Analysis of Time-Series Variation Patterns of Total Fertility Rate for the Last 20 Years (2000-2020) in 335 Secondary Medical Areas in Japan and its Association with Social Status of Women

Abstract

Japan’s Total Fertility Rate (TFR) is below the population replacement level, and the government is calling on local governments to develop policies to improve the TFR. However, a comparison of TFR values by prefecture shows that the highest value in 2022 is 1.70 in Okinawa Prefecture, while the lowest is 1.04 in Tokyo, indicating notable regional differences.

This study analyzes the relationship between time-series changes in TFR and geographical and socioeconomic background factors for all municipalities in Japan. In particular, we will examine the relationship between TFR and gender indicators in order to clarify the relationship between women’s social advancement and TFR.

In order to utilize demographic data of small municipalities as raw data, we organized information on municipal mergers over the last 20 years (2000-2020) and used a method to aggregate data by secondary medical area. The time-series variation of TFR in each secondary medical area was converted into a comprehensive index by factor analysis, and the two groups were compared by Mann-Whitney U test (MWU-test), dividing them by high and low factor scores. In addition, the geographic distribution of TFR-trend-factors was evaluated by choropleth maps.

Eleven factors were extracted, and two of them, TFR-trend-factor 1 (maintenance of high TFR) and TFR-trend-factor 3 (linear upward trend of TFR), showed positive trends for TFR. MWU-test results showed that both TFR-trend-factor 1 and 3 were significantly associated with gender indicators (economy, education, health, and political participation). This suggests that the higher social status of women in society may increase the TFR.

On the other hand, the strength and direction of the impact on TFR tended to vary depending on the TFR-trend-factor and the social advancement indicator items of each region. Therefore, it is also important to analyze the relationship between the sub-indexes and individual data and TFR.

Introduction

Many developed countries have experienced declining birthrates since around 1965, and the Total Fertility Rate (TFR) of most developed countries has declined to below the population replacement level of 2.07 [1]. In particular, Japan’s TFR declined earlier and more rapidly than those of other countries, and as of 2021 it stood at 1.30 (35^th place among OECD member countries) [2]. The so-called “TFR 1.57 shock of 1989” triggered a number of analyses of the causes of the decline in the TFR. Mainly by national agencies such as the National Institute of Population and Social Security Research, and measures to increase TFR and to improve the child-rearing environment have been studied and implemented [3].

The “Basic Direction of Policies to Support Child-Rearing in the Future” [4] formulated jointly by the Ministry of Education, Science, Sports and Culture; Ministry of Health and Welfare; Ministry of Labor; Ministry of Construction in 1994 cited the rising rate of unmarried couples due to late marriages and the declining fertility of married couples as factors behind the decline in TFR. According to this official report, the five main background factors behind the decline in TFR are as follows: (1) women entering the workplace, (2) severe housing conditions, especially in metropolitan areas, (3) difficulties in balancing child-rearing and work, (4) increased psychological and physical burden of child rearing, and (5) increased cost of child rearing. To address these issues, the Ministry of Health, Labour and Welfare (MHLW) requested prefectures and municipalities to actively consider and implement measures for TFR value recovery. Kamata et al. [5] noted that the development of such policies in local governments was uniform with few regional differences across the country, although the economic base and population size of each municipality had a certain degree of influence. However, there are still large regional differences in the actual TFR values. A comparison by prefecture shows that the highest value in 2022 is 1.70 in Okinawa Prefecture, while the lowest value is 1.04 [6] in Tokyo metropolitan area. In particular, high TFRs are reported to be maintained in a number of municipalities in the Kyushu region and Okinawa prefecture [7, 8].

Many of the studies [9, 10, 11, 12] related to TFR use models that equalize TFR variations that are highly coincidental for small municipalities such as islands and depopulated areas in order to obtain accurate national values. This makes it difficult to grasp the actual situation of TFR ups and downs in small municipalities after eliminating coincidental factors. In other words, in the case of the studies [7, 13] by municipality, the individuality and commonality of the factors increasing or decreasing the TFR can be approached. However, we have not established a method of measurement that separates chance and inevitability in the numerical variation related to the demographics of municipalities with small populations.

Based on our analysis of and response to these issues [14, 15], we have attempted to examine the factors that increase the TFR by aggregating municipal data into units of secondary medical areas (SMAs) composed of several neighboring municipalities. In our previous studies [14, 15], following three advantages were argued that aggregating data in units of SMAs (1) allows aggregation of raw data without equalizing TFR values, (2) reduces the risk of errors when the population size is small [11], and (3) allows data to be aggregated in units similar to living area [16]. As a results, it possible to take advantage of common geographical and socioeconomic background factors. However, these previous analyses were limited to Kyushu region and Okinawa prefecture, and only extracted characteristics of areas where TFR is rising or maintained. In comparing geographical characteristics, it is considered useful to analyze not only areas where TFR is increasing but also areas where TFR is decreasing, and an analysis covering a wider area by utilizing the analysis method for each SMA is required.

