counties with a lower share of Jews. These patterns are consistent with the hypothesis that Jews increase religious activity during Hanukkah because of the presence of Christmas and that this response is primarily driven by the presence of children. Jews with children at home may celebrate Hanukkah more intensively so their children do not feel left out and/or because they are concerned their children will convert or intermarry.

Taken together, this article demonstrates that religious activity is at least partially endogenous to the environment in which it takes place, and in particular to the religious activities of ‘competing’ religions. We thus contribute to the literature that incorporates economic analysis into the study of religions (Iannaccone, 1991, 1992, 1998; Iannaccone et al., 1997; Berman, 2000; Gruber, 2005). Our work is also related to the literature that incorporates identity into economics. For example, Akerlof and Kranton (2000) define identity as a person's sense of self, which ‘is associated with different social categories and how people in these categories should behave’. They model identity as altering the payoffs from different actions, such that following the behavioural prescriptions for one's identity enhances one's identity and violating these prescriptions results in anxiety and discomfort. Our article contributes to this literature by providing evidence of parents investing in shaping the identity of their children. Similar and related issues have also been the focus, albeit from a different perspective, of other disciplines, including sociology (Cavan, 1971a, b; Finke, 1990; Finke and Stark, 1992; Kaufman, 2002) and law (Dershowitz, 1997).

1. Background

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References

Hanukkah, also known as the festival of lights, is an annual eight-day Jewish holiday beginning on the 25th day of the third Jewish month of Kislev, which falls between late November and late December, depending on the particular year.1 Hanukkah is celebrated by the lighting of candles on each night of the holiday – one on the first night, two on the second, and so on.2

Hanukkah is not mentioned in the Tanakh (old testament), and it is considered a minor holiday in Jewish tradition. In Israel, where Jewish holidays are recognised officially, Hanukkah is observed as a vacation only in the state's elementary and high schools. Other institutes and companies, private and public, operate as usual. In the US, Hanukkah is considered important as it occurs during the national winter holiday season. Many American Jews regard Hanukkah as the Jewish alternative to Christmas, thus giving it special importance.

This stark difference between Israel and the US in the relative importance of Hanukkah as a Jewish holiday is witnessed by each Israeli immigrant to the US (including ourselves). To provide a more quantitative statement of this difference, we also conducted a short survey among undergraduate students in economics in both Israel and the US, and asked them to list the three most important Jewish holidays. The results are reported in Table 1. They clearly show that Passover and Rosh Hashana (Jewish new year) are consistently ranked as the most important holidays in both Israel and the US, and that other holidays except Hanukkah are secondary and less important. The perceived importance of Hanukkah, however, is very different in the two countries. While in Israel it is ranked together with the other secondary holidays, in the US it is viewed as just as important as Passover and Rosh Hashana, and sometimes even more so.

Table 1. Survey Results Regarding the Perceived Importance of Jewish Holidays
RespondentsIsrael SurveyUS Survey
84123
  1. The survey participants are undergraduate students of economics in Tel Aviv University and Stanford University. The Table reports the percentages of times each holiday was checked (as one of the 3 most important) by each participant. Note that the percentages do not add up to 300% exactly; this is because a small number of respondents did not mark a full list of 3 holidays. We did not adjust the way we count their responses (e.g., by reweighting).

Do you consider this holiday among the 3 most important Jewish holidays? (%)
Rosh Hashana90.578.9
Sukkot34.58.1
Hanukkah38.168.3
Purim8.38.9
Passover96.493.5
Shavuot26.211.4
Don't know1.215.4
Do you think your classmates consider this holiday among the 3 most important Jewish holidays? (%)
Rosh Hashana88.178.0
Sukkot42.91.6
Hanukkah29.895.1
Purim10.74.1
Passover95.291.1
Shavuot21.45.7
Don't know4.85.7

2. Evidence I: Individual-level Survey

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References

2.1. Data

We use the US National Jewish Population Survey, which was collected between August 2000 and August 2001 for the United Jewish Communities and the Jewish Federation System. The data contain information on 5,148 Jewish households. The survey provides individual-level information on the intensity of Hanukkah celebration (defined as the number of candles lit during the most recent Hanukkah) and Passover celebration (defined as whether Passover dinner – the ‘seder’– was celebrated during the most recent Passover). Households are also asked other questions regarding aspects of their Jewish life, such as the degree of their Jewish identity.

Households provide information about their denomination, which often means affiliation with one of three main synagogue movements (orthodox, conservative, reform). While all three are religious movements, they differ in the manner in which they implement their religious observance (Lazerwitz et al., 1998). Orthodox Jews (which are the vast majority of non-secular Jews in Israel) largely follow traditional religious practices, similar to those observed by Jews in Europe in the Nineteenth century. Reform Jews, on the other hand, are more adaptive to changes in the environment, and have adopted practices that are more open and more similar to their Christian neighbours. Reform Jews are more likely to live in mixed neighbourhoods, because unlike orthodox Jews they are permitted to drive on Saturday and thus they do not have to live within walking distance of their synagogue; their children are more likely than orthodox Jewish children to attend public day schools as opposed to Jewish day schools,3 and they are more likely than orthodox Jews to work in and interact with the outside community. Conservative Jews are somewhere in between.

The survey also provides demographic information.4Table 2 lists the key variables we use for the subsequent analysis, and reports their summary statistics.

Table 2. Jewish Population Survey – Variables List and Summary Statistics
 Obs*Mean
  1. *Overall, there were 5,148 survey respondents. Different survey questions, however, are associated with different (low) frequencies of no response, which causes the actual number of observations to vary across variables.

