Effect of religious rules on time of conception in Romania from 1905 to 2001
Claudiu Herteliu, Bogdan Vasile Ileanu, Marcel Ausloos, Giulia Rotundo
EEffect of religious rules on time of conceptionin Romania from 1905 to 2001
Claudiu Herteliu , Bogdan Vasile Ileanu , Marcel Ausloos , , , Giulia Rotundo The Bucharest University of Economic Studies, Bucharest, Romania School of Management, University of Leicester,University Road, Leicester LE1 7RH, UK GRAPES, r. Belle Jardini`ere, 483,B-4031 Li`ege, Wallonia-Brussels Federation Royal Netherlands Academy of Arts and SciencesJoan Muyskenweg 25, 1096 CJ Amsterdam, The Netherlands Sapienza University of Rome, Faculty of Economics,Department of Methods and models for Economics, Territory and Finance,via del Castro Laurenziano 9, I-00161 Roma, ItaliaOctober 26, 2015
Abstract
Population growth (or decay) in a country can be due to various fsocio-economic constraints, as demonstrated in this paper. For exam-ple, sexual intercourse is banned in various religions, during Nativityand Lent fasting periods. Data consisting of registered daily birthrecords for very long (35,429 points) time series and many (24,947,061) a r X i v : . [ phy s i c s . s o c - ph ] O c t abies in Romania between 1905 and 2001 (97 years) is analyzed. Thedata was obtained from the 1992 and 2002 censuses, thus on personsalive at that time. We grouped the population into two categories(Eastern Orthodox and Non-Orthodox) in order to distinguish reli-gious constraints and performed extensive data analysis in a compar-ative manner for both groups. From such a long time series dataanalysis, it seems that the Lent fast has a more drastic effect than theNativity fast over baby conception within the Eastern Orthodox popu-lation, thereby differently increasing the population ratio. Thereafter,we developed and tested econometric models where the dependentvariable is the baby conception deduced day, while the independentvariables are: (i) religious affiliation; (ii) Nativity and Lent fast timeintervals; (iii) rurality; (iv) day length; (v) weekend, and (vi) a trendbackground. Our findings are concordant with other papers, provingdifferences between religious groups on conception, - although reach-ing different conclusions regarding the influence of weather on fertil-ity. The approach seems a useful hint for developing econometric-likemodels in other sociophysics prone cases. Babies are born and therefore conceived non-uniformly over the year. Theworld-wide trend is affected by seasonality (Quetelet, 1826; Lam and Miron,1991; Cancho-Candela et al., 2007). The determinant factors producing theseseasonal effects can be grouped (Friger et al., 2009) into so called naturalcharacteristics (latitude, weather conditions, day-length (DL)) and artificialones (demographic, economic, socio-cultural, including religion).The seasonal effects of natural factors have been extensively studied. Thetarget populations are spread all around the world (Lam and Miron, 1994;Martinez-Bakker et al., 2014; Friger et al., 2009; Dor´elien, 2013; Hubert,2014), but very rarely were they focused on Romania (except for Huberand Fieder, 2011). The findings on climate effects vary across time andgeographical location. They have been extensively analyzed, the importanceattributed to them ranging from high (Villerme, 1831) to as unimportant ine.g. some French population (Regnier-Loilier and Rohrbasser, 2011) or seenas acting jointly with economic development (Huber and Fieder, 2011; Seiver,1985) due to the reasoning that poor populations are negatively affected bylow temperatures during winter or, on the contrary, by hot summers without2ir conditioning. Other researchers (Macfarlane, 1974; Lam and Miron, 1994)stated that weather affects the seasonality of births as long as there are majortemperature differences over the year. Other natural factors like day-length(Bronson, 2004; Rojansky et al., 1992) or moon phases (Criss and Marcum,1981; Jongbloet, 1983) were also tested for their possible connection to birthseasonality.From the artificial factors which were recorded as having some influenceover the seasonality of births, let us mention the demographic ones: age,marital status, education, intergenerational effects (Bobak and Gjonca, 2001;Huber and Fieder, 2011) and economic ones: income, rurality (Kestenbaum,1987; Vitzthum et al., 2009), as well as socio-cultural characteristics: religion,ethnicity (Friger et al., 2009; Hubert, 2014). It turns out occurs that researchon the influence of the artificial factors is rare, perhaps due to some lack ofprecise data. In this context, our paper aims to answer the following questionconcerning the influence of religiosity on the seasonality of dates of conceptionand of birth. More precisely, we raise questions as: did the interdiction ofsexual intercourse during fast (Lent and Nativity) periods have any effect onthe date of conception in twentieth century Romania? Has that socio-cultualconstraint an effect on population growth?
