Fertility Level, Changing Trends and Possible Relaxation of the One-Child Policy in China

Transcription

1 Fertlty Level, Changng Trends and Possble Relaaton of the One-Chld Polcy n Chna Abstract -specfc fertlty rates n Chna decrease fro 199 to 2. Fro 2 onward, these rates for frst-born babes ncrease everywhere and at all ages ecept for woen under 27 n ctes, under 26 n towns, and under 25 n countes rural vllages, ecludng towns because young people arry late. For second and above babes, fertlty rates have also ncreased, eceedng the lts set by the one-chld polcy. Coparsons of the predcted age-specfc rates for 29 and the observed rates fro an annual survey of the.1% of the entre populaton n 29 valdate the conclusons on the assessents of the fertlty level and the changng trends. Based on the estated age-specfc fertlty rates over years and current woen's age dstrbuton, Chna s supposed to be eperencng a peak or peaks of new born babes, and t s not a good tng to rela the one-chld polcy rght now. KEY WORDS: age-specfc fertlty, cubc splne, Lee-Carter odel, logstc regresson odel, perod total fertlty rate, one-chld polcy

2 1. Introducton Fertlty n Chna s affected by the governent's one-chld polcy, whch s one for woen wth non-agrculture household regstraton, 1.5 chldren for woen wth agrculture household regstraton, and 2+ chldren for ethnc nortes. The polcy was launched n 1979 when the country counted a quarter of the world's populaton. In 211, though the populaton s stll ncreasng, t represents one ffth of the world's populaton. Hesketh, Lu, and Xng and Wang revew the pleentaton and the consequences of the polcy regardng the decreasng TFRs, unbalanced se rato, the growng proporton of elderly people, and the ncreasng rato between elderly parents and adult chldren. Hesketh, Lu, and Xng and Greenhalgh and Bongaarts 1987 suggested that everyone could be allowed to have up to two chldren, wth a brth space of at least fve years. Wang suggested an edate change of the polcy. We shall assess the age-specfc fertlty rates and the TFR over years n Chna. The data are fro the Chna Populaton Statstcs Yearbooks Natonal Bureau of Statstcs of Chna and The Ffth Populaton Census Data Natonal Bureau of Statstcs of Chna 2, and the odel s a odfed Lee-Carter odel hao Fertlty Data n Chna 2. Fertlty Level and Changng Trend n Chna

3 In recent years n Chna, censuses were carred out n 1982, 199, 2, and 21, surveys of about 1% of the entre populaton were conducted n 1987, 1995, and, and saller surveys of about.1% of the entre populaton were conducted n the rest of the years. The Chna Populaton Statstcs Yearbooks Natonal Bureau of Statstcs of Chna and The Ffth Populaton Census Data Natonal Bureau of Statstcs of Chna 21 contan vald age-specfc and brth order-specfc fertlty data of four groups of woen, the country, ctes, towns, and countes for 1987, 199, 1995, 1997, and , where the three no overlappng groups: ctes, towns, and countesrural vllages, ecludng towns, ake up the country. Brth order n the orgnal data contans four categores, all babes, frst-born babes, second babes, and thrd and above babes. In ths study, the data for second babes and thrd and above babes are put n one group, the second and above babes. There are 5 age groups for woen of ages between 15 and 49, and we use the ddle pont of an age perod to represent the age of the group. For nstance, 15.5 s used to represent the age of the group of woen between 15 and 16. The data for 1996 and 1998 are dropped, as the data for 1996 are n 5-age groups and the nubers of babes born n 1998 n ctes, towns and countes do not add up to the nubers n the country where the nuber for ctes and towns are recorded as the sae for all ages and brth orders by stake. Besdes the nuber of babes, we have also added up the average nubers of woen n ctes, towns, and countes, and have checked f they atch the correspondng nubers n the country. The data for 29 were not avalable when we started the analyss; the census data for 21 are not