  2. Denomination is given by the respondent's answer to ‘Identification with Jewish religious denominations’. In practice, individuals could list more than a single denomination but fewer than 2% did so, so we assign the first mention to each individual. Throughout the article, we only code the four most frequent denominations, with the rest classified as ‘other’.

  3. Income is a categorical variable, taking the values of 1–11. The standard deviation is 2.369, the 10th percentile is 1 (corresponding to income of less than inline image15,000), the median is 5 (i.e., inline image50,000–75,000), and the 90th percentile is 8 (i.e., inline image150,000–200,000).

  4. §Age has a standard deviation of 18.2, with 10th, 50th, and 90th percentiles equal to 26, 49 and 77, respectively.

Hanukkah celebration: Number of nights lit candles last Hanukkah
 None of the nights (coded as 1)5,1190.291
 Some nights (coded as 2)5,1190.160
 Most nights (coded as 3)5,1190.089
 All eight nights (coded as 4)5,1190.460
Passover celebration: Held/Attended Seder last Passover5,0990.672
Jewish denomination
 Orthodox4,3830.094
 Conservative4,3830.246
 Reform4,3830.323
 ‘Just Jewish’4,3830.202
 Other4,3830.136
Belonging: Answer to ‘Belong to Jewish People?’
 Strongly Disagree4,4450.049
 Somewhat Disagree4,4450.090
 Neutral4,4450.010
 Somewhat Agree4,4450.287
 Strongly Agree4,4450.564
Number of children (under 18) in the household
 05,1140.716
 15,1140.127
 25,1140.107
 >25,1140.050
Other demographics variables:
 Income (categorical)3,7514.771
 Age§5,01450.2
 Male Dummy5,1480.443

2.2. Empirical Strategy

Ideally, we would run a regression of Hanukkah celebration on the extent to which households view themselves as trying to provide a Jewish alternative to Christmas but the latter is not directly observed. We thus identify groups that are more likely to be affected by the presence of Christmas and test whether they celebrate Hanukkah more intensively than other groups. Specifically, it seems natural to view Jews with children under 18 as more likely to be affected by the presence of Christmas. Christmas is a gift-giving holiday and Jewish parents might worry that their children would feel left out. Moreover, the intermarriage rate of American Jews is over 40% and it is a key concern of American Jewry. Jewish parents may be concerned about their children's intermarriage down the road. Hanukkah, which falls close to Christmas, gives parents the opportunity to give their children an exciting alternative and ‘compete’ with Christmas. Thus, we expect that Jewish parents will celebrate Hanukkah more intensively.

There are two potential problems, however, with interpreting the effect of children on Hanukkah celebration as a response to the presence of Christmas. First, Jewish parents may generally be more likely to celebrate holidays (for example, they may want to instill Jewish identity in their children). To account for this possibility, we use as a control the intensity of Passover celebration, which does not fall close to Christmas.5

Second, even if Jewish parents are more likely to celebrate Hanukkah but are not more likely to celebrate Passover, this could be because Hanukkah is a more ‘fun’ holiday for children rather than due to the presence of Christmas. To account for this possibility, we use a difference-in-differences approach whereby we identify groups that a priori seem more likely than other groups to be responsive to the presence of Christmas. We then test whether having children is associated with more Hanukkah celebration in these groups.

In particular, Jewish individuals may be more responsive to Christmas if their children are at a higher ‘risk’ of intermarriage, conversion, or feeling envy and left out during Christmas. Individuals affiliated with the various Jewish denominations naturally differ in this ‘risk’. Specifically, it seems reasonable to assume that, all else equal, reform and conservative Jews are at a higher ‘risk’ of intermarriage and conversion because they (and their children) interact more with the non-Jewish population. Indeed, orthodox Jewish children are the least likely to convert or to outmarry from Judaism (intermarriage rate of 6%), followed by conservative Jews (32%), reform Jews (46%), and secular Jews (49%) (Gordon and Horowitz, 2007; Dershowitz, 1997). Therefore, we expect the effect of having children on Hanukkah celebration to be largest for reform Jews and smallest for orthodox Jews, with conservative Jews in between. Such a differential effect is not likely to occur if Hanukkah is simply more fun for all children.

Specifically, we run the following individual-level OLS regression:

  • image(1)

where HanukkahCelebi is the intensity of Hanukkah celebration by household i, PassoverCelebi is a dummy variable that equals 1 if household i celebrated Passover seder, Childreni is a dummy variable that equals 1 if household i has children (under 18),6Denominationki are five dummy variables for the different Jewish denominations (see Table 2), (Childreni × Denominationki) are interaction variables of the children dummy variable and the various Jewish denomination dummies, and Xi are control variables such as age, gender, and income. The main coefficients of interest are the δ’s.

At the same time, Jewish individuals may be more responsive to Christmas if they view possible intermarriage or conversion more negatively. Specifically, we expect Jewish parents who care more about their Jewish identity to be more likely to celebrate Hanukkah. Therefore, we expect the effect of having children on Hanukkah celebration to be larger for Jews who feel more strongly about their Jewish identity. We run the following individual-level OLS regression:

  • image(2)

where HanukkahCelebi, PassoverCelebi, Childreni, and Xi are as described earlier, JewishIdentityki are five dummy variables for individual i ’s self-reported feeling of belonging to Judaism (see Table 2), and (Childreni × JewishIdentityki) are interaction variables of the children dummy variable and the Jewish identity dummies. The main coefficients of interest are the elements of the vector δ′.