Romania is a country with a large majority (more than 86% nowadays) ofEastern Orthodox population as can be observed in Fig. 1, which is based oncensuses performed across Romanian territories in the last 150 years. Otherreligious denominations have been present in different proportions during thelast 150 years in regions that are now part of Romania. For example, it isapparent that Transylvania is the most heterogeneous Romanian territorywith regard to religious diversity (Herteliu, 2013). However, this effect isapt to be minor, and our current analysis is conducted at the national level,without distinguishing regions.It is known that the operation of complex mechanisms acting on or re-lated to religion is not an easy task to investigate (Herteliu, 2007; Herteliuand Isaic-Maniu, 2009; Ileanu et al., 2012). In this context, to get some sta-tistical balance given the vast Eastern Orthodox majority, the non- Orthodoxsegments are grouped together, thus transforming the religious affiliation intoa simple yes or no variable. Even if some other religions grouped into the3igure 1: Religious affiliation in Romania Source of data: Censuses per-formed across Romanian territories in the last 150 years. Please note thatin communist era four censuses were performed (1948, 1956, 1966 and 1977).Due to political bias, none of them included the questionnaire items regardingreligious affiliation. Note: The data on Romanian territories are organizedor grouped differently from today. The codification is the following: T Tran-sylvania; P Romanian Principalities (Moldavia and Wallachia); GR GreaterRomania 4ategory non- Orthodox have periods of fasting during the year, due to theirparticularities - such as the use of a different calendar, different dates forEaster, different rules for fast periods, or other features specific to variousreligious sects or denominations, their exclusion from the well defined EasternOrthodox group was preferred for the sake of simplicity.It is rather well-known that in the Eastern Orthodox tradition, in ad-dition to abstaining from particular types of food and drinks during fast-ing periods, individuals should also avoid sexual intercourse. Support forthese extensive fasting interdictions specific to the Eastern Orthodox tradi-tion ( http : //orthodoxinf o.com/praxis/f asting s ex.aspx retrieved in March2015) are based on interpretations of certain quotes from the Bible (such asRomans 8:12-14 or Galatians 5:16- 17). One of our goals is to find whetherthese constraints have been implemented. The analysis is based on a very long (35,429 points = 365 (days/year) ? 97years + 24 leap years) daily time series representing all births (24, 947,061)of persons alive at the 1992 or 2002 censuses (similar to Kestenbaum, 1987).Other long data series have been studied before (Cancho-Candela et al., 2007;Ausloos et al., 2015; Rotundo et al., 2015) using different methodologies. Inthe present case, data points were recorded for 01/01/1905 to 31/12/1991(31,776 points) from the 1992 census, while the other 3,653 points (corre-sponding to the period 01/01/1992 to 31/12/2001) are from the 2002 census.The data source is the Romanian National Institute of Statistics (NIS) viaa query tool available within NIS’s intranet ( http : //happy : 81 /P HC http : (retrieved on December2013)). The dates for Easter in the Eastern Orthodox religion during the last110 years is obtained from http : (retrieved on December2013). 5 .2 Variables (a) Variables from censuses : Data was based on persons’ self-declared birth-day. In addition, we use data about: (i) religion: distinguishing Eastern Or-thodox (EOx) and Non-Orthodox (NOx) affiliation. We must assume thatthe religious affiliation of the parents is well defined, based on the informa-tion given by their offspring at census time, and is the same for both par-ents. Even so, the available statistical information showed that the intensityof changing religious affiliation across the Romanian population is very low(Ileanu et al., 2012); (ii) rurality: the percentage of people located in ruralor non rural areas. This introduces a double filter: Eastern Orthodox (rural-ity EOx: rEOx) and Non-Orthodox (rurality NOx: rNOx), the alternativebeing Eastern Orthodox in Urban communities (EOxU) and Non-Orthodoxin Urban communities (NOxU).After subtracting 280 days from the registered birthday, an estimate of theconception day was obtained (Seiver, 1985; Lam et al., 1996). However, date-of-birth evidence can only point to to sexual intercourse leading to successfulpregnancies. It is known that Lent is 48 days long while the Nativity fastlasts 40 days. The uncertain conception date, since not every child is bornexactly 280 days after conception (Pana et al., 2015), overlaps the 40 or48 day fast. We assume that the birth rate is correlated with the rate ofintercourse. Practically, family planning and contraception methods havenot been very common in Romania (it was certainly not encouraged by thecommunist regime between 