Ambient temperature variation does not influence regional proportion of human male births in New Zealand

Abstract

Fluctuations in the mean annual ambient temperature have been associated with temporal changes in the proportion of males in total human births. Our aim was to test if changes in mean annual ambient temperature have influenced regional variation in the proportion of males born in New Zealand from 1961 to 2009. We used time-series analyses to test for positive relationships between current or lagged mean annual ambient temperature and the proportion of male births within 13 regions. Only two regions showed a significant effect of either current year or lagged temperature on the proportion of males born, with each showing a decreasing male-bias with temperature (contrary to our hypothesis). Additionally, a meta-analysis revealed that the standardized effect size of temperature on the proportion of male births across regions was negative for both current and previous years, but was not statistically significant. These findings did not support the hypothesis that mean annual ambient temperature affects the proportion of males born.

Keywords:

• meta-analysis
• sex allocation
• sex ratio
• time series analysis

Introduction

Natural selection predicts a 1:1 sex ratio (Fisher 1930). However, many mammals do not conform to this pattern, prompting much theoretical speculation into the mechanisms underpinning this anomaly. Trivers and Willard (1973) proposed that natural selection may have favoured mechanisms in females that select in utero for the offspring that will be most reproductively successful in given environmental circumstances. Thus, skews in the proportion of male births, or alternatively analysed as the secondary sex ratio (SSR: the number of male births divided by the number of female births), should be predictable based on biological, sociological and abiotic events.

The global human SSR is consistently male-biased and is currently estimated to be 1.07 (Central Intelligence Agency 2010). It has been proposed that political unrest, natural disasters and maternal stress are among a long list of traits that lower the male bias in SSRs (James 2010; Navara 2010), whereas periods of war favour a more male-biased SSR (James 2009). Climatic differences across populations, particularly fluctuations in ambient temperature, may also affect the human SSR (McLachlan & Storey 2003). Although human births are globally male biased, the degree of male bias has been shown to vary across latitudinal gradients and climates, so that in countries with tropical climates the SSR is less male biased than in countries with temperate or subarctic climates (Navara 2009). Changes in ambient temperature may cause physiological stress and therefore could play a determining role in sex allocation in humans (Catalano et al. 2008). Thus, it is suggested that fewer males will be born during stressful colder periods, as a weaker male might not survive and reproduce (Catalano et al. 2008; Helle et al. 2008). Studies conducted on a smaller spatial scale using time-series analyses to test for temporal variation in the SSR have found that more males are born in warmer years (Catalano et al. 2008; Helle et al. 2008, 2009), suggesting that within countries mean annual ambient temperature has a significant effect on the SSR.

Previous studies using time-series analyses to test for effects of ambient temperature on the proportion of male births have used national data sets, including an analysis demonstrating no influence of ambient temperature on the proportion of male births for the New Zealand population as a whole (Dixson et al. 2011). However, that analysis used a national composite measure of ambient temperature, which may have hidden or masked the localized differences in temperatures that could potentially influence the proportion of males born at a regional level within New Zealand. Indeed, New Zealand consists of two long islands, on which the ambient temperature varies considerably. The warmest region is Northland, with an overall mean ambient temperature of 15.15 °C, while Southland is the coldest region with an overall mean ambient temperature of 9.87 °C (Figs 1, 2). This 5.28 °C difference is of a similar magnitude to that observed to influence the proportion of male births in Europe (Catalano et al. 2008; Helle et al. 2008, 2009). Here, we analyse a dataset for which we have birth statistics for specific regions of New Zealand. Thus, the aim of this study was to investigate the influence of regional variation in mean annual ambient temperature on the proportion of male births from 1961 to 2009 in New Zealand. We tested for effects of ambient temperature in the same year of birth, or in the previous year, as the time around conception is considered to influence the proportion of live male births. Based on previous published research, we hypothesized that increases in mean annual ambient temperature will be associated with more male births, whereas colder mean temperatures will be associated with fewer male births (Catalano et al. 2008; Helle et al. 2008, 2009).

Figure 1  Map showing the location of different regions in New Zealand used for the analysis. Note that these regional boundaries are those defined by Statistics New Zealand in 1961. n _t =total number of births in dataset; n _y =average number of births per year.

Figure 2  The relationship across regions between the mean annual ambient temperature and average proportion of male births (both averaged over 1961–2009). Error bars are 95% confidence intervals.

Methods

In 1961 Statistics New Zealand established 13 statistical areas across New Zealand (Fig. 1) to generate accurate local demographic data. Data on the numbers of male and female live births in these regions were provided by Statistics New Zealand from vital statistics in annual reports. We extracted the numbers of males and females born annually in each region during the period 1961–2009, from which we calculated the annual proportion of male births (the number of male births divided by the total number of male and female births). To obtain data on mean annual ambient temperature in °C for the 13 statistical areas, we used data collected by the National Institute of Water and Atmospheric Research (NIWA) at multiple weather stations within each region. We calculated the average temperature across the weather stations within each region for each year from 1961 to 2009.