2.3. Results

Figure 1 presents the overall average intensity of Hanukkah and Passover celebration for each group (the two left panels), as well as the incremental effect of having children (the two right panels). That is, a point in the left panels represents the average intensity of celebration (of Hanukkah or Passover) of individuals in a given group, and the right panels present the difference, within each group, between those with children and those without. Since Hanukkah is a categorical variable with four categories and Passover is a dummy variable (see Table 2), we standardise both to have an overall mean of zero and standard deviation of one, so that units are comparable. As could be expected, the left panels of Figure 1 show that Orthodox Jews are on average more likely than reform Jews to celebrate both holidays and that celebration of both holidays is much more likely for Jews who feel more strongly about their Jewish identity. Importantly, the intensity of Hanukkah and Passover celebrations is almost identical within each group. The right panels of Figure 1 show that, for all groups, having children increases the intensity of Hanukkah celebration by 0.2 to 0.5 standard deviations. Children also make Passover celebration more likely for almost all groups but the (standardised) effect is not as large. Most importantly, individuals who are more likely to be affected by Christmas are affected more. In both right panels of Figure 1, the groups of individuals are ordered from those who are (a priori) least likely to be affected by the presence of Christmas to those who are most likely to be affected. Indeed, the effect of children on Hanukkah celebration increases in all panels as we move to the right. In contrast, the increased intensity of Passover celebration due to the presence of children does not show any obvious pattern.

image

Figure 1. Intensity of Hanukkah and Passover Celebrations In both panels on the left, the plotted points represent the sample means (of the standardised Hanukkah and Passover variables) of the corresponding category defined on the horizontal axis. These do not average to zero in the top left panel because there is an omitted ‘other’ category (which is hard to interpret, so is not in the Figure). In both panels on the right, the plotted points represent the difference in means (of the standardised Hanukkah and Passover variables) between those households with children and those with no children for the corresponding category defined on the horizontal axis.

Download figure to PowerPoint

Table 3 subjects the relationship between Hanukkah celebration and having children to a regression analysis, as described earlier. In both panels of the Table, columns (1) and (2) present linear probability models and columns (3) and (4) present probit regressions.7 The results are remarkably stable across all columns. Panel (a) suggests that having children is associated with more Hanukkah celebration and that orthodox Jews celebrate Hanukkah most intensively, followed by conservative Jews, reform Jews and unaffiliated Jews. The key coefficients of interest are the interactions between having children and the various denominations. Panel (a) shows that, consistent with our hypothesis and with Figure 1, the effect of having children on Hanukkah celebration is highest for reform Jews and those without affiliation, and lowest for orthodox Jews. Notice that the regressions control for Passover celebration, which, as expected, is positively correlated with Hanukkah celebration.

Table 3. Determinants of Hanukkah Celebration (a) The differential effect by Jewish denomination
 OLS RegressionsProbit Regressions
 Dep. Var.: Hanukkah Celebration (standardised)Dep. Var.: Hanukkah Celebration (binary)
 (1)(2)(3)(4)
 Coef.Std. Err.Coef.Std. Err.dF/dXz-statdF/dXz-stat
Passover Celebration (standardised)0.337(0.014)0.345(0.017)0.158(18.47)0.164(15.93)
Orthodox Jewish0.502(0.057)0.571(0.073)0.333(9.89)0.352(8.28)
Conservative Jewish0.216(0.029)0.283(0.039)0.246(9.05)0.248(7.44)
Reform Jewish−0.003(0.026)0.001(0.034)0.143(5.17)0.108(3.20)
‘Just Jewish’−0.206(0.032)−0.139(0.040)0.054(1.78)0.044(1.22)
(Children > 0) × Orthodox Jewish0.166(0.082)0.126(0.100)0.177(2.75)0.134(1.65)
(Children > 0) × Conservative Jewish0.277(0.059)0.225(0.070)0.188(5.14)0.171(3.76)
(Children > 0) × Reform Jewish0.448(0.050)0.499(0.059)0.244(8.97)0.269(8.09)
(Children > 0) בJust Jewish’0.477(0.065)0.427(0.074)0.232(7.13)0.224(5.72)
Income (standardised)  0.018(0.016)  0.020(1.96)
Age (standardised)  −0.023(0.016)  −0.006(−0.59)
Male  −0.146(0.030)  −0.083(−4.28)
Number of observations (respondents)4,3213,0734,3213,073
R-Squared/Pseudo R-Squared0.2740.2850.1760.190
Log Likelihood    −2,354.2−1,669.3
Table 3.    (b) The differential effect by Jewish belonging
 OLS RegressionsProbit Regressions
 Dep. Var.: Hanukkah Celebration (standardised)Dep. Var.: Hanukkah Celebration (binary)
 (1)(2)(3)(4)
 Coef.Std. Err.Coef.Std. Err.dF/dXz-statdF/dXz-stat
  1. ‘Standardised’ implies that the value of the variable was standardised to have mean zero and a standard deviation of one (in the entire sample) to ease interpretation of the coefficients.

  2. All regressions in (a) also include the ‘other’ denomination category (alone, and interacted with the children dummy), so (in both panels) the coefficients on the interaction terms should be interpreted as the incremental effect of children for each category.

  3. The dependent variable in the probit regressions is equal to 1 if Hanukkah celebration is equal to 3 or 4 (see Table 2). The results are very similar if we define the dependent variable to be 1 for only values of 4, for values of 2–4, or if we run the regression as an ordered probit.