1966 and 1989) (Chelbegean, 2010). We assumethat the birth rate is correlated with the rate of intercourse. Recall thatthis a posteriori estimate of sexual intercourse resulting in a birth a is madeprimarily to correlate to natural variables, i.e., weather conditions and daylight (or night) duration at the time of the sexual activity.For each year, the fraction of births y i on given day, i , were computedseparately for the orthodox y EOxi , and non-orthodox populations y NOxi , using: y EOxi = n EOxi (cid:80) n EOxi (1)and y NOxi = n NOxi (cid:80) n NOxi (2)where n EOxi and n NOxi is the number of births on the specific day i for EOxand NOx respectively, while (cid:80) n EOxi and (cid:80) n NOxi is the total number of6irths during that year for the respective populations.The dependent variables in the regression models are calculated by divid-ing the daily fraction of births (1) and (2) by the daily expected (uniform)distribution (UD) across the year which is y UD = 1 /
365 for non-leap yearsand y UD = 1 /
366 for leap years. Therefore, we define the following twodependent variables: d EOxi = y EOxi y UD (3)and d NOxi = y NOxi y UD , (4)respectively and using the appropriate y UD .(b) Nativity and Lent fast variables: We used a countdown variable (indays) to Christmas or Easter for each day of Nativity or Lent fast, insteadof a simple binary variable indicating whether or not conception occurredin the fasting period. We preferred this because it is possible that religiousconstraints were observed more scrupulously the closer one was to eitherEaster or Christmas. This countdown variable was computed having theconception date as the reference point. Subsequently, Days Before Christmas(DBC) was computed from: DBC = 0 or t (5)if the conception occurs within or outside Nativity fast time interval; t rep-resents the number of days to Christmas.Similarly, Days Before Easter (DBE) was defined but where t representsthe number of days to Easter: DBE = 0 or t (6)For completeness, we note that Romania adopted the Gregorian calendarin 1919. Therefore, the 31st of March 1919 was followed by the 14th ofApril 1919. Moreover, the Romanian Eastern Orthodox Church is using anupdated form of the Julian calendar. Thus, Christmas is celebrated in Ro-mania on the 25th of December. However, Easter for the Romanian EasternOrthodox Church is usually not synchronized to that of the Roman CatholicChurch. Even so, it is emphasized that these specific calendar change details7o not interfere with the results of our research, since the data regarding reg-istered birthdays in the 1992 and 2002 censuses were based on the Gregoriancalendar.(c) Time interval related variables: Some binary variables were insertedin order to mark some particular time windows during which the concep-tion date might have occurred and influenced by global events or days of theweek. These are: (i) First World War (1stWW) taking the value 1 for con-ception days between 28 July 1914 and 11 November 1918 and 0 outside thistimespan; (ii) Second World War (2ndWW) having a 1 value for conceptiondays between 1 September 1939 and 2 September 1945 and 0 outside thistimespan; (iii) weekend’s days having a value = 1 if the conception day ison a Saturday or Sunday and 0 for Monday, Tuesday, Wednesday, Thursdayand Friday and (iv) squared trend ( t for t =1 to 35,429).(d) Natural variables: The effect of some natural variables is also ana-lyzed through: (i) day length (DL) measured as the number of minutes ofdaylight during each 24 hours of a day. This information refers to Bucharestas the capital city of Romania, but is taken as to be the same for wher-ever place in Romania; (ii) temperature in Celsius degrees (mainly mea-sured for Bucharest) (Due to the usual intrinsic error bars on meteorologicalmeasurements (and variation across the country and throughout the day),using Bucharest temperatures is just a rough approximation to local temper-ature.) , and (iii) proportion of the moon disc being illuminated (having alsoBucharest as the reference point), within the same approximation. The analysis began in qualitatively to determine whether births (or concep-tion) vary throughout the year. Graphs were drawn with various softwares(Kaleidagraph, EViews, Excel, Power Point). Subsequently, the daily av-erage conception within and outside fasting periods was computed. Priorto the econometric modeling, we checked a possible periodicity via a FastFourier Transformation performed with SPSS for 2 cases (the last 32,768days).The final step in the data analysis consists in the design of several econo-metric models. Breaking points in the trend of the dependent variables wereidentified with Cusum SQ Test performed with EViews. Regressions wereperformed with IBM-SPSS (Ordinary Least Square - OLS estimations), - thesame model being run for each phase/cycle (identified after a Cusum SQ8ndicator EOx NOxN Valid 366 366Missing 0 0Mean 59219.15 9082.180Median 59418.00 9050.0000Std.Deviation 8906.85 1049.23Skewness 