For each region we used transfer function (ARIMA) models (Box et al. 2008) to estimate the dynamic effects of temperature on the proportion of male births. In addition, in order to identify any unusual outlying observations or structural changes that are best modelled explicitly as functions of time, we used an intervention analysis approach (Box & Tiao 1975; Tsay 1986). Models were tested for residual autocorrelation using the Ljung-Box portmanteau statistic (Ljung & Box 1978). Computations used SPSS (SPSS 2009).

Our first analyses were designed to test for dynamic effects of temperature on the proportion of male births within each statistical region. Khashan et al. (2009) raised the issue that in some research on effects of stress on the human proportion of male births where multiple comparisons were made, one or two statistically significant results could be due to chance alone. One approach to this problem would be to use a (sequential) Bonferroni correction. However, Nakagawa (2004) has criticized the use of Bonferroni corrections, since they increase the risk of Type II errors and hence reduce power. Instead a meta-analytic approach is recommended, and consequently we conducted a meta-analysis to estimate the average standardized effect size of temperature on the proportion of male births across regions. Standardized effects were estimated by dividing the estimated coefficients by their standard errors to take account of differences in levels of inherent variability in the proportion of male births between regions, principally due to differences in numbers of births per year. If the hypothesis that increased temperature is associated with more male births is supported, the average standardized effect of temperature should be positive. In this meta-analysis, a P value<0.05 would indicate that there was a significantly more positive or negative average effect than would be expected by chance.

Results

Despite a difference in mean annual temperature of up to 5.28 °C between regions within New Zealand, the mean proportion of male births across all 13 regions showed no relationship with mean annual ambient temperature during 1961–2009 (Fig. 2).

Temperatures within regions also varied between years. For example, in Westland the mean annual temperature varied from 9.42 °C in 1969 to 12.54 °C in 1989. Time series analyses revealed that over the period of 1961–2009, in 11 of the 13 regions, changes in the annual proportion of male births were unrelated to changes in the current or previous years' mean annual ambient temperature (Table 1). All selected models required no ARIMA parameters, since there was no residual autocorrelation (Ljung-Box P≥0.162). In the Nelson region, there was a statistically significant negative effect of both current year (n=49, ß=−0.011, SE=0.004, P=0.008) and previous year (n=49, ß=−0.009, SE=0.004, P=0.046) temperature on the proportion of male births. This significant negative effect indicated that as mean annual ambient temperatures increased, significantly fewer males were born. Likewise, in Southland there was a statistically significant negative effect of temperature on the proportion of male births for the current year (n=49, ß=−0.014, SE=0.004, P=0.002), but not the previous year (n=49, ß=−0.003, SE=0.004, P=0.432). Figure 3 shows the proportion of male births and mean annual ambient temperature in Northland, the warmest region in New Zealand (overall mean 15.15 °C) and Southland, which is the coldest region (overall mean 9.87 °C). While the mean annual ambient temperatures are markedly different in these two regions, the proportions of male births were similar and show similar variation in both regions (Fig. 3).

Figure 3  Time series of the proportion of male births and average annual ambient temperature in two contrasting New Zealand regions. A, The warmest region, Northland (overall mean ambient temperature of 15.15 °C). B, The coldest region, Southland (overall mean ambient temperature of 9.87 °C).

Table 1  Model coefficients from time series analysis using transfer function (ARIMA) models to estimate the dynamic effects of mean annual ambient temperature on the proportion of males born. This analysis tested for a positive relationship between current or lagged mean annual ambient temperature and the proportion of male births. Only two of 13 coefficients were statistically significant (P < 0.05) for the current year temperature, and just one for the first lag of temperature, with those three coefficients all negative n=49 years of data.

Region	Current year temperature			First lag of temperature
	ß	SE(ß)	P	ß	SE(ß)	P
Northland	0.001	0.003	0.856	0.005	0.003	0.113
Central Auckland	0.002	0.001	0.077	0.002	0.001	0.125
South Auckland/ Bay of Plenty	−0.003	0.002	0.053	−1.4×10^−4	0.002	0.927
East Coast	1.1×10^−4	0.004	0.978	−0.004	0.004	0.338
Hawke's Bay	0.002	0.003	0.591	0.004	0.003	0.125
Taranaki	−0.004	0.003	0.194	−0.003	0.003	0.33
Wellington	−0.002	0.002	0.179	0.002	0.002	0.394
Marlborough	0.005	0.008	0.531	−0.008	0.008	0.284
Nelson	−0.011	0.004	0.008	−0.009	0.004	0.046
Westland	−0.008	0.005	0.102	−0.003	0.005	0.533
Canterbury	−0.003	0.002	0.145	0.001	0.002	0.689
Otago	0.003	0.003	0.297	−0.001	0.003	0.633
Southland	−0.014	0.004	0.002	−0.003	0.004	0.432

From our meta-analysis the estimated average standardized effects of current (standardized effect size=−0.739, P=0.111) and previous year (standardized effect size=−0.083, P=0.806) mean annual ambient temperatures were both negative, with neither being statistically significant. These results support the conclusion that there is no consistent positive effect of temperature on the proportion of male births across regions in New Zealand from 1961 to 2009.