Passover Celebration (standardised)0.338(0.014)0.341(0.017)0.158(18.530)0.163(15.740)
Belong to Jewish People? Strongly Disagree’−0.426(0.067)−0.382(0.079)— dropped —— dropped —
Belong to Jewish People? Somewhat Disagree’−0.376(0.049)−0.328(0.058)0.012(0.230)0.005(0.090)
Belong to Jewish People? Neutral’−0.387(0.138)−0.405(0.162)−0.026(−0.260)−0.137(−1.03)
Belong to Jewish People? Somewhat Agree’−0.128(0.028)−0.101(0.035)0.140(3.180)0.133(2.440)
Belong to Jewish People? Strongly Agree’0.156(0.020)0.209(0.029)0.296(6.650)0.298(5.460)
(Children > 0) בBelong? Strongly Disagree’0.089(0.124)0.005(0.137)0.066(0.900)0.007(0.070)
(Children > 0) בBelong? Somewhat Disagree’0.419(0.094)0.349(0.107)0.201(4.250)0.189(3.280)
(Children > 0) בBelong? Neutral’0.568(0.296)0.606(0.351)0.287(2.240)0.336(2.340)
(Children > 0) בBelong? Somewhat Agree’0.421(0.052)0.419(0.060)0.223(8.160)0.233(7.010)
(Children > 0) בBelong? Strongly Agree’0.401(0.037)0.361(0.047)0.263(10.870)0.243(7.970)
Income (standardised)  0.008(0.016)  0.011(1.100)
Age (standardised)  −0.047(0.016)  −0.022(−2.14)
Male  −0.122(0.030)  −0.067(−3.49)
Number of observations (respondents)4,3813,1274,3813,127
R-Squared/Pseudo R-Squared0.2660.2790.1740.187
Log Likelihood    −2,402.7−1,709.9

Panel (b) repeats a similar analysis, where instead of denominations, individuals are classified to different groups according to their sense of belonging to Judaism. The pattern is similar. Individuals who feel more strongly about their Jewish identity celebrate Hanukkah more while the correlation of having children at home with Hanukkah celebration is the lowest for individuals who have the weakest sense of belonging to Judaism. The key finding in panel (b) is that the effect is smallest for people with the weakest Jewish identity, consistent with our hypothesis. Although the point estimates suggest a non-monotone effect of having children, peaking for individuals who are neutral with respect to their Jewish identity, this pattern is statistically insignificant as only 1% of individuals are neutral (see Table 2).8

It is important to notice a key conceptual difference between the two panels of Table 3. In panel (b) the effect of having children is highest for the groups who celebrate Hanukkah most intensively even in the absence of children. One possible concern is that these results could be driven by a level effect. That is, if the effect of children were multiplicative, rather than additive, the results may change. For this reason, we are reassured by the results in panel (a), in which the individuals who celebrate Hanukkah less are those who are most responsive to the presence of children.

One possible concern with our empirical strategy and our interpretation of the results is that the two panels of Table 3 may confound each other. That is, if denomination and Jewish identity are correlated with each other, it is possible that the estimated denomination effect (panel (a)) is confounded by an omitted Jewish identity variable and the estimated Jewish identity effect (panel (b)) is confounded by an omitted denomination. Indeed, Table 4 reports the joint distribution of denomination and Jewish identity, and shows that while almost all orthodox and conservative have a strong Jewish identity, there are significant portions of reform and unaffiliated Jews who feel less strongly about their Judaism.

Table 4. The Joint Distribution of Denomination and Jewish Identity (%)
  Jewish Denomination
  OrthodoxConservativeReform‘Just Jewish’Total
  1. The cross-tabulation is based on 3,860 survey respondents (who responded to the belonging question and identified themselves with one of the above four denominations).

Feel a Strong Sense of Belonging to the Jewish PeopleStrongly Disagree11373
Somewhat Disagree149138
Neither Agree nor Disagree00021
Somewhat Agree622333728
Strongly Agree9172544060
Total100100100100100

To ameliorate these concerns, we repeat the analysis of Table 3 for different subsamples. Columns (2) and (3) of Table 5 report the results.9 In panel (a) we show that the denomination effect (reported in Table 3(a) and replicated in column (1) of Table 5) remains essentially the same even when we restrict our sample to those with strong Jewish identity. That is, we exclude from the sample those with weak Jewish identity, who are mostly reform and unaffiliated Jews, and this does not change the results. Similarly, in Table 5(b) we show that the Jewish identity effect is the same even within each denomination.

Table 5. Robustness (a) The differential effect by Jewish denomination
SampleOLS Regressions; Dependent Variable: Hanukkah Celebration (standardised)
Base sampleBelonging? Somewhat or Strongly AgreeBelonging? Strongly AgreeCelebrated PassoverDid not Celebrate Passover
(1)(2)(3)(4)(5)
Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.
Passover Celebration (standardised)0.337(0.014)0.326(0.016)0.310(0.02)    
Orthodox Jewish0.502(0.057)0.528(0.056)0.568(0.06)0.762(0.06)−0.140(0.18)
Conservative Jewish0.216(0.029)0.245(0.030)0.300(0.03)0.476(0.03)−0.394(0.08)
Reform Jewish−0.003(0.026)0.031(0.027)0.065(0.03)0.193(0.03)−0.324(0.07)
‘Just Jewish’−0.206(0.032)−0.154(0.036)−0.017(0.05)−0.016(0.04)−0.619(0.06)
(Children > 0) × Orthodox Jewish0.166(0.082)0.154(0.081)0.131(0.08)0.157(0.08)−0.082(0.40)
(Children > 0) × Conservative Jewish0.277(0.059)0.286(0.059)0.265(0.07)0.273(0.06)0.226(0.20)
(Children > 0) × Reform Jewish0.448(0.050)0.481(0.052)0.515(0.07)0.453(0.05)0.523(0.15)
(Children > 0) בJust Jewish’0.477(0.065)0.471(0.071)0.440(0.10)0.545(0.08)0.370(0.12)
Number of observations (respondents)4,3213,6742,4483,2671,054
R-Squared/Pseudo R-Squared0.2740.2860.3360.2800.279
Table 5.    (b) The differential effect by Jewish belonging
 OLS Regressions; Dependent Variable: Hanukkah Celebration (standardised)
Base sampleReform Jews‘Just Jewish’Celebrated PassoverDid not Celebrate Passover
 (1)(2)(3)(4)(5)
SampleCoef.Std. Err.Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.Coef.Std. Err.
  1. All regressions above use the same specification as column (1) of Table 3. For comparability, the first column above replicates Table 3, and the other columns estimate the same specification on different subsamples. The results are essentially the same if we repeat a similar analysis for the other specifications reported in Table 3.