0.93 2.21Kurtosis 13.52 37.59Minimum 11291 1904Maximum 131311 20016Sum 21629796 3317625Percentiles 20 53554.80 8452.840 57996.20 8851.860 60449.60 9209.280 65141.40 9.618.6Table 1: Descriptive statistics for daily aggregated data of population growthTest). At this stage, factors suspected of being collinear were removed (e.g.day length was chosen instead of temperature) from the regression models.The same procedure was performed for factors with low statistical signifi-cance (e.g. moon phases). As shown by Huber and Fieder (2011) for women born between 1920 and1955 and older than 45 years (completed fertility), Romanian births registera non-uniform distribution across the year with a maximum in June andminima in December and January, as obtained from monthly data.Our raw data contains the number of births on each day of the year (EOxand NOx) for a long period (97 years). In order to present this extensive datain a concise way, a pivot table was designed separately containing monthlyand daily aggregated values for EOx and NOx. For both populations, themean and median are very close to each other (less than 0.4%) while the stan-9ard deviations for the means are quite low (15% of the mean for EOx and11.6% for NOx). The statistics of both populations are slightly asymmetricand leptokurtic. Other descriptive statistical characteristics are available inTable 1. Next, the aggregated date were sorted into quintiles (Dedu et al.,2014; Giuclea and Popescu, 2009) obtained by quintiles. The following inter-vals were obtained for EOx (Fig. 2): (q1) days having under 53,554.8 birthslabeled in red; (q2) days having between 53,554.8 and 57,996.2 births labeledin orange; (q3) days having between 57,996.2 and 60,449.6 births labeled inyellow; (q4) days having between 60,449.6 and 65,141.4 births labeled in lightgreen and (q5) days having over 65,141.4 births labeled in dark green. Sim-ilarly for NOx (Fig. 3), the quintile intervals are (q1) days having under8,452.8 births (red); (q2) days having between 8,452.8 and 8,851.8 births(orange); (q3) days having between 8,851.8 and 9,209.2 births (yellow); (q4)days having between 9,209.2 and 9,618.6 births (light green) and (q5) dayshaving over 9 618.6 births (dark green).This sorting by quintiles of almost one century’s data can (by avoidingcompensation through averaging over unequal size month) provide some sta-tistically significant differences among daily distributions across the year. Itcan easily be seen that this compensation does not presently occur is ourcase. Therefore, it may be deduced that the outlined monthly or daily dif-ferentiation does not lead to a random outcome.Referring to Fig 2, for the EOx population, there are important differ-ences between November, December and January and the rest of the year,and also between the first two thirds of a month compared to the last third.These monthly differences were tested by a χ test and are statistically sig-nificant (computed χ was 131,285.4 being significant with a p value lowerthan 0.001). Again referring to Fig. 2, the lowest aggregate number of birthsfor a specific day was, naturally, on 29th of February (11,291 births) whilethe maximum was on the 1st of January (131,311 births). We will commenton the large number of births on 1 Jan below.In the case of the NOx population, (See Fig. 3) the lower birth figuresin November and December are consistent with lower number of births inthose months in the EOx population, while January is not so distinct as inthe NOx case. The distribution of the daily figures outside the November-January time span seems to be more equally balanced for NOx, - no persistentpattern is visible, in contrast to the EOx case. Similarly in the EOx and theNOx population, the minimum aggregate number of births was registered on29th of February (1,904 births) while the higher one occurs on 1st of January10able 2: The logarithm of the ratio of conceived orthodox babies relativeto assumed uniform daily value for 1 year. The regression coefficients shownare the standardized ones (prior to estimation all independent variables arestandardized). Factors (independent variables) are: Days Before Christmas(DBC), Days Before Easter (DBE), Day Length in minutes (DL), Ruralityfor ORTHD, squared trend ( t for t = 1 to 35,429), Binary variables: (i)FirstWorldWar (1stWW)taking the value 1 for conception days between 28July 1914 and 11 November 1918 and 0 outside this timespan; (ii) SecondWorldWar (2ndWW) having a 1 value for conception days between 1 Septem-ber 1939 and 2 September 1945 and 0 outside this timespan; (iii) weekenddays having a value 1 if the conception day is on a Saturday or Sunday and 0for Monday, Tuesday,Wednesday, Thursday and Friday. Bold values denotethe models goodness of fit, adjusted by number of independent variables. ∗ Statistically significant at level 10%; ∗∗ Statistically significant at level 5%; ∗∗∗