4 avalable yet. Overall, there are 12 subsets of data, three brth orders all babes, frst-born babes, and second and above babes for each of the four groups of woen the country, ctes, towns, and countes. In each subset, the average nuber of woen aged n year t and the nuber of babes these woen gve brth to can be wrtten as nt,,dt,. For all babes n the country, the observed age-specfc fertlty rates dt,/nt, for the ajor years, 1987, 199, 1995, 2 and are shown n Fgure 1. For each age group, over the years, the fertlty rate reaches ts peak and dp n 199 and 2, respectvely, ecept for woen of ages between 2.5 and 25.5, for whch the lowest fertlty rates are acheved n. The peak n 199 reflects the phenoenon of the so called etra-brth guerrlla era n Chna, durng whch soe fales wth daughters ove secretly fro place to place to escape the brth-control regulators hopng to produce a son. There s a draatc drop n fertlty after 199, and the an reason s the proulgaton of the Provsonal Regulatons on the Enrollent of Tuton-payng Students n the Regular Hgher Educaton Insttutons by the Natonal Educaton Cottee 199, effectve as of July 9, 199, whch ade parents feel the fnancal pressure of rasng a chld. The hgher educaton four years college used to be tuton free n Chna but was accessble only for the elte. Those students who dd not have a chance to enter college before used to get arred earler and to gve brth to babes. Attendng college postponed ther ages for havng a baby. In addton, soe of those who had pad tuton by theselves preferred the DINK

5 Double Incoe No Kds lfestyle after ther graduaton, whch also contrbutes to the low fertlty n the late 199s. The observed age-specfc fertlty rates dt,/nt, are good estates of the true age-specfc fertlty rates for census years, but not for the survey years of the.1% or 1% of the entre populaton. Usng the well-known Lee-Carter odel Lee and Carter 1992, Lee 199 analyzed age-specfc fertlty n US. For the fertlty data n Chna 2-28, we estate the epected age-specfc fertlty rates usng a odfed Lee-Carter odel hao Modelng -specfc Fertlty For a fed set of knots Lee-Carter odel s,, 1, the general for of the odfed p t, ln 1 p t, t t t t 1 { } 1 { t } 1 where pt, s the unknown fertlty rate at age and year t, and, j, j,j=,1,2,,,1,, are all unknown paraeters; the functons of age,, 2,, 1,, are defned as a1, a2 and a see 1 Append for odels wth cubc, quadratc and lnear functons above the last knot, respectvely. and are defned for odels wth 2 quadratc and lnear restrctons below the frst knot, respectvely. For each of the nne scenaros of the lnear, quadratc, and cubc restrcton cobnatons at the two tals, odel 1 s a logstc regresson when ftted to annually collected bnoal data

6 nt,,dt,. For each of the 12 subsets of fertlty data, followng a two-step procedure, a fnal best odel s selected. The procedure s: a. For a fed scenaro and a fed nuber of knots, an optal set of knots that produces the sallest resdual devance s selected aong all possble nteger knots by runnng an optal search progra, and the odel wth the optal knots s stored as a canddate for the fnal best odel; b. Aong all the canddates odels, a odel wth relatvely saller resdual devance and saller p values s chosen as the fnal best odel. Note that usng dfferent odel selecton crtera, we ay have dfferent fnal best odels, but the epected age-specfc rates do not dffer uch. Table 1 lsts the fnal best odel wth the optal knots, the scenaro of restrctons at the tals, and the resdual devance for each of the 12 subsets. For eaple, the fnal odel for all babes n the country s: p t, ln p t, <.1.7t < < <.1 -.6t < <.1 -.2t.151.1t < t <.1 4 < t < t <.1 5 <.1 Where 15, 21 19, 22 21, , 4 2, 5 2 and the values n the brackets below the estates are the correspondng p-values. For each of the three brth orders, all babes, frst-born babes, and second and above babes, Fgures 2-4 show the observed and ftted 2, and 28