In this study, we expand the target to all municipalities in Japan and analyze the relationship between time-series changes in TFR and the geographical and socioeconomic background factors. In particular, to clarify the relationship between women’s social advancement and TFR [14, 17], we will examine the relationship with the Gender Empowerment Measure [18] and other gender indicators.

Materials and Methods

1. Ensure continuous demographic data of local governments before and after the Heisei Era Great Mergers

Given the shrinking size of municipalities due to the decline in Japan’s rural population, mergers of municipalities across the country were implemented from 1999 to 2010. These were called that “Heisei Era Great Mergers” [19]. As a result, the number of municipalities was consolidated from approximately 3,200 to approximately 1,700, stabilizing the financial bases of municipalities, which had been threatened due to the declining birthrate and aging population, and increasing the efficiency of policy effects through wide-area mergers. However, from the perspective of time-series research, the continuity of demographic data for merged and newly merged municipalities has been disrupted, making long-term tracking impossible.

To address these issues, this study follows the methodology of our 2023 study [14] by using municipal merger records from 2000 to 2020 [20], assuming that mergers during this period took place in 2000, and using weighted averages of demographic indicator of the municipalities from before the formation of the municipality. Sado City, Niigata Prefecture, is given as an example of a specific method for calculating 21-year TFR values from 2000 to 2020 for a SMA. This city was merged with 10 municipalities in 2004. The TFR values for the SMA were calculated by adding the population of women aged 15-49 and the number of births in each of the 10 constituent municipalities for each year from 2000 to 2020 according to the common method.

2. Calculation of TFR

TFRs for SMAs were obtained by weighting the TFRs calculated from the number of births to women aged 15-49 years per municipality based on the TFR calculation method presented by the United Nations: ESCAP [21].

3. Demographic Adjustment in Small Municipalities (Significance of Consolidation into Secondary Medical Areas)

In small municipalities with small populations, problems often arise when the number of births changes only slightly in terms of the actual number of births, but changes excessively when the fertility rate is calculated. There are two conventional methods used to address these problems: the first method, adjustment using the ratio of the standard population to the actual number of children and women aged 15-29: the child-woman ratio (CWR) [11]; and the second method, equalization of demographic information (birth rate, mortality rate, etc.) of small-population municipalities such as islands using the Bayesian method [22, 23] presented by the Japanese MHLW. In the cases of using these methods, it may be impossible that specific factors in small-population municipalities such as islands, can be detected.

Therefore, in this study, TFR values were calculated for each year from 2000 to 2020 for 335 SMAs using raw data without equalizing the TFRs of surrounding municipalities using the methods as described above. This enabled the detection of factors that increase fertility rates common to multiple islands by using the percentage of the population living on an island within a SMA as an indicator.

4. Basic Statistics and Trend Indicators of Long-Term Time-Series Variation of TFR

The characteristics of the data in this study are time-series variation data consisting of 21 TFR points at one-year intervals spanning 2000-2020 for each SMA. Therefore, basic statistics consisting of a few parameters such as mean and standard deviation (sd) based on normal distribution, which are generally applied to cross-sectional data, cannot capture the time-series variation trend of TFRs. Therefore, the TFR-trend was captured by a total of 26 various parameters shown in i), ii) and iii) below.

i) Eleven basic statistics: mean, minimum, maximum, range, sd, mean of the last 3 years (2018-2020) (mean3y), coefficient of variance (CV), correlation coefficient (r), contribution ratio (r²), linear regression equation (slope of αx+β), and quadratic regression equation (slope of αx²+β)

ii) Table 1 shows 13 candidate regression curves including with linear lines to show the TFR-trends. The 13 regression curves consist of the following functions: linear, quadratic, cubic, quartic, pentadecimal, inverse (1/x), logarithm (log(x)), compound, power, sigmoid, grow, exp(exponential), and logistic. The optimal regression curve is the equation that minimizes the Akaike Information Criterion (AIC) [24]. Mathematically, the relationship between the convex or concave extrema of the polynomial and the dimension of the function is formed by three years. In the case of 21 points for years 2000-2020, the maximum number of extreme values showing convexity and concavity every three years is 10. In this study, the maximum degree of the polynomial was set to 5 and dimensions 6-10 were omitted to avoid an overly detailed analysis.

iii) Two other parameters: The difference between the linear equation and the optimal regression equation (AIC minimum), and the difference between the quadratic equation and the optimal regression equation (AIC minimum). These are indicators of the complexity of TFR-trend as they represent the difference in goodness of fit between the simple basic trend (linear, convex, or concave) and the optimal regression curve.