Statistically significant at level 1%.11able 3: The logarithm of the ratio of conceived Non-Orthodox babiesrelative to assumed uniform daily value for 1 year. The regression coeffi-cients shown are the standardized ones (prior to estimation all independentvariables are standardized). Factors (independent variables) are: Days Be-fore Christmas (DBC), Days Before Easter (DBE), Day Length in minutes(DL), Rurality forNORTHD, squared trend ( t for t = 1 to 35,429), Binaryvariables: (i) FirstWorldWar (1stWW)taking the value 1 for conception daysbetween 28 July 1914 and 11 November 1918 and 0 outside this timespan; (ii)Second WorldWar (2ndWW) having a 1 value for conception days between1 September 1939 and 2 September 1945 and 0 outside this timespan; (iii)weekend days having a value = 1 if the conception day is on a Saturday orSunday and 0 for Monday, Tuesday,Wednesday, Thursday and Friday. Boldvalues denote the models goodness of fit, adjusted by number of independentvariables. **Statistically significant at level 5%; *** Statistically significantat level 1%. 12igure 2: Quintile distribution of daily aggregated births for the EasternOrthodox (EOx) population during 1905-2001.13igure 3: Quintile distribution of daily aggregated births for the Non-Orthodox (NOx) population during 1905-2001.1420 016 births). For this group, the computed χ was 9,184.2 showing thatthe differences are statistically significant with a p value lower than 0.001.For both (EOx and NOx) populations, Figs 2- 3 show a noticeably highnumber of births (almost dark green) on the 1st day of any month and a lownumber of births) on the last day of each month.As noted above for both populations, there is an anomalously high num-ber of births on Jan. 1st. In fact there are twice as many births on Jan 1stthan on Jan 2nd and almost 4 times as many births as on the 31st of Decem-ber. We call this anomaly the 1st of January effect; further work should bedone on this subject. A similar effect was noted by MacFarlane (1974). Inbrief, possible explanations for this effect might be: (i) parents’ psychologi-cal comforting thought that their child is considered to be one year youngerif registered on January 1st instead of December 31st (a quite comfortablesituation for both girls and boys since there would be a delay of one year tothe compulsory military training, - effective in Romania prior to 2007); (ii)municipal recorders not working during winter holidays therefore increasingthe likelihood of an incorrect record (despite the fact that children can beregistered on the 5th of January mentioning the correct birth date); (iii)on the 1st of January there is a very important holiday for the EOx (SaintGreater Basil), evidenced by many children born around this day namedBasil. (According to http : − − romani − isi − serbeaza − ziua − onomasticade − sf antul − vasile − .html ,retrieved on December 2014, there are almost 600 000 people in Romanianamed Basil or the female version); and (iv) mistaken registration. (Evenif it is not statistically important, one co-author of the current paper hasa brother registered as born on January 1st, even if the event certainly oc-curred, on 2nd January 1:00 AM).Since the variation of the number of births across the year is undoubtable,we now address the primary concern of this research: did major fast periods(i.e. before Christmas and Easter) affect conception or not ? First, notethat several religions (Roman Catholic, Greek Catholic, Old Rite EasternOrthodox) grouped within NOx category still have different Lent and Nativ-ity fast periods. Moreover, the fast periods (especially for Lent) are usuallynot similar to EOx. Therefore, a visual inspection of this variability (Figs. 4- 5) was conducted, only for EOx. Prior to this, it seems useful to computefor each year the daily average conceptions within such fast (or out of fast)periods.Despite the long time series which certainly must depend on many local15igure 4: Daily average number of conceptions during Lent fast/out of fastperiod for Eastern Orthodox (EOx)or general factors, Fig. 4 shows that the daily mean number of conceivedchildren by orthodox is significantly lower during Lent than during the restof the year. In the whole range of almost 100 years only two exceptions tothis rule can be seen: (i) during the First World War (more exactly, in 1916)and (ii) during 1989, a year which had a strong emotional impact on theRomanian population, when there was a revolution and a major change ofpolitical system. Even if births increased because of Ceausescu’s decree, thenumber of conceptions out of fast is still larger than the number of concep-tions during fasting periods. Thus, even if in the short term (1967-1968)the Ceausescu Decree reduced the gap between ”out of fast” and ”duringthe fast” on the long term, the decree had not enough force to completelyeliminate the effect of ecclesiastical admonitions.Similar plots of conception within and without the Nativity fast (Fig. 5)show dramatically different behavior. In this case the rule of fewer concep-tions during fast periods was broken in 67 from all 97 