7 age-specfc fertlty rates for woen n each of the four groups of people. In general, the observed and ftted rates ft each other well for years 2 and - the year wth a larger saple sze plays a bgger role n estatng the unknown paraeters n the odel, but not so well for year 28. It s worth notng that the so called observed rates are actually the estates of the real rates ecept for census years, therefore the observed rates for 28 and ehbt larger volatlty than that for 2. In Fgure2, there s a hup around n each of the fgures for the country, towns and countes, but not for ctes, whch reflect the specal 1.5 chldren polcy for woen holdng an agrcultural household regstraton,.e., a second brth s allowed f the frst chld s a daughter over fve years old. The woen wth agrcultural household regstraton are n the group of towns or countes whle none of the are n the group of ctes by defnton. In Fgure, for frst-born babes, the fertlty reaches ts peak at about age 2 for countes' group, 24 for towns' group and 25 for ctes' group. Over the years, the ftted fertlty ncreases for woen over 28 n ctes, over 26 n countes and over 27 n the country and towns, and decreases sgnfcantly for woen ages n the ctes' group. For woen under 26 n ctes and under 25 n towns and countes, the observed rates are lower than the ftted values, a sgn that the fertlty rates ght have reached ther peaks n. In Fgure4, for second and above babes, the observed and the ftted fertlty rates ft each other well for towns, but not so well for ctes, countes, and the country n 28. The fertlty reaches ts peak at -1 for ctes, 29- for towns, and for

8 countes, whch s a fve-year shft fro that of the frst-born babes, and hence produces the hups around for all babes as n Fgure4, as the fertlty rate for all babes s the su of the fertlty rates for frst-born babes and second and above babes. 2. Perod Total Fertlty Rates To avod confuson, TFR1 and TFR2+ are used to denote the total frst and second and hgher fertlty rates respectvely,.e., TFR s the su of the correspondng TFR1 and TFR2+. The observed 2-28 and the ftted and predcted perod total fertlty rates over the years based on the fnal odels lsted n Table 1 are shown n Fgure 5, where the ftted 2-28 and the predcted rates are presented n lnes. For second and above babes, the observed and the ftted TFR2+s ft well for towns, but the ftted TFR2+s see a lttle hgher than the observed values for ctes and countes after. Eanng the values of the observed TFR2+s, we fnd that n the ost recent years, -28, the values are over.15 and.65 for ctes and countes, respectvely, whch have eceeded the brth lts set by the one-chld polcy of one for ctes and 1.5 chldren for towns and countes. Though there are soe rch people who prefer to pay the penalty to have a second baby, the nuber of such people s lted. The observed and the ftted TFR2+s for the towns' group are reachng.5. For frst-born babes, the observed and the ftted TFR1s ft well for all four groups

9 for For towns and countes, the observed values are already beyond 1 n 28, and there s not uch space to grow. For all babes, the observed and the ftted TFRs ft well for ctes and towns, but the ftted TFRs appear hgher than the observed TFRs for countes for 27 and 28, as both TFR1 and TFR2+ eceed the lts set by the one-chld polcy. In the country, the observed and the ftted TFRs ft well for all the brth orders. The coparson of the observed and ftted total fertlty rates provde a reason why the observed and ftted age-specfc fertlty rates do not atch each other well for soe of the subsets. 2.4 Predcted and Observed -specfc Fertlty Rates for 29 The dscrepancy between the observed and epected age-specfc rates n 28 ndcates the estence of nonlnear change of fertlty over perod 2-28 and reflects the fact that data n 2 and play larger roles n estatng the unknown paraeters than that n 28. For frst-born babes and all babes, the fertlty rates ay have reached ther peaks n for young woen, whch ay be caused by the rebound after etree low fertlty around 2 and the fact that young woen tend to get arred relatvely late n the very recent years hao, Hou, and hao 211. Because of the delay, we can re-estate the age specfc rates based on the -28 data. For eaple, n the country, a fnal best odel s chosen usng the two-step procedure, and the observed and ftted age-specfc rates for and 28

10 are presented n Fgure 6. They atch each other well. After we had fnshed the analyss, we obtaned the Chna Populaton Statstcs Yearbook 21, whch contans the fertlty data for 29 fro an annual survey of.1% of the entre populaton. The observed age-specfc fertlty for the country are calculated and added to Fgure 6 wth the predcted rates for 29 based on -28 data. They ft each other well for ages under 25, whch confr the delay n gvng brth to babes. Unfortunately, there s stll a dfference between the observed and predcted rates for ages over 25 n 29, whch ay be a result of possble under-countng of new born babes n or 29, or over-countng fro or nonlnear change over the years for these ages. We added the observed TFRs, TFR1, and TFR2+s for 29 to Fgure 5. As we have epected, due to the lts set by the one-chld polcy, for second and above babes, the observed TFR2+ and age-specfc fertlty rates for 29 are at about the sae level as that n 28, lower than the predcted values for 29 based on the odel, but hgher than the level n 2 for all the four groups. For frst-born babes and all babes, the observed TFR1 and TFR for 29 are close to the predcted values n ctes but lower than the predcted values n towns and countes. Dscusson and Concluson Predcton of age-specfc fertlty s dffcult especally for a huge and constantly changng country lke Chna. Yet, we have no doubt that wth the contnued econoc prosperty n the recent years n Chna, the fertlty s ncreasng and has grown