5. Comprehensive indexing of TFR time-series trend indicators (factor analysis)

Factor analysis was performed on the 26 indicators listed in i), ii), and iii) in Section 4.1 to extract factors that were then converted into a comprehensive index. IBM SPSS Statistics 18.0.1.0 (142) was used as the analysis software, and factors with eigenvalues greater than 1.0 were extracted by using the principal component analysis method as the initial analysis and adding varimax orthogonal rotation. Interpretation of each factor was based on the positive/negative sign and absolute magnitude of the factor loadings, which are the single correlation coefficients between the 26 indicators and each factor.

6. Association of TFR-trend with geographic and socioeconomic information and gender indicators of SMAs (Mann-Whitney U test)

The top 25% and bottom 25% of SMAs were extracted using the factor scores for each factor extracted in section 5. For the two groups divided by high and low factor scores, the following 11 indicators were compared by Mann-Whitney U test as geographic and socioeconomic indicators and gender indicators for each SMA. The indicators were selected with reference to the global gender gap index (GGGI) [25], the gender inequality index (GII) [26], and its predecessor, the gender empowerment measure (GEM) [18] (used by the United Nations Development Programme), and included items related to the economy, education, health, and politics. The selected indicators were as follows:

1: Rate of workers in primary industry, 2: Rate of workers in secondary industry, 3: Rate of workers in tertiary industry, 4: Rate of graduates from junior college or higher, 5: Rate of nuclear family households, 6: Rate of elderly population, 7: Average number of hospital beds per capita, 8: Rate of residents living on islands, 9: Proportion of women in municipal assemblies, 10: Women’s labor force participation rate, 11: Number of obstetrics and gynecology hospitals per 100,000 women aged 15-49.

7. Choropleth map

The strength and weakness of the factor scores of representative examples of TFR-trend-factors obtained are shown as a choropleth map of Japan consisting of 335 SMAs. The ranks are defined as quartiles of the factor scores. This visualization allowed us to confirm the geographic distribution of the TFR-trend-factors and complemented the quantitative analysis described above.

All data in this study were obtained from secondary sources published on the Internet by the Cabinet Office, the Ministry of Internal Affairs and Communications, and MHLW.

Results

1. Typology of TFR-trends

Table 1 shows the frequency at which each of the 13 regression curves is optimal for the TFR-trends of each of the 335 SMAs. Three-dimensional polynomial curve was optimal for 38.21% (128) of the SMAs, followed by four-dimensional polynomials for 32.84% (110) and 5-dimensional for 12.24% (41). The best fit to the TFR-trend for a total of 83.28% (279) of the SMAs were 3-5 dimensional models with repeated concavo-convexity. Straight lines were judged optimal for just 3.28% (11) of SMAs, and simple unimodal concavo-convexity was optimal for just 2.09% (7). Logarithm type was optimal for 3.28% (11), and inverse type was optimal for 2.99% (10). Sigmoid type showed optimal fit in 2.39% (8) of SMAs, logistic type was the best fit in 2.09% (7) of the SMAs, the compound and exponential types were optimal in 0.30% (1) of SMAs, and the other two types were not optimal in any SMAs.

2. TFR-trend-factors

Table 2 shows the results of factor analysis (varimax orthogonal rotation method) of the basic statistics of TFR-trend for SMAs and 24 indicators, excluding two exponential types (power and grow) that were not applicable out of 26 trend indicators. Factor loadings with an absolute value of 0.4 or greater are indicated in Table 2. Eleven factors with eigenvalues greater than 1.0 were extracted, with a maximum eigenvalue of 3.72 and a cumulative contribution rate of 84.59%. Labels for each factor were interpreted using high loading variables. Factors with eigenvalues greater than 2.0 were Factor 1 of the maintenance of long-term high TFR values, Factor 2 of the magnitude of amplitude of TFR values and Factor 3 of the linear increasing trend of TFR. Interpretations of the remaining factors (4 to 11) are shown in the lower part of Table 2.

3. Relationships among TFR-trend and geographic and socioeconomic characteristics of SMAs and gender indicators (Mann-Whitney U test)

To explore the factors contributing to the TFR-trend, we compared 11 items of geographic and socioeconomic information and gender indicators (Mann-Whitney U test) between two groups: the top 25% and bottom 25% of factor scores. In particular, two of the extracted factors (Factor 1: the maintenance of long-term high TFR values, Factor 3: linear increasing trend of TFR) were used, as they indicated a positive trend in TFR-trend.