available years. Tocomplete the basic statistical analysis, an enhanced econometric analysis isdescribed and performed in the Econometric models sub-section, below.16igure 5: Daily average number of conceptions during Nativity fast/out offast period for Eastern Orthodox (EOx) The analytical approach of our paper is an econometrical one. However,before applying econometric model, it is useful to test whether the maindependent variable of interest EOx is affected by some periodicity (figure6) in order to avoid such a specific trend. A spectral analysis is therebyperformed, using SPSS (Tukey-Hamming method Brillinger, 2002). Such aspectral analysis may be done when the number of data points is a power of 2.Therefore, in order to have the maximum set of points, 2 (= 32,768 cases),the first 2 661 data points were deleted, whence shortening our dataset tothe 14/4/1912 - 31/12/2001 interval.The major peak (we use a log scale in order to slightly enhance the x-axis) is about half a year (183 days or a little bit more than 26 weeks). Thenext largest peaks occur ca. 372 days (ca. 53 weeks), four months (120 days)and one month (30 days); such ”periods” are quite similar to those found in(Cancho-Candela et al., 2007). 17 .3 Econometric modelling As presented in the above paragraphs, the number of conceptions in a pe-riod of time was assumed to be determined by a set of factors, more or lessknown and more or less understood. In the current analysis, classical multi-ple regression models are considered, with various components and variablesmeasured on a daily time scale. Several versions of the models were testedstarting from the general linearized multifactorial model (Andrei and Bour-bonnais, 2008). All factors mentioned in the data and methods section abovewere initially included in the models. Based on the goodness of fit of these,taking into account the statistical significance of the regression coefficients(with a Student t-test), an optimum, model was obtained and it will only bethe only model outlined below. As in other research (e.g., Lam and Miron,1996), the best results are obtained when considering the logarithmic variableas a dependent variable.(A) Model for the orthodox population; components and variables:In such a case, the dependent variable, d EOx and covariates are: Days be-fore Christmas (DBC), Days before Easter (DBE), rurality EOx (rEOx), daylength (DL), World wars, through 1 stW W and 2 ndW W variables, weekends(WE), trends (long term tendencies, a stable component, core of the timeseries) and a stochastic component ( (cid:15) ). The variables are described in a sub-section of the Material and methods section above. A formalized version ofthe model can be easily written; after applying a logarithmic transformation,the model reads in obvious notations, ln ( d EOx ) is equal to= α + α DBC + α DBE + α ln ( DL )+ α rEOx + α stW W + α ndW W + α W E + α ( trend ) + (cid:15) (7)(B) Model for the non-orthodox population; components and variables:with almost the same explanatory variables but instead having the depen-dent variable, d NOx , i.e. the normalized number of non-orthodox conceptions.Moreover, instead of covariates share of orthodox in the rural area, the shareof non-orthodox in the rural area is introduced (rNOx), which may be moresuggestive. Thus, in obvious notations, ln ( d NOx ) is equal to= β + β DBC + β DBE + β ln ( DL )+ β rN Ox + β stW W + β ndW W + β W E + β ( trend ) + (cid:15) (8)18igure 6: Fast Fourier Transformation (FFT) for the d EOx . This dependentvariable in the regression models is calculated by dividing the daily fractionof births by the daily expected (uniform) distribution across the year.19igure 7: Breaking points identification using CUSUM SQ test(C) Structural break identification:Since the series covers a very long time interval (97 years=35,429 days),parameters might vary somewhat. Perhaps as a result the overall model (M0,presented in Table 1) appears not to be very efficient in that some variablessuch as the Christmas fast seem to be insignificant. As in (Cancho-Candelaet al., 2007), a segmentation of the long time span into shorter sub-periodsseems an interesting procedural test.In order to do so, the CUSUM SQ (Young, 2011) test was applied to theoverall estimated model (M0). The CUSUM SQ graphic, presented in Figure7 suggests that a 5% level of significance multiple changes is likely during theanalyzed timespan. It can be observed that the breaking points are ca. 1911,1918; 1947, 1966, 1990. Consequently, there are 6 suggested sub-periods ofstability for the models (Mi): 1. 1905-1910 (M1); 2. 1911- 1918 (M2); 3.1919-1946 (M3); 4. 1947-1966 (M4); 5. 1967-1989 (M5); 6. 