11 beyond the lts set by the one-chld polcy for certan groups of woen. Over the years, ore and ore woen n ctes attended hgher educaton, four years of college, and even graduate school over the years, hence they postponed gvng brth. But why are the changes for the woen n towns and countes relatvely slow? In Chna, those who were n towns or countes, as long as they passed the college entrance ea and entered a regular college, ther regstratons hukou were changed fro agrculture household regstratons to non-agrculture household regstratons ctes. As long as they can fnd a job n a cty, they wll rean holdng ther non-agrculture household regstratons, thus classfyng the n the group of ctes. For those who dd not pass the ea, they ay have a chance to be adtted to a short-ter college two-year or three-year to receve techncal tranng f ther scores are not too low, but ther regstratons wll not be changed. After two or three years' tranng, ost of the prefer to fnd a job n a cty or town, and few would lke to return to countes. Therefore, the percentage of woen n the groups of ctes and towns are ncreasng over the years, but the percentage n the group of countes s decreasng. Fro the 2 census, there were totally 2,76,216 woen between ages 15 and 49 n the country, 8,681, % fro ctes, 4,62, % fro towns, and 19,4, % fro countes. Fro the survey, the total nuber s 4,676,587 n the country, 1,45,645.7% fro ctes, 85,6 17.9% fro towns and 19,4, % fro countes. That s, soe woen n the countes' group have oved nto the ctes' or towns' group; soe woen n the towns' group have oved nto the ctes' group; those who stay behnd

12 would observe the local culture of gettng arred and gvng brth to babes at the tradtonally epected ages, though there s a slght delay n ages of gettng arred hao, Hou, and hao 211 whch certanly results a slght delay of gvng brth to new babes. There are sudden drops n the observed TFR1 n 29 n towns and countes, besdes the possble under-countng of new born babes n the annual survey of.1% entre populaton n 29, the rapd urbanzaton that s happenng n Chna ay be another possble reason. In the very recent years, ore and ore people n towns or countes had been relocated to spare ther house or land for assve constructon projects such as new roads, hghways, shoppng centers and resdental buldngs. Most of these people have to lve n rented apartents or houses for 2- years before ovng n the new apartents copensated by the governent. The relocatons ay also caused delay n producng new babes. The 21 census data n Chna are not avalable yet, to get a prelnary populaton projecton n the country, the ftted rates for 2-24 for all babes n Fgure2 and the ftted rates for -28 n Fgure 6 can be used as the fertlty rates for the years, and the estated fertlty rates for 29 can be corrected by replacng the rates over 25 wth the rates of the correspondng ages n 28. After the correcton, the estated TFR s 1.58 for 28 and for 29. Should Chna rela the one-chld polcy rght now? Besdes the fertlty level, other factors that are related to the sze of populaton should also be consdered, such as the rapdly decreasng ortalty n Chna hao et al. 211, whch contrbutes to

13 the ncrease of total populaton n Chna. In fact, the populaton n Chna s stll ncreasng each year and the ncrease wll last for over a decade accordng to our prelnary results n a populaton projecton study. Also, the dstrbuton of woan's age should also be consdered. In Chna, feales born around 199 etra-brth guerrlla era are about to gve brth to new babes. Though we do not have access to the 21 census data, based on 2 census data, a bar plot of the nuber of woen for dfferent age groups n 2 s presented n Fgure 7, whch shows that there were uch ore woen aged 1-15 n 2. These woen should be 2-25 n 21, and they are about to at the ages of hgh fertlty. Therefore, there s supposed to be a peak n the nuber of new born babes n the very near future, and t s not a good tng to rela the one-chld polcy rght now. Snce the urbanzaton and the shft of the coposton of woen's regstraton do affect the fertlty n the long run, further studes are needed to fnd a good te to rela the polcy n the future.