Table 3 shows the results of the comparison of the two groups for Factor 1. Among the index values for which the Mann-Whitney U test was significant, those that were high in the top 25% group are highlighted in pink, and those that were high in the bottom 25% group are highlighted in blue. The indicators that showed significant differences were rate of workers in primary industry (p<0.001), rate of workers in tertiary industry (p<0.01), rate of graduates from junior college or higher (p<0.01), average number of hospital beds per capita (p<0.001), rate of residents living on islands (p<0.001), proportion of women in municipal assemblies (p<0.001), women’s labor force participation rate (p<0.05), and number of obstetrics and gynecology hospitals per 100,000 women aged 15-49 (p<0.01).

Table 4 shows the results of the comparison of the two groups in Factor 3. As in Table 3, index values that were significant in the Mann-Whitney U test are highlighted in pink and blue. The indicators that showed significant differences were rate of workers in primary industry (p<0.001), rate of workers in secondary industry (p<0.001), rate of workers in tertiary industry (p<0.001), rate of graduates from junior college or higher (p<0.001), rate of nuclear family households (p<0.001), rate of elderly population (p<0.001), proportion of women in municipal assemblies (p<0.001), women’s labor force participation rate (p<0.05), number of obstetrics and gynecology hospitals per 100,000 women aged 15-49 (p<0.001).

4. Choropleth map

Figure 1 shows the factor scores of the largest eigenvalue, Factor 1 (the maintenance of long-term high TFR values), divided into quartiles in a choropleth map. The distribution of the 84 SMAs belonging to the highest quartile is geographically concentrated, with a large number of SMAs in the Chugoku, Kyushu regions and Okinawa prefecture. Of these, 27 SMAs include outlying islands within the area.

Similarly, Figure 2 shows the factor scores of Factor 3 (linear increasing trend of TFR), divided into quartiles in a choropleth map. The distribution of the 84 SMAs in the highest quartile is concentrated in metropolitan areas (Sapporo, Tokyo, Nagoya, Osaka, and Fukuoka) and the region surrounding the Seto Inland Sea. Of these, 24 SMAs include isolated islands within the area.

Both of these two factors tend to have positive TFR-trend, but Factor 1 is centered in the southwest islands group and Factor 3 in the metropolitan area, and there is a significant difference in the characteristics of their geographic distribution. In other words, the geographic distributions of Factor 1 and Factor 3 suggest that the background factors of TFR-trend may differ.

Discussion

1. TFR-trend factors

The results of the factor analysis shown in Table 2 imply that 24 statistical indicators for TFR-trend are aggregated into 11 types in 335 SMAs nationwide from 2000 to 2020. The cumulative contribution of 84.6% indicates that the indicators are highly complementary. In particular, the first factor indicates that TFR tends to remain high as an independent trend separated from the increase/decrease and concavity trends. The second factor represents the magnitude of the fluctuation range of TFR, and the third factor represents the strength of the rightward linear change. These TFR-trend factors are independent (uncorrelated) of each other, as shown by the results of the varimax orthogonal rotation analysis. After the fourth factor, the optimal regression curves were markedly different, indicating that there is a diverse distribution of TFR-trend factors in SMA units.

2. Regional characteristics of TFR-trend Factor 1 (background factor analysis and analysis with maps)

Among the factors listed in Table 2, we analyzed the background factors for SMAs that were correlated with TFR-trend Factor 1 and Factor 3, which indicated a positive TFR-trend. The results in Table 3 show the background factors strongly related to TFR-trend Factor 1. One of the characteristics is that SMAs with a high Factor 1 (meaning high long-term TFR) value have a high rate of residents living in islands, which is consistent with the findings of Muramatsu N [27] and Masuda J [8]. Next, as an economic characteristic of TFR-trend Factor 1, in SMAs with high long-term TFR, the rate of workers in primary industry tends to be high and the rate of workers in the tertiary industry tends to be low. According to a document from the Regional Development Promotion Secretariat [28], the rate of tertiary industry is highest in Tokyo, followed by 74 government-designated cities and core cities, and then the nation as a whole, indicating that the areas where TFR remains high are not in urban areas but in rural areas. Figure 1 shows that the areas with high factor scores for the first factor are mostly in the southwestern region, which has many islands, and many metropolitan areas and core cities are excluded. In the education sector, there is a significant trend toward higher TFR values at lower rates of higher education. The tendency of higher education to lower the TFR was mentioned by Masuda J [8] in Japan, and similar trends were reported by [29] in Korea, Sweden [30] in Sweden, and Turkey [31] in other countries. On the other hand, a similar study in Finland found no significant relationship between education and TFR [32]. Other characteristics of SMAs with long-term high TFR include a high average number of hospital beds per capita, a high number of obstetrics and gynecology hospitals per 100,000 women aged 15-49, and a high female labor force participation rate. On the other hand, there was a significant trend toward lower percentages of female members of municipal councils in SMAs with long-term high TFR.