1990-2001 (M6).Each break point is in fact well related to a social/ political/historical event:e.g., during the period for 1907-1910, there were peasants’ uprisings; at theend of the second and third sub-period, there were events which determinedthe decrease of the number of births: the First World War and the Second20orld War respectively. Also, this observation suggests how to specify thetrend component of the models. Therefore, the econometric models were re-worked for each of the sub-periods taking into account these remarks: (i) formodel M1 a non-linear (let us take it squared, within usual series expansionapproximations) trend of births during the entire period is introduced; (ii) formodel M2 and M3, similar squared trends can be also assumed.; (iii) the sameassumptions are introduced for models M4 and M5. According to nationalreports, during the period of 1956-1966, the nativity rate decreased dramat-ically due to many factors such as: abortion liberalization, better access towork and education for women; moreover, between 1985 and 1990, the socialcondition of Romanians significantly decreased. The Decree of Ceausescuoutlawing abortion (1st October 1966) had effects only for a short periodof time; (iv) finally, for the 6th model, the trend in the number of babiesconceived and born decreased according to many factors, like abortion liber-alization (992 000 such operations occur in 1990), but also post-communistsocio-economic crisis, migration etc.After applying estimations using SPSS methodology, the optimized re-sults are given in Table 2 and Table 3.After performing a F test - ANOVA it has been found that the modelfor every sub-period for EOx and NOx population are valid and statisticallysignificant. The proportion of variance (R2) explained for the dependentvariable is between 12.7% (M0 and M5) and 32.3% (M3), - values similar toother studies (Friger et al., 2009). Almost all regression parameters are sta-tistically significant (from a Student t-test) with slightly different degrees ofsignificance (most of them having the p value lower than 0.01). As expected,the model performed on NOx population, registered levels of R2 lower thanthe previous ones. The maximum deviation occurs for M6 (10.5%) and theminimum for M2 (2.3%). Regression coefficients in this case fail to be slightlymore often significant.In addition, the sensitivity of the statistical models was tested by takinginto consideration a gestation duration shorter by one and two weeks, respec-tively. The models outcomes in these scenarios are found to be quite similarto the presented ones. 21
Discussion and Conclusions
Romanians following Eastern Orthodox religious beliefs or rules, i.e. the ma-jority of religious persons in Romania, are taught to avoid sexual intercourseand to abstain from particular food and drinks during fasting periods. Theresults of our time series analysis on birth rates show different behaviorsduring the major fasts. Conceptions, during Lent Fast, are consistent withreligious constraints. The negative value of the coefficient (DBE) for Days toEaster during the Fast shows that as the Holiday gets closer, the estimatednumber of conceptions significantly decreases. It is worth mentioning herethat the Orthodox religion grants some exception periods to believers for rea-sons mainly linked to health. In these grace periods, there might be increasedconception. It is important to note that the standard absolute value of thecoefficient of the variable which measures the impact of Lent fast on con-ception (DBE) remains relatively stable in [-0.25; -0.35] for all sub-periods.This points to the continuity and persistence of the Eastern Orthodox faithin general, and particularly regarding interdictions of Lent fasting through-out the almost 100 years of this study. This continuous adhesion to faithis also highlighted by the values estimated in the second model (Table II)for the Non-Orthodox sub sample (NOx) where the estimated coefficient ofthe variable has reasonable values, though half [-0.04;-0.19] that for the EOxpopulation. On the other hand, the Nativity Fast no-sex constraint doesnot seem to have a great significance for Orthodox Romanians nowadays:the coefficient of the variable (DBC) which measures this impact is positiveafter the 1st World War, showing that as the Christmas day gets closer, thenumber of conceptions increases significantly. If in the case of the Lent Fastthe coefficient remains stable in time, in the case of the Nativity Fast, thestandard value is increasing in time, showing that when approaching the Na-tivity date, the population practices a more liberal lifestyle, obeying fewerconstraints during the fasting period. For example, in the period 1990-2001the value of the coefficient is approximately three times higher than the oneestimated for the period before the Second World War. This can be likelyexplained from a rural habit: weddings are traditionally performed in ru-ral Romania in the last part of autumn, after the harvest (Trebici, 1979).Moreover, the fermentation process of the new wine is stopped, thus render-ing the liquid more drinkable, in particular at weddings. It has been shown(Trebici, 1979) that usually about 9-10 months later, the first baby of a newcouple is delivered. Another point is based on the psychological idea that22he