  2. Note that we do not stratify the regressions in Panel (b) by Orthodox and Conservative Jews, because there is hardly any variation in the belonging variable for these groups (see Table 4).

Passover Celebration (standardised)0.338(0.014)0.196(0.028)0.280(0.030)    
Belong to Jewish People? Strongly Disagree’−0.426(0.067)−0.330(0.158)−0.127(0.133)−0.228(0.103)−0.887(0.096)
Belong to Jewish People? Somewhat Disagree’−0.376(0.049)−0.296(0.088)−0.613(0.101)−0.241(0.066)−0.760(0.080)
Belong to Jewish People? Neutral’−0.387(0.138)0.705(0.374)−0.434(0.262)−0.216(0.195)−0.819(0.213)
Belong to Jewish People? Somewhat Agree’−0.128(0.028)0.015(0.047)−0.335(0.061)0.099(0.031)−0.589(0.057)
Belong to Jewish People? Strongly Agree’0.156(0.020)0.112(0.037)−0.017(0.054)0.397(0.020)−0.351(0.054)
(Children > 0) בBelong? Strongly Disagree’0.089(0.124)0.092(0.281)−0.215(0.263)0.137(0.176)0.058(0.192)
(Children > 0) בBelong? Somewhat Disagree’0.419(0.094)0.504(0.162)0.923(0.218)0.387(0.124)0.464(0.156)
(Children > 0) בBelong? Neutral’0.568(0.296)−0.244(0.700)1.219(0.944)0.814(0.436)0.396(0.444)
(Children > 0) בBelong? Somewhat Agree’0.421(0.052)0.467(0.085)0.565(0.110)0.502(0.058)0.215(0.111)
(Children > 0) בBelong? Strongly Agree’0.401(0.037)0.527(0.071)0.448(0.120)0.403(0.037)0.359(0.135)
Number of observations (respondents)4,3811,3958593,2821,099
R-Squared/Pseudo R-Squared0.2660.1870.2020.2720.261

Finally, columns (4) and (5) of Table 5 address a different possible concern, regarding the linear and additively separable way by which we control for Passover celebration. In these columns we show that both the denomination effect and the Jewish identity effect are quite stable, even when we run the exercise separately for those who celebrate Passover and for those who do not.

3. Evidence II: County-level Expenditure

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References

In this Section we supplement the survey data analysis with data on actual purchasing behaviour. We use three sources to construct the data. First, we collected weekly store-level data from a large grocery retail chain, which operates stores in various parts of the US. In particular, we obtained data on the weekly sales of ‘Jewish products’ (as categorised by the retailer). The data we obtained covers sales in 1,109 stores between October 2004 and October 2005. We aggregated these data to the county level (to match the other data sets described below) based on store zip codes and classified sales into the different Jewish holidays based on dates.10 We then matched these data with county-level data on 150 religious bodies collected through the ‘Religious Congregations Membership Study’ in 2000. These data contain the number of adherents and the number of congregations in each county. We supplemented these with county-level census data. Table 6 and its notes describe all the variables used for this analysis and provide summary statistics.

Table 6. County-level Analysis – Variables List and Summary Statistics
 Obs.MeanStd. Dev.10th pctile50th pctile90th pctile
  1. The data cover all counties in which we observe at least one retailer store. There are 105 counties, covering the following states (number of counties in parentheses): CA (36), WA (11), TX (10), MD (9), IL (6), VA (6), AK (5), HI (4), NJ (4), PA (4), MT (3), NV (3), DE (2), DC (1), ID (1).

  2. Sales of Jewish products is measured by the total dollar value of sales of all products categorised (by the retailer) as ‘Jewish products’ in all stores operating in the corresponding county. There are almost 3,000 distinct products (UPCs or ‘barcodes’) that are classified as Jewish, although only a small fraction of them would typically be available in a given store. The products cover a range of food items (Matzo balls, Geflite fish etc.), although they also include kosher drinks and a small number of non-food items typically sold in grocery stores, such as Shabat and Hanukkah candles.

  3. Total sales is the sum of all Jewish product sales over the entire period we observe it (10/3/2004 – 8/16/2005). The holiday-specific sales are the sum of sales of all Jewish products over the week of the holiday and the week that preceded it. For Passover we use one additional preceding week because Passover preparation is typically longer. (The subsequent results are similar if we use the same window for all holidays.) We note that Purim sales may well be confounded with early Passover sales due to the proximity between the holidays.

  4. Adherents is the number of Jewish, Protestant and Catholic adherents in the county, based on the ‘Religious Congregations Membership Study’ from the year 2000. Adherents of other streams/religions are excluded from the analysis. The excluded adherents account for 3.2% of the total adherents in the counties we use for the analysis. We note that total adherents account for only 34.4% of the total population in the counties we use.