These data on economy, education, health, and political participation are included in the calculation of gender indices such as GGGI [25] and GII [26], which are used as indicators to measure women’s advancement in society. The results of this study showed that all of these indicators were associated with the TFR, but the data tended to differ in whether the effect was toward increasing or decreasing the TFR. Although there are several studies on the relationship between gender indicators and TFR, the direction of the relationship is not consistent, with a mixture of studies showing that TFR decreases [33, 34] or increase [35, 36] as women advance in society. Under these circumstances, Nakagaki [37] analyzed the relationship between the four sub-indexes of the GGGI [25] and TFR, and found that the strength and direction of the impact differed for each sub-index. In other words, indicators related to social advancement do not simply result in decreasing TFR, but rather have different impacts depending on the TFR-trend and social advancement indicator items that each region has. Therefore, it is considered important not only to directly compare TFRs with comprehensive indicators related to gender, but also to analyze the relationship between TFRs and the sub-indexes and individual data that make up the comprehensive indicators, and to do so for each region separately.

3. Regional characteristics of TFR-trend Factor 3 (background factor analysis and analysis with maps)

The results in Table 4 show background factors strongly associated with regions with linearly increasing TFR (Factor 3). The results differ from those in Table 3 in that they show that a low ratio of primary and secondary industries, a high tertiary industry rate, a high level of education, a high rate of nuclear families, and a low rate of elderly population are associated with linearly increasing TFR. All of these are characteristics that are common in urban areas. From these results, Factor 3 showed a trend of increasing TFR in highly urbanized areas, which has never been confirmed in previous studies. In other words, while Factor 1 indicates a trend in the Kyushu region and Okinawa Prefecture, which show stable high TFR values over a 20-year period, Factor 3 indicates a trend in urban areas, where TFR increases linearly over time. Figure 2 shows that the areas with high scores for Factor 3 are distributed mainly in the five major metropolitan areas of Sapporo, Tokyo, Nagoya, Osaka, and Fukuoka, as well as in the area around the Seto Inland Sea in the Chugoku and Shikoku regions. Previous studies on TFR disparities by prefecture [9, 10] have discussed a contrast between high-TFR regions centered in the Southwest Islands and urban areas with low TFR. However, this study instead finds that these two regions both have distinctly positive TFR trends, though they are defined by distinctly different TFR trend patterns.

The difference between TFR-trend Factor 1 and Factor 3 is noteworthy here: TFR-trend Factor 1 is not characterized by an upward trend in the factor analysis results shown in Table 2. However, in Table 2, the coefficient α of the primary regression equation (No. 11): y=αx+β represents the slope of the regression line. Thus, it indicates that TFR is increasing in a gradual but linear manner in the five major metropolitan areas and the area surrounding the Seto Inland Sea related to TFR-trend factor 3. Although the change is not large in absolute value, it is a significant and definite change. Furthermore, in the TFR-trend Factor 1, a high rate of students going on to junior college or higher was associated with a lower TFR, while in Factor 3, a high rate of students going on to higher education contributed to an increasing trend in the TFR. These trends, observed in both island and urban areas, are consistent with the results of our previous study [14] in the Kyushu region and Okinawa prefecture.

The relationship between the gender index and the TFR for Factor 3 shows that a higher ratio of female legislators is significantly associated with increasing TFR, a result that differs from that of the TFR-trend Factor 1. Kaneko Y [38] pointed out that the characteristics of regions with a high ratio of female councilors in local councils are a high level of education and metropolitan areas, which is consistent with the characteristics of the areas shown in Figure 2 of this study. Other characteristics pointed out include the fact that residents in regions with a high ratio of female councilors are more politically active, and that election outcomes are influenced by residents who actively volunteer their efforts. These trends suggest that the Seto Inland Sea area shown in Figure 2 may have a high level of political participation among its residents even though it is not as urbanized as the other areas in that figure. In order to understand the relationship between regional characteristics and TFR, it is important to conduct a survey of individual characteristics, attitudes toward childbirth and childrearing, and the current status of women who were included in the TFR calculation across SMAs, where the time-series trends of TFR differ significantly.

Only the female labor force participation rate was significantly associated with higher TFR for both TFR-trend Factor 1 and Factor 3. This suggests the strength of the impact of economic factors on childbearing and childcare. According to Kawaguchi’s review of trends in female employment and fertility in 21 OECD member countries [17], the correlation analysis of female labor force participation rate and TFR shows a reversal of the slope from negative to positive between 1970-1975 and 2000-2005. In other words, the OECD as a whole suggests that the relationship between various background factors such as women’s labor force and TFR differs from country to country (region to region) and from era to era. As outlined in the Introduction, the national proposal “Basic Direction of Policies to Support Child-Rearing in the Future” [4], posits that women’s advancement in society is a factor in the decline of the TFR. However, the results of this study showed that gender indicators exerted both an upward and downward influence on the TFR. Therefore, an analysis that takes into account time series and regional characteristics is considered to be useful. Regional correlation studies on gender indicators and TFR have analyzed the relationship from an international perspective [33, 34, 35, 36], but it is highly novel and extremely significant that this study clarified the relationship at the SMA level.