Christmas holiday is a more joyful event compared to Easter (in agree-ment with others, like Seive (1985)). We can also propose as a cause that,in rural Romania, there are fewer household duties (see Lam and Miron,1991; tangential Becker, 1991) occurring outside the house during the Nativ-ity fast than during spring Lent. And lastly, the day length (and of coursenight length) needs to be mentioned as there was a lack of electricity untilthe middle of the 20th century and also a lack of other indoor sources ofentertainment, thus increasing the likelihood of sexual encounters at thosetimes.Therefore, the proposed models highlight the impact of religiously con-trolled fasting periods on conception, besides the existence of other factorswith persistent or temporary influence. For instance, the day length variablereveals the significant influence of several factors such as: the regularity ofdaily activities determined by the season and the temperature (these twofactors were included in the initial models and eliminated due to the multi-colinearity effect). With one exception (M1), our analysis reveals a photope-riodicity related cause for baby conception as found for European Countrieslike France, England, Sweden (Lam and Miron, 1994) on monthly data. Ofcourse, due to multicolinearity, the photoperiodicity could have only an ap-parent effect as a proxy for factors like the temperature, outside householdduties or the lack of available inside sources of entertainment.Weekend is a factor inserted into our models that tends to function asproxy for holidays, and more permissive sexual activity. Previous research(Bobak and Gjonca, 2001) suggests that during holidays, the frequency ofconception seems to be higher indeed. Our data analysis generally agreeswith these findings. Therefore, for EOx population, weekends seem to havea positive impact on conception with the exception of M4. Furthermore, thesame influence is reflected on NOx population also until 1966 (models M1 toM4). Thereafter, weekends are shown to have a negative influence.It also needs to be stated that despite other researchers’ findings (Jong-bloet, 1983), the proportion of illuminated moon disc was found to be in-significant from a statistical view point (along a Student t-test).Furthermore, the values of the regression coefficients specific to the Or-thodox persons in the rural environment, illustrate that behavioral patternsin these geographical areas are much more prone to seasonal elements inchild conception as compared to urban areas, where these differentiationsare diminished.The trend of the series for each analyzed sub-period confirms the expec-23ations stated above. Thus, in each major timespan here analyzed, there areyears of maximum increases followed by a descending evolution. The onlyexception to this rule are the years between 1948-1966, when the maximumpoints of the number of conceived babies are at the ends of the interval. Onthe one hand, this could be determined by the necessity to cover the disastercaused by the war; on the other hand, it shows the significant impact ofCeausescu’s decree.Finally, the splitting of the populations into EOx and NOx categories isproven relevant. The model for all sub-periods analyzed is much weaker (fromeconometrics viewpoint) for the NOx population which does not have Nativ-ity or Lent fasts, or if it does, with time delays versus the EOx population.Moreover, the standardized coefficient that measures the impact of Lent Faston conception has in general the largest absolute value. It is followed in ab-solute value by the rurality (rEOx) coefficient, both values strengthening thevalue of the orthodox tradition and faith, factors which are underestimatedor missing in previous human conception researches.In brief, the primary result of the current investigation is that the non-uniformity of the conception of babies, in Romania, for the last century,across the year is affected not merely by seasonality, but by a double filter:season and religion.
Acknowledgements
During the long gestation period for this paper several relatives, friends,colleagues of the authors provided valuable feed-back, suggestions, comments:Robert Ancuceanu, Tudorel Andrei, Radu Chirvasuta, Gurjeet Dhesi, Vic-tor Dragota, Anca Herteliu, Roxana Herteliu-Iftode, Alexandru Isaic-Maniu,Ionel Jianu, Michael S. Jones, Adrian Pana, Gheorghe Peltecu. And lastbut not least Dr. David Berman (Iowa University) reviewed the paper forits grammar, vocabulary and style content. He has made also many veryinteresting suggestions improving the scientific content.This paper is part of scientific activities in COST Action TD1210 ’An-alyzing the dynamics of information and knowledge landscapes’. This workby CH was co-financed by the European Social Fund through the SectorialOperational Programme Human Resources Development 2007-2013, projectnumber POSDRU/1.5/S/59184 Performance and excellence in postdoctoralresearch in Romanian economics science domain24 eferenceseferences