Sales of Jewish products ($)
 Total10526,34559,6655065,67375,160
 Around Hanukkah1051,0412,383142133,752
 Around Passover1058,48921,7541081,48121,943
 Around Rosh Hashana1051,5513,951132184,635
 Around Purim1052,3054,823425459,524
Adherents (000s)
 Jewish10518.860.80.12.338.3
 Catholic105180.1436.08.068.3306.4
 Protestant105108.6167.412.357.7196.5
Median income ($000s)10550.411.834.748.667.0

If the presence of Christmas is important, then we expect that Jewish households who live in areas with a large fraction of Jews are likely to live in Jewish communities, so the concern of Christmas may be less important. In contrast, it is natural to expect that Jews who live in mostly Christian locations will celebrate Hanukkah (compared with other holidays) more intensively. To test this hypothesis, we investigate whether expenditures on Jewish products during Hanukkah (compared with other Jewish holidays) are lower in counties that contain fewer Jews. We run the following county-level regression:

  • image(3)

where HanukkahExpenditurej and PassoverExpenditurej are the expenditures on Jewish products in county j around Hanukkah and Passover respectively, (JewishAdherents/TotalAdherents)j is the fraction of Jewish adherents out of the overall adherents in county j, and Xj is a vector of control variables. The main coefficient of interest is η1, which we expect to be negative.

The main reason that we work with ratios of the variables rather than with levels is the large variation in county sizes and even larger variation in the number and size of stores of the retailer in different counties (see also Table 6). Some of the additional variables in the regressions control for the size of the county and the overall volume of sales.

An obvious concern about this exercise is selection. It seems likely that Jews who are concerned about their children converting would choose to live in larger Jewish communities, or in counties with a higher fraction of Jews. While it is hard to fully address this selection problem using the data we have, we note that this possible selection issue will confound the analysis and work against our hypothesis. If individuals who care more about Judaism and therefore live in larger Jewish communities celebrate Hanukkah more intensively, this will bias our estimate of η1 upwards.

The results are presented in Table 7. Panel (a) presents the results when we use Passover as the ‘control holiday’, while panels (b) and (c) repeat the same analysis using Rosh Hashana and Purim respectively instead. Across all specifications, the coefficient on the ratio of Jewish adherents to total adherents is negative, with elasticities ranging from 1% to 6%, which are statistically significant or very close to it, depending on the specification. Interestingly, we also find a larger, very stable and sometimes statistically significant effect of the ratio of Catholic adherents to total adherents. This effect is consistent with the work of Rebhun (1999), who suggests that while both Jewish-Catholic and Jewish-Protestant (Protestants are the omitted category in all the regressions of Table 7) marriages have a negative effect on Jewish identity, the effect is more pronounced when Jews marry Catholics. This suggests that Catholic people may impose a higher ‘conversion threat’ than Protestants, consistent with a positive coefficient on the fraction of Catholic adherents. Another possible explanation for the large and stable effect of Catholics on Hanukkah expenditure among Jews is that Christmas celebration among Catholics might be more intense and visible than it is among Protestants. Overall, we conclude that individuals who live in larger Jewish communities or in smaller Catholic communities, who are presumably less affected by the presence of Christmas, celebrate Hanukkah less intensively compared with how much they celebrate other Jewish holidays.

Table 7. County-level Regressions of Hanukkah Sales
  1. * Statistically significant at a 10% confidence level; ** Statistically significant at a 5% confidence level.

  2. Total sales of Jewish products contains all holiday sales. The results remain essentially the same if this is replaced by sales of Jewish products over the entire year except these holidays.

(a) Dependent variable: Log (Hanukkah Sales/Passover Sales + 1)
 (1)(2)(3)
Log (Jewish adherents/Total adherents)−0.008(0.005)−0.014**(0.006)−0.014*(0.007)
Log (Catholic adherents/Total adherents)0.031(0.021)0.024(0.023)0.024(0.024)
Log (Total adherents)  −0.013(0.010)−0.013(0.010)
Log (Total sales of Jewish products)  0.012*(0.007)0.013*(0.008)
Log (Adherents Herfindahl Index)    −0.007(0.073)
Log (Median county income)    −0.013(0.036)
Number of observations (counties)105105105
R-Squared0.0380.0680.069
(b) Dependent variable: Log (Hanukkah Sales/Rosh-Hashana Sales + 1)
 (1)(2)(3)
Log (Jewish adherents/Total adherents)−0.055**(0.020)−0.051**(0.025)−0.062**(0.029)
Log (Catholic adherents/Total adherents)0.074(0.087)0.111(0.094)0.117(0.095)
Log (Total adherents)  −0.043(0.038)−0.038(0.040)
Log (Total sales of Jewish products)  0.009(0.034)0.007(0.035)
Log (Adherents Herfindahl Index)    −0.222(0.293)
Log (Median county income)    0.006(0.134)
Number of observations (counties)979797
R-Squared0.0760.0960.102
(c) Dependent variable: Log (Hanukkah Sales/Purim Sales + 1)
 (1)(2)(3)
Log (Jewish adherents/Total adherents)−0.011(0.013)−0.030*(0.016)−0.025(0.018)
Log (Catholic adherents/Total adherents)0.152**(0.053)0.132**(0.057)0.132**(0.057)
Log (Total adherents)  −0.043*(0.023)−0.048**(0.024)
Log (Total sales of Jewish products)  0.041**(0.017)0.047**(0.019)
Log (Adherents Herfindahl Index)    0.091(0.176)
Log (Median county income)    −0.063(0.088)
Number of observations (counties)104104104
R-Squared0.0780.1300.137