Discussion

For 11 of the 13 New Zealand (NZ) geographic regions, from 1961 to 2009, we found no significant relationship between fluctuations in mean annual ambient temperature during the current or previous year and the proportion of males born. Our results therefore did not support the hypothesis that temperature is positively related to changes in the proportion of male births in NZ. Previous studies conducted on a similar spatial scale in Scandinavia, which have also employed time-series analyses to investigate the role of mean annual ambient temperature on the proportion of male births, or the secondary sex ratio (SSR: the number of male births divided by the number of female births), have found a positive relationship (Catalano et al. 2008; Helle et al. 2008, 2009). We were unable to replicate these findings from Northern Europe in NZ. Further, where we did find a relationship, as was the case in the regions of Nelson and Southland, we found a negative relationship, so that more males were born in colder years. If anything, these results contradict the hypothesis that there would be a positive relationship between mean annual ambient temperature and the proportion of male births.

The effect of annual ambient temperature on sex specific mortality may be stronger when looking at the extremes of temperature differences between latitudinal gradients. When Navara (2009) analysed data on the proportion of male births from 202 countries from 1997 to 2006, the proportion of male births was significantly less male-biased at tropical latitudes than in temperate and subarctic latitudes. Other studies conducted in tropical sub-Saharan African countries support this finding (Garenne 2002; Kaba 2008). The extreme cold climate of the arctic affects infant survival and when temperatures are lower in January–July, infant mortality increases significantly (Young & Mäkinen 2010). When comparing across latitudes in Europe, in warmer climates more males were born; however, this trend was not supported when comparing data on the SSR at different latitudes in North American cultures (Grech et al. 2002). Thus, while our results for NZ follow the general global pattern of male-bias in the proportion of male births that has been observed in other temperate climates (Navara 2009), we found no effects of fluctuations in ambient temperature between years on the proportion of males born.

In another study, Catalano et al. (2008) compared a composite measure of SSR from Denmark, Finland, Norway and Sweden with the mean annual ambient temperature measured from two sites in Sweden. Their study found that the sex ratio rose with ambient temperature during the period 1865–1914. We found little temporal structure to the proportion of males born in NZ and hence when testing it against the more structured ambient temperature data, we found no relationship between them. Interestingly, we found that in the Southland region of New Zealand, the coldest part of the country with a mean ambient temperature of 9.87 °C, the opposite pattern was found to that identified by Grech et al. (2002) in Europe and Catalano et al. (2008) in Scandinavia. It is possible that improvements in modern housing conditions, such as insulation and central heating, may have changed the effect that ambient temperature exerted on sex determination in the early 1900s. However, even in modern NZ, household central heating is still uncommon. In a survey conducted in 2004 in which 397 houses were randomly selected from throughout NZ, only 5% had central heating (French et al. 2006). Thus, we think that our failure to find a positive effect of mean annual ambient temperature on the proportion of male births is not an artefact of modern housing in NZ. We conclude that while differences in the SSR have been found when comparing global differences in mean annual ambient temperature elsewhere, these findings do not extend to NZ despite substantial differences in ambient temperature between the northern- and southern-most regions. Results of this current analysis support our previous work wherein we found no seasonal or annual relationship between ambient temperature and the proportion of males born in the country as a whole over the much longer period of 1876–2009 (Dixson et al. 2011).

It is likely that stressful events occurring at the time of conception, other than changes in mean annual ambient temperature, play stronger roles in sex allocation in New Zealand. For example, Catalano et al. (2010) have evidence using time-series analysis that economic stress, such as high numbers of layoffs from employment, is associated with significantly fewer males being born. Another potential mechanism of sex-specific mortality is the change in female somatic condition around the time of conception. In mammals, female condition at the time of conception has been shown to affect sex determination, so that females in fitter condition give birth to more males (Cameron 2004; Cameron & Linklater 2007). In humans, it is also possible that at the time of conception the nutritional status of the mother is an important predictor of the SSR (Rosenfeldt & Roberts 2004). Some studies have suggested that maternal condition and diet influence the SSR in humans (Gibson & Mace 2003; Matthews et al. 2008). Thus, it is likely that several social, economic and medical factors interact to determine the SSR (Catalano et al. 2010; James 2010; Navara 2010). These hypotheses deserve further testing as alternatives to the temperature-dependent selection hypothesis in order to uncover the possible underlying mechanisms driving the SSR in New Zealand.

Acknowledgements

We thank Victoria University of Wellington for funding this work, Statistics New Zealand for access to the birth data and NIWA for the use of the temperature data.