4. Future Tasks

The results of this study reveal an association between TFR-trend and gender indicators in SMA units. There is no clear correlation between social advancement of women and decreasing TFR; rather, gender indicators have different effects depending on the TFR-trend indicator used and the specific social advancement indicators in each region. Therefore, further analysis of the relationship between the individual data comprising the gender indicators and the TFRs is required in the future. In doing so, it will be important to conduct related studies based on the individual characteristics of women’s social advancement in each region. It will be necessary to clarify the relationship between life events that women go through (employment, marriage, childbirth, and childcare), gender indicators, and indirect factors of each region by conducting surveys on women’s childbearing and childcare environments and attitudes in regions with high and low TFR.

The results of the review by Kawaguchi [17] and this study suggest that the relationship between TFR and background factors such as female labor force participation rate and educational level varies over time. Since the relationship with the TFR may change depending on time period of the data under study, a time series analysis is required. Specifically, it is considered necessary to analyze changes in the relationship between the TFR and gender indicators by organizing which gender indicators are involved in periods when the TFR is rising and, conversely, in periods when the TFR is declining.

Conclusion

Municipal information related to large-scale municipal mergers in the last 20 years (2000-2020) was aggregated and organized by SMAs to analyze the relationship between time-series changes in TFR and geographic, socioeconomic, and gender indicators in all SMAs (335) in Japan.

There are three notable findings from this study:

i) The validity of the method of analyzing TFR values in small municipal units nationwide by aggregating the data into SMA units without equalizing them based on TFR values of surrounding municipalities was confirmed.

ii) Significant associations between time-series variation in TFR and gender indicators such as GII and GGGI were demonstrated, and an association between women’s social advancement and TFR was revealed.

iii) In addition to the Southwest Islands and other regions identified in previous studies as having high TFR, the five metropolitan areas (Sapporo, Tokyo, Nagoya, Osaka, and Fukuoka) and the region surrounding the Seto Inland Sea were newly identified as regions with linearly increasing TFR.

The results of this study showed that in both TFR-trend Factor 1 (maintaining high TFR) and TFR-trend Factor 3 (TFR linearly increasing), an association was found in all areas of data on gender indicators (economy, education, health, political participation). This suggests that the improvement in social status that accompanies women’s entry into society may increase TFR. According to the national proposal “Basic Direction of Policies to Support Child-Rearing in the Future” [4], women’s entry into society is a factor in the decline of TFR, but the results of this study provide different and important findings due to improved analytical methods. Increment of TFR requires not only measures such as maternity and childcare benefits and parental leave, which have traditionally been taken by the national and local governments, but also a wide range of support related to women’s social advancement is required.

On the other hand, the main sources of variation in TFR tended to differ according to the TFR trends and social advancement indicators of each region. Therefore, it is considered important not only to directly compare TFRs with comprehensive indicators related to gender, but also to analyze the relationship between TFRs and the sub-indices and individual data that make up the comprehensive indicators. In the future, it will be necessary to conduct surveys on women’s childbearing and childrearing environments and on their attitudes to clarify the relationship among life events that women go through (employment, marriage, childbirth, and childrearing), gender indicators, and indirect factors specific to each region.

Acknowledgements

We would like to thank the teachers working in the Department of Health Informatics at Niigata University of Health and Welfare for their valuable guidance. We would also like to thank Dr. Jeffrey Huffman for his great contribution to the English language editing.

Conflicts of interest

This research received no grant from any funding agency in the public, commercial, or not-for-profit sectors. The authors declare that they have no relevant conflicts of interest.