A possible concern is that the retailer from which we obtained the data primarily (although not only) sells Jewish food, while Hanukkah have many non-food items, such as candles, candleholders, decorations, toys and chocolate coins, which are also sold elsewhere, possibly leading to differential shopping patterns during Hanukkah and other holidays. This is the main reason we use (in Table 7(c)) Purim as an alternative ‘control holiday’. Purim is a Jewish holiday heavily associated with non-food items, such as Halloween-like customs and graggers (noise-makers). We are encouraged that the qualitative results are similar.11

A related concern with our empirical strategy is that in counties with higher proportions of Jews, shopping for Jewish products may be carried out outside of the retail chain on which we have data, possibly in retail outlets aimed specifically at Jewish consumers, such as Jewish bakeries and kosher butchers. If such shopping is carried out differentially around Hanukkah and Passover, this may confound our interpretation. To account for this possibility, we obtained data on sales of Jewish products from the Nielsen Homescan data, which consist of a panel of households who record (at home) all their food purchases, from all stores and channels.12 Thus, the advantage of the Homescan data over the store-level data we used is that they allow us to observe sales of Jewish products in a wide variety of store types. The chief (and important) disadvantage, however, is that we cannot aggregate sales to the county level but only to one of 50 large urban markets (roughly, MSAs), which may be too large and heterogeneous to obtain an empirically meaningful measure of the relevant community. Still, as a robustness check, we regress the ratio between the (log) expenditure on Jewish products around Hanukkah to the Jewish product expenditure in each of the other holidays (as in Table 7) on the (log) fraction of Jewish adherents out of the overall adherents in the market (with no additional controls). We are encouraged that in all three cases, the coefficients are negative, although (probably due to the small number of observations) largely statistically insignificant.13

4. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References

In this article we present evidence that is largely consistent with a story that the importance of Hanukkah among American Jews is driven by its proximity (in the time dimension) to Christmas, and that many American Jews use Hanukkah as a way to provide their children with an exciting alternative. Extrapolating this story out of the data, it may also explain why Hanukkah is such a popular and important holiday among Jews living in the US, even though it is a much less important Jewish holiday in Israel, where ‘competition’ from Christmas is largely absent.

The effect of Christmas on other cultures goes beyond its effect on Jews and on Hanukkah celebration. Morean and Skov (1993), for example, document the effect of Christmas in Japan. Another example is Kwanzaa, an African-American holiday celebrated around Christmas (almost entirely in the US), which also ‘competes’ with Christmas.14 These effects may become even more widespread if the importance of Christmas continues to increase as it has over recent decades (Scott, 1995). It is also worth noting that Christmas itself and the dates of its celebration were influenced by earlier pagan winter celebrations.

One natural idea for further research is to investigate the behaviour of Jews who live in predominantly Muslim countries and analyse whether Jews in such countries respond to ‘attractive’ Muslim holidays. More broadly, we think that this article highlights the fact that religious behaviour is endogenous to the environment in which it takes place. We looked at Judaism but it is natural to speculate that other religions respond in other contexts in similar ways. This seems a promising avenue for future research.

Footnotes
  • 1

     In principle, this variation in the exact timing of Hanukkah could produce very useful variation for the question at hand. Unfortunately, as described later, the relevant data sets we could find are cross-sectional, so at least this article cannot exploit this excellent variation.

  • 2

     The interested reader may wonder what is being celebrated in Hanukkah. Hanukkah commemorates the rededication of the Temple of Jerusalem after its desecration by Antiochus IV, king of Syria and ruler of the land of Israel. Around 200 bce Jews lived autonomously in the land of Israel. The Jews paid taxes to Syria and accepted the king's legal authority. By and large, they were free to follow their own faith. By 175 bce, Antiochus IV Epiphanes ascended to the Seleucid throne and Jews were gradually forced to violate their faith. Jews rebelled, the Temple in Jerusalem was looted, and Judaism was outlawed. In 167 bce, when Antiochus ordered an altar to Zeus brought to the Temple, a Jewish priest (Mattathias) and his five sons led a rebellion against Antiochus. The Jewish revolt against the Seleucid monarchy was eventually successful and the Temple was liberated. The festival of Hanukkah was instituted by Judah Maccabee and his brothers to celebrate this event. According to the Talmud, at the re-dedication of the Temple in Jerusalem following the victory of the Maccabees over the Seleucid Empire, there was only enough oil to fuel the menorah in the Temple for one day. Miraculously, the oil burned for eight days, the time needed to prepare a new supply of oil. Hanukkah commemorates this miracle and symbolises the miraculous survival of the Jewish people through millennia of suffering and persecution.

  • 3

     However, the children of reform Jews are more likely than orthodox Jews’ children to attend Jewish Sunday schools.

  • 4

     This also includes information about the Metropolitan Statistical Area (MSA) in which the household resides. However, with most surveyed households living in only few locations (almost half of the sample lives in the New York City area), the geographic variation is quite limited, and we do not use it. A detailed description of the survey by the Federation of North America can be found at http://www.ujc.org/page.html?ArticleID=9451. A methodological Appendix can be found at http://www.ujc.org/page.html?ArticleID=46185.

  • 5

     One possible concern is that Passover falls close to Easter. Note, however, that to the extent that this is a problem (i.e., that Passover intensity is increased in response to Easter), this should make us less likely to find what we report below. Moreover, our results remain qualitatively the same when we use the intensity of celebrating Rosh Hashana as a control (instead of Passover).

  • 6

     While we have information about the number and ages of children in the household, it turned out that incorporating this additional information into the subsequent regression analysis made little difference to the results.