14 Append Three sets of forulae used n the odfed Lee-Carter odel,,,1,, 1,2,, j j j a1, 1;,1,, 1,2;, j j j a2 and. 2;,1,, ; ; a where j s defned as j for and for. References [1] Agrest, Alan. 22. Categorcal Data Analyss 2nd edton. New York: Wley. [2] Greenhalgh S. and J. Bongaarts Fertlty Polcy n Chna: Future Optons. Scence 25: [] Hesketh, Therese, L Lu, and hu We Xng.. The Effect of Chna's One-Chld Faly Polcy After 25 Years. The New England Journal of Medcne 5: [4] Lee, Ronald D Modelng and Forecastng the Te Seres of US Fertlty: Patterns, Range, and Ultate Level. Internatonal Journal of Forecastng 9:

15 [5] Lee, Ronald D. and Lawrence Carter Modelng and Forecastng U.S. Mortalty. Journal of Aercan Statstcal Assocaton 87: [6] Natonal Bureau of Statstcs of Chna Eds Chna Populaton Statstcs Yearbook of Bejng: Chna Statstcs Press n Chnese. [7] Natonal Bureau of Statstcs of Chna. 21. The Ffth Populaton Census Data 2, n Chnese, retreved Aprl 25, [8] Natonal Educaton Cottee Provsonal Regulatons on the Enrollent of Tuton-Payng Students n the Regular Hgher Educaton Insttutons, n Chnese, retreved May 4, [9] Wang, Peng.. Can Chna Afford to Contnue Its One-Chld Polcy? Asa Pacfc Issues, East-West Center #77. [1] hao, Bojuan Barbara A odfed Lee-Carter Model for Analyzng Short Base Perod Data, Populaton Studes, onlne: 16 Dec 211, [11] hao, Bojuan B., Xanglang Lang, Wenke hao, and Delong Hou Modelng of Group-Specfc Mortalty n Chna Usng a Modfed Lee-Carter Model, Scandnavan Actuaral Journal, onlne: 24 Oct 211, [12] hao, Bojuan B., Delong Hou, and Wenke hao Trends and Patterns n Martal Status n Recent Chna under revew.

16 fertlty rates Fgure 1: For all babes, observed age-specfc fertlty rates for the ajor years n the country.

17 a fertlty n the country, all b fertlty n ctes, all c fertlty n towns, all d fertlty n countes, all Fgure 2: For all babes, the observed and ftted age-specfc fertlty rates for each of the four groups of woen for 2,, and 28 based on odels lsted n Table 1.

18 a fertlty n the country, frst b fertlty n ctes, frst c fertlty n towns, frst d fertlty n countes, frst Fgure : For frst born babes, the observed and ftted age-specfc fertlty rates for each of the four groups of woen for 2,, and 28 based on odels lsted n Table 1.

19 a fertlty n the country, second+ b fertlty n ctes, second c fertlty n towns, second+ d fertlty n countes, second Fgure 4: For second and above babes, the observed and ftted age-specfc fertlty rates for each of the four groups of woen for 2,, and 28 based on odels lsted n Table 1.

20 a TFRctes, towns, countes b TFR1ctes, towns, countes ctes towns countes.6.5 ctes towns countes Year-2 Year-2 c TFR2+ ctes, towns, countes d the country TFR, TFR1, TFR2+ 1. ctes towns countes 2. TFR2+ TFR TFR Year-2 Year-2 Fgure 5: Observed, ftted and predcted perod total fertlty rates based on the odels lsted n Table 1 based on 2-28 data. Newly obtaned observed rates for 29 are added and arked wth sold ponts.

21 fertlty n the country Fgure 6: Observed, ftted and 28 and predcted 29 age-specfc fertlty based on the -28 fertlty data n Chna. Newly obtaned observed rates for 29 are added and arked wth crosses.

22 nuber of woen n llons n year 2 Fgure 7: Bar graph of the nuber of woen n dfferent age groups n 2 Data source: The Ffth Populaton Census Data n Chna 2.