References

• 1.  Cabinet Office. Choosing the Future - The Future Visible from Population Estimates - Chapter 3: Current Situation and Issues Surrounding Population, Economy, and Local Communities [Internet]. Cabinet Office [updated 2015 Oct 28; cited 2024 Aug 25]. Available from: https://www5.cao.go.jp/keizai-shimon/kaigi/special/future/sentaku/pdf/p030105.pdf (in Japanese)
• 2.  Ministry of Health, Labour and Welfare. 2015 Edition. Annual Health, Labour and Welfare Report — Consideration of a depopulating society — [Internet]. Ministry of Health, Labour and Welfare [updated 2015; cited 2024 Aug 25]. Available from: https://www.mhlw.go.jp/english/wp/wp-hw9/dl/summary.pdf
• 3.  Atoh M. Policy responses to lowest-low fertility in Japan: a twenty-year trajectory and evaluation. Human Science Research. 2010; 23(2): 187-207. (in Japanese)
• 4.  The Ministry of Education, Science, Sports and Culture, Ministry of Health and Welfare, Ministry of Labor, Ministry of Construction. Regarding the basic direction of future child-rearing support policies (Angel Plan) [Internet]. National Institute of Population and Social Security Research [updated 1994 Dec 16; cited 2024 Aug 25]. Available from: https://www.ipss.go.jp/publication/j/shiryou/no.13/data/shiryou/souron/19.pdf (in Japanese)
• 5.  Kamada K. Policy process of measures to combat the declining birthrate in local governments: Econometric analysis of the timing of policy output using the “Local Government Survey on Measures to Support Raising the Next Generation”. The Review of Economics & Political Science. 2010; 78(3-4): 213-242. (in Japanese)
• 6.  Ministry of Health, Labour and Welfare. Vital Statistics Survey: Annual total fertility rates by prefecture [Internet]. Ministry of Health, Labour and Welfare [updated 2024 Mar 29; cited 2024 Aug 25]. Available from: https://www.e-stat.go.jp/dbview?sid=0003411598 (in Japanese)
• 7.  Ito W. The environment of the bringing up children in Tarama Village in Okinawa Prefecture — the life and culture in Tarama Island with the highest birthrate in Japan. The Jumonji Journal of Human Life Sciences. 2005; 3: 63-88. (in Japanese)
• 8.  Masuda J. Why is the total fertility rate of Kyushu region High? An essay on birth and parenting in Kyushu region, Japan. Bulletin of Nakamura Gakuen Developmental Support Center. 2021; 13: 18-27. (in Japanese)
• 9.  Tanabe K, Suzuki T. Analysis of prefectural differences in fertility rates. Journal of Health and Welfare Statistics. 2016; 63(5): 13-21. (in Japanese)
• 10.  Hiroshima K, Mita F. Prefectural differentials in recent fertility. Journal of Population Problems. 1995; 50(4): 1-30. (in Japanese)
• 11.  Yamauchi M. The new indirect estimation method of TFR using standardized CWR. Journal of Population Studies. 2009; 45: 35-44. (in Japanese)
• 12.  Koike S. Comparison of Actual and Bayesian Estimates of Total Fertility Rate in Vital Statistics by Municipality. FY1990 Ministry of Health, Labour and Welfare, Policy Science Research Project [Internet]. Ministry of Health, Labour and Welfare [updated 2021 Mar; cited 2024 Aug 25]. Available from: https://mhlw-grants.niph.go.jp/system/files/report_pdf/202001018A-sonota1.pdf (in Japanese)
• 13.  Katoh H. Analysis of fertility rate disparities and their causes by municipality. Financial Review. 2017; (131): 6-23. (in Japanese)
• 14.  Kinoshita N, Luong THY, Takano K, et al. Regional differences in total fertility rates (TFRs) and background factors in Kyushu region and Okinawa prefecture - Time-series study (2000-2020) using 63 Secondary Medical Areas -. Fukai Institute of Health Science (FIHS). (in press)
• 15.  Luong THY, Kinoshita N, Namizuka A, et al. Comparison of environment and attitudes toward childbirth and childcare of women living in the lowest- and highest-TFR (total fertility rate) areas in Japan: synergy analyses of quantitative and qualitative data. Niigata Journal of Health and Welfare. 2023; 23(2): 45-64.
• 16.  Ministry of Land, Infrastructure, Transport and Tourism. Concept of Living Area 3. What is a living area [Internet]. Ministry of Land, Infrastructure, Transport and Tourism [updated 2009 Mar 9; cited 2024 Aug 25]. Available from: https://www.mlit.go.jp/sogoseisaku/region/21seikatuken/04/03.pdf (in Japanese)
• 17.  Kawaguchi A. Trends in women’s employment and fertility. Social Policy and Labor Studies (Shakai-seisaku). 2005; 14: 18-37. (in Japanese)
• 18.  WHO. Definition of Gender-empowerment measure (GEM) [Internet]. WHO [updated unknown; cited 2024 Aug 25]. Available from: https://www.who.int/data/gho/indicator-metadata-registry/imr-details/3671