  • 7

     In the latter we code the Hanukkah variable as a dummy variable that is equal to 1 when the original Hanukkah variable takes a value of 3 or 4. Other ways to code the variable do not affect the results. This is to be expected, as Hanukkah mostly takes values of 1 and 4 (see Table 2).

  • 8

     Specifically, we cannot reject the test that all four coefficients on the interaction terms (except the first) are equal to each other (F(3,4370) = 0.13, p-value = 0.94).

  • 9

     We report the results using the specification of column (1) in Table 3. Using any of the other specifications reported in Table 3 leads to essentially identical results.

  • 10

     We initially planned to also categorise the products by holidays but it turned out that those Jewish products that had the most sale volume were hard to associate with specific holidays, leaving us with too little volume for the products we could categorise.

  • 11

     While it is encouraging that we obtain similar results, we also note that this is a weak robustness exercise. First, Purim is a minor holiday in the US and it is not celebrated as widely as Hanukkah, Passover or Rosh Hashana (indeed, it is one of the two least important holidays in the survey we conducted among US University students; see Table 1). Second, Purim falls about a month before Passover, so much of the Jewish product sales around Purim may merely reflect early Passover preparations.

  • 12

     See http://www.nielsen.com/clients/index.html for additional information about Nielsen Homescan data. We obtained the 2004 data. To focus on sales of Jewish products, we limit our analysis to those products classified (by our retailer, as in the primary data) as ‘Jewish products’.

  • 13

     Specifically, the estimated coefficients (standard error in parentheses) are −0.78 (0.96), −3.76 (1.92) and −1.69 (1.29), when the ‘control holiday’ (the denominator of the dependent variable, as in Table 7) is Passover, Rosh Hashana and Purim, respectively.

  • 14

     The founder of Kwanzaa stated that ‘…it was chosen to give a Black alternative to the existing holiday and give Blacks an opportunity to celebrate themselves and history, rather than simply imitate the practice of the dominant society’ (Karenga, 1977; p. 21).

References

  1. Top of page
  2. Abstract
  3. 1. Background
  4. 2. Evidence I: Individual-level Survey
  5. 3. Evidence II: County-level Expenditure
  6. 4. Conclusions
  7. References
  • Akerlof, G. and Kranton, R. (2000). ‘Economics and identity’, Quarterly Journal of Economics, vol. 115(3) (August), pp. 71553.
  • Berman, E. (2000). ‘Sect, subsidy, and sacrifice: an economist's view of ultra-orthodox jews’, Quarterly Journal of Economics, vol. 115(3) (August) pp. 90553.
  • Cavan, R. (1971a). ‘Jewish student attitudes toward interreligious and intra–jewish marriage’, American Journal of Sociology, vol. 76(6) (May), pp. 106471.
  • Cavan, R. (1971b). ‘A dating marriage scale of religious social distance’, Journal for the Scientific Study of Religion, vol. 10(2) (June), pp. 93100.
  • Dershowitz, A. (1997). The Vanishing American Jew: In Search of Jewish Identity for the Next Century, Boston, MA: Little, Brown and Company.
  • Finke, R. (1990). ‘Religious deregulation: origins and consequences’, Journal of Church and State, vol. 32(3) (Summer), pp. 60926.
  • Finke, R. and Stark, R. (1992). The Churching of America, 1776–1990: Winners and Losers in Our Religious Economy, New Brunswick, NJ: Rutgers University Press.
  • Gordon, A. and Horowitz, R. (2007). ‘Will your grandchildren be Jews?’, Jewish World (online), available at http://www.aish.com/jewishissues.
  • Gruber, J. (2005). ‘Religious market structure, religious participation, and outcomes: is religion good for you?’, Advances in Economic Analysis and Policy, vol. 5(1), pp. 130.
  • Iannaccone, L.R. (1991). ‘The consequences of religious market structure: Adam Smith and the economics of religion’, Rationality and Society, vol. 3(2) (April), pp. 15677.
  • Iannaccone, L.R. (1992). ‘Sacrifice and stigma: reducing free-riding in cults, communes, and other collectives’, Journal of Political Economy, vol. 100(2) (April), pp. 27191.
  • Iannaccone, L.R. (1998). ‘Introduction to the economics of religion’, Journal of Economic Literature, vol. 36(3) (September), pp. 146596.
  • Iannaccone, L.R., Finke, R. and Stark, R. (1997). ‘Deregulating religion: the economics of church and state’, Economic Inquiry, vol. 35(2) (April), pp. 35064.
  • Karenga, M. (1977). Kwanzaa: Origin, Concepts, Practice, Los Angeles, CA: Kawaida Publications.
  • Kaufman, J. (2002). ‘The political economy of interdenominational competition in late-nineteenth-century American cities’, Journal of Urban History, vol. 28(4) (May), pp. 44565.
  • Lazerwitz, B., Dashefsky, A. and Tabory, E. (1998). Jewish Choices: American Jewish Denominationalism, Ithaca, NY: SUNY Press.
  • Morean, B. and Skov, L. (1993). ‘Cinderella Christmas: kitsch, consumerism, and youth in Japan’, in D.Miller (ed.), Unwrapping Christmas, pp. 10533, Oxford: Oxford University Press.
  • Rebhun, U. (1999). ‘Jewish identification in intermarriage: does a spouse's religion (Catholic vs. Protestant) matter?’. Sociology of Religion: A Quarterly Review, vol. 60(1), pp. 7188.
  • Scott, A. (1995). ‘Why is consumption so seasonal?’, CEP Discussion Paper No. 0269, London School of Economics.
  • United Jewish Communities (2000). National Jewish Popoulation Survey, available at http://www.ujc.org/page.aspx?id=33650.