目標：主要探討台灣地區從1980年開始到2005年為止，359個鄉鎮市區因子和鄉鎮層次青少女生育率的相關性，及鄉鎮市區因子可否預測五年及十年後的青少女生育率。方法：資料來源為台閩地區人口統計、縣市要覽以及戶口及住宅普查資料。樣本為台灣地區359個鄉鎮市區，從1980年到2005年，每五年抽樣一次，共計2,154個樣本。分析方法使用雙變項分析、重複測量的混合模型(Mixed Model)分析以及時間遲滯模式(Time-lagged Model)。本研究的區域因子包含人口組成(人口密度、原住民人口百分比、65歲以上人口百分比和離婚率)、社經地位(高等教育人口百分比)、社會醫療福利服務(每萬人口醫師數和社會福利支出)以及其他因素(遷徙率)。結果：在同年模式中，人口密度越低、每萬人口醫師數越低、高等教育人口百分比越低、65歲以上人口百分比越高、原住民人口百分比越高的地區，青少女生育率是越高的。區域因子影響青少女生育率也有出現時間遲滯的情形，也就是說，在控制了當年青少女生育率後，區域的人口密度、65歲以上人口百分比、原住民人口百分比、高等教育人口百分比這四個變項，可穩定預測五年及十年後的青少女生育率；而區域的離婚率則是要經一段時間才能影響青少女生育率，以十年的效果最大，離婚率愈高，青少女生育率愈高；遷徙率只有在五年及十年時間遲滯模式中有達到顯著，遷徙率愈高則青少女生育率愈高。結論：影響青少女生育率的因子，以社會結構層面因子較為明顯，像是人口密度、65歲以上人口百分比、原住民人口百分比、高等教育人口百分這四個區域因子，都跟當年的青少女生育率有關，也可穩定預測五年後及十年後的青少女生育率。未來政策制訂者應該要看每個影響因子背後所帶有的意義，也應該要努力去達到地區平均發展機會。

Objectives: This study examined the relationship between township-level
factors and teenage birth rates from 1980 to 2005 in 359 townships in
Taiwan. We also examined whether township-level factors were associated
with five-year and ten-year teenage birth rates. Methods: The data came
from Taiwan Demography, The City and County Statistics, and Census. The
study used repeated measures of township-level characteristics every 5
years from 1980 to 2005 (N=2154). We conducted bivariate analyses and
used mixed models with repeated measurements and time-lagged models to
analyze the data. The variables included population composition
(population density, the percentage of aborigines, the percentage of
people over 65 years old, and divorce rates), socioeconomic status (the
percentage of people with college degrees), medical and social services
(the number of physicians per 10,000 people and social welfare spending)
, and other factors (residential mobility). Results: In the contemporary
model, population density, the percentage of physicians, and the
percentage of college educated people were negatively associated with
teenage birth rates, whereas the percentage of elderly and the
percentage of aborigines were positively associated with teenage birth
rates. In the time-lagged models, after controlling for current teenage
birth rates, population density, the percentage of college educated
people, the percentage of elderly, and the percentage of aborigines
were associated with five-year and ten-year teenage birth rates. The
relationship between the percentage of divorced people and teenage
birth rates was strongest in the 10-year model. Residential mobility
was not significant in the contemporary model, but was significant
in the 5-year and 10-year models in which higher residential mobility
predicts higher teenage birth rates. Conclusions: These findings
suggest that township-level social structural characteristics have
more significant effects on teenage birth rates than do other
township-level characteristics. Population density, the percentage of
elderly, the percentage of aborigines, and the percentage of college
educated people were associated with teenage birth rates in the
contemporary model, the 5-year model, and the 10-year model. Future
policy should focus on examination of the theoretical meaning of each
township factor and improvement in regional equity as well as
fulfillment of regional needs.