• 19.  Shimada A. A comprehensive review of the “Heisei Great Mergers”. Local Autonomy Fukuoka. 2018; 64: 3-37. (in Japanese)
• 20.  Geospatial Information Authority of Japan. Municipal Mergers by Prefecture [Internet]. Ministry of Internal Affairs and Communications [updated unknown; cited 2024 Aug 25]. Available from: https://www.soumu.go.jp/gapei/gapei_h11iko.html (in Japanese)
• 21.  Lee E. A comparison of fertility measures from censuses and a civil registration and vital statistics system: The case of Republic of Korea [Internet]. United Nation: ESCAP [updated 2020 June; cited 2024 Aug 25]. Available from: https://www.unescap.org/sites/default/files/APS2020/SC3_63_Comparison_of_fertility_measures_from_censuses_and_vital_statistics.pdf
• 22.  Ministry of Health, Labour and Welfare. Methods of Bayesian estimation in the 2010 Municipal Life Tables [Internet]. Ministry of Health, Labour and Welfare [updated 2013 July 31; cited 2024 Aug 25]. Available from: https://www.mhlw.go.jp/toukei/saikin/hw/life/ckts10/dl/03.pdf (in Japanese)
• 23.  Saeki N, Hirako T, Nakata T. Application of Bayesian statistics to demographic municipal statistics (2) application to total fertility rate. Journal of Health and Welfare Statistics. 1999; 46(11): 3-10. (in Japanese)
• 24.  Alexandre Zajic. What Is Akaike Information Criterion (AIC) [Internet]. Built in [updated 2022 Nov 29; cited 2024 Aug 25]. Available from: https://builtin.com/data-science/what-is-aic
• 25.  World Economic Forum. Global Gender Gap Report 2023 [Internet]. World Economic Forum [updated 2023 June 20; cited 2024 Aug 25]. Available from: https://www3.weforum.org/docs/WEF_GGGR_2023.pdf
• 26.  United Nations Development Programme. Human Development Reports [Internet]. United Nations Development Programme [updated 2024 Mar 13; cited 2024 Aug 25]. Available from: https://hdr.undp.org/system/files/documents/global-report-document/hdr2023-24reporten.pdf
• 27.  Muramatsu N, Kumano K, Nishikawa T. Construction of analysis base for total fertility rate (TFR) time course: An attempt to search association for TFR by municipality. Musashino University Academic Institutional Repository. 2020; 5: 72-89. (in Japanese)
• 28.  Cabinet Office. Materials from the 4th Regional Attractiveness Creation Experts Conference: Strengthening urban industry and employment structures and urban functions [Internet]. Cabinet Office [updated 2018 Oct 11; cited 2024 Aug 25]. Available from: https://www.chisou.go.jp/sousei/meeting/chiikimiryoku_souzou/h30-10-11-shiryou1.pdf (in Japanese)
• 29.  Song JE, Ahn JA, Lee SK, et al. Factors related to low birth rate among married women in Korea. PLoS One. 2018; 13(3). doi.org/10.1371/journal.pone.0194597
• 30.  Eggert J, Sundquist K. Socioeconomic factors, country of birth, and years in Sweden are associated with first birth fertility trends during the 1990s: A national cohort study. Scandinavian Journal of Public Health. 2006; 34(5): 504-514.
• 31.  Eroğlu K, Koruk F, Koruk I, et al. Women’s reproductive behaviour and perspectives on fertility, and their modifying factors, in a Turkish province with a high fertility rate. Eur J Contracept Report Health Care. 2021; 26(2): 139-147.
• 32.  Oura P. Sociodemographic indicators of birth rate in a low fertility country-A nationwide study of 310 Finnish municipalities covering > 5,000,000 inhabitants. Frontiers in Public Health. 2021; 9: 1-7. doi.org/10.3389/fpubh.2021.643561
• 33.  Gotmark F, Andersson M. Human fertility in relation to education, economy, religion, contraception, and family planning programs. BMC Public Health. 2020; 20(1): 265. doi.org/10.1186/s12889-020-8331-7
• 34.  Lesthaeghe R, Permanyer I. European sub-replacement fertility: Rrapped or recovering? University of Michigan, Population Studies Center Research Reports. 2014; 14-822.
• 35.  Amra F. Gender inequality and fertility: Empirical evidence for Bosnia and Herzegovina [Internet]. Academia. [updated unknown; cited 2024 Aug 25]. Available from: https://www.academia.edu/44661622/Gender_inequality_and_fertility_Empirical_evidence_for_Bosnia_and_Herzegovina
• 36.  Suzuki T. Declining birth rates in Japan, East Asia, and Europe: trends, causes, and policy responses. J of Population Problems. 2012; 68(3): 14-31. (in Japanese)
• 37.  Nakagaki Y. Convex relationship between fertility and gender gap. Economics Bulletin. 2019; 39(3): 2014-2026.
• 38.  Kaneko Y. Why are there so few female members of local assemblies in Japan? -A study of local assemblies in Yamagata Prefecture-. Annual Report of Political Science. 2010; 61 (2): 151-173. (in Japanese)