Fourteenth Annual Conference of the European Society for Population Economics

June 14-17, 2000, Bonn

Scenarios of living arrangements, household and family structures

The proportion of the families of the
entire population decreased in the last decades drastically. The *Gemeinschaften*-type
communities, whose member in various relations are connected, lose
quantitatively at importance in relation to one-person households and the
living arrangement of the singles. People marry in later age years than decades
ago, more frequently they do not marry at all. At the same time divorces
increased. Likewise the low birth numbers since the 60‘ and even more clearly
the decline in the birth-rate in the new Länder in Germany after the turn
demonstrate lasting behaviour modifications. Therefore the demographic aging
will strengthen in the next decades further and the questions about future
living together of the families and the generations become more urgent. This is
accompanied by the concern accompanies whether and in which extent the
subsidiary supports between generations, partners and in families will be at
all possible (Lowenstein 1999). With the following scenarios of the population
structure in Germany bases are to be supplied to the quantitative future
forecast.

With the model calculations which can be submitted here new ground was entered in two different regard: The scenarios go beyond past work, because the size and structure of the population are to be calculated not only after age and gender, but also after living arrangements and households. And it uses a so-called dynamic projection procedure, i.e. that the estimate units result from transitions of one age to the next and particularly from transitions between the living arrangements, assuming Markovian chains. Contrary to conventional procedures of the household projections in general no structure ratios were set.

The project "Projections of living arrangements, household and family structures" was executed together by the Federal Institute for Population Research (BiB, Wiesbaden), the Max Planck Institute for Demographic Research (MPIDR, Rostock) and the Federal Statistical Office (StBA, Wiesbaden). The BiB in 1993 and 1995 had compiled model calculations of the household and family structures for the seniors, then for the total population (Höhn, Roloff 1994; Dorbritz, Hullen, Schiener 1995). Also the 1996 final network study of the BiB (Schulz 1996) offered views of the assistance and interactions within the family networks of the future seniors. The Federal Statistical Office submitted 1996 a household forecast on the base of the most recent population projection (Voit et al. 1996).

Zeng Yi (University of Peking and since 1997 head of the Research Unit on Family Dynamics and Aging in the Max Planck Institute for Demographic Research at Rostock) developed a computer program for the integrated projection of the population and household structures (Zeng Yi, Vaupel, Wang Zhenglian 1997). "ProFamy" was distributed to different users as Windows application since 1999.

Household projections so far predominantly are performed with the headship-rate method or with the household member method derived from it (Linke 1983; as example for Germany cf. Roloff, Höhn 1994). Projected populations are divided into subpopulation in accordance with ratios of household types. Such quota approach thus does not deal at all with the endogenous factors of household development.

The micro-analytic approach for household projections follows a completely different procedure. For all individual cases of the population or a population sample its eventual demographically relevant changes are simulated. Apart from births, deaths and partnerships the household formations and household solutions can belong to it.

Macrosimulations differ from microsimulations by the fact that persons‘ groups of same statuses are aggregated. Thus the amount of computation reduces itself anyhow. Microsimulation and macrosimulation are equally exposed to the objection that they can hardly consider all possible feature combinations in a population and not even all possible transitions between selected living arrangements (Arminger, Galler 1991). Their methodical procedure frequently was called "dynamic approach" or "causal appraoach", because they – differently to the „static" or "non causal" quota approach – consider also also the endogenous change factors of the population structure.

Population projection which take use of a base population and asumptions concerning the fertility, the mortality, and if necessary over the migration, are to be carried out today with simple computer programs. Batch-programs on large computers lost their earlier monopolistic position. A population projection differentiating after gender and age can be created today easily with spread-sheet programs on the PC. The numbers can be transferred in addition easily in graphs, even if desires remain open to the graph types.

Scenarios become more difficult naturally, if they partition the populations after other features. Programs for microsimulation and macrosimulation are supposed to follow the abundance of the data in multi-state population models and supply at the same time a clear pleasing result output.

The PC-program „ProFamy“ was developped by Zeng Yi together with Wang Zhenglian. It uses a procedure which proceeds from the family-status life-table of Bongaarts (1990), thereby benefiting from the advances in the multi-dimensional demography (Rogers 1975) and especially the multistate life-table model (Willekens et al. 1982; Willekens 1990). Zeng Yi extended Bongaarts' model, by including both nuclear families and three-generation households (Zeng 1988, 1991). In the life table models of Bongaarts and Zeng Yi the population development is determined only by one gender, namely women, and it assumed that the age structure of the transition probabilities remains constant. During the far development of Zengs family life table Zeng, Vaupel and Wang (1997) came to a two-sexual dynamic projection model. In it demographic behaviour patterns can change, and one does not need data, which - in the opinion of the authors - were not available from conventional data sources.

The modelling a Markovian process is the base of the model. It is assumed that changes of living arrangements are dependant on the age and on the current arrangement, not however of their prehistory and expressly also not on the duration of the past living arrangement. In addition the statistical assumption is applied that persons with same features are subject to the same risks of the change to other statuses.

We summon some phrases of the program authors on the computational strategy (Zeng, Wang 1998):

Let l_{i}(x,t) denote the number
of persons of age x with composite state (i=1,2, ..., T) in year t. Let P_{ij}(x,t)
denote the probability that a person of age x with composite state i in year t
will survive and be in composite state j at age x+1. Thus,

_{}

If P_{ij}(x,t), which are the
elements of a TxT matrix, were properly estimated. the calculation of l_{j}(x+1,t+1)
would be straightforward. Unfortunately, the estimation of P_{ij}(x,t)
is usually not practical when the total number of states T is large, which is the
case in our model.

Following Bongaarts (1990: 209-211) it is assumed that births occur throughout the first half and the second half of the year, and the birth probabilities used refer to the corresponding half year and they depend on the status at beginning and middle of the year respectively the other demographic events occur at middle of the year: deaths, migration, changes of coresidence with parents, marital status transitions, and changes in the number of surviving and co-residing children. The are dependant on transitions probabilities of the whole year and the status on beginning of the year.

Because the model deals with two sexes and both children and parents, some procedures are adopted to ensure the consistency.

(1) Equal numbers of male marriages and female marriages, furthermore equal numbers of cohabiting males and the number of female counterparts are ensured by using the arithmetic mean of the estimation aggregates.

(2) The status of children living with their parents might change by leaving the parental home, the death of on parent or by the own death. The adequate changes of the living arrangements of the parents are given by consistency algorithms.

(3) Births of children in the two-sex-model are computed both for the female and the male population. The number of births produced by the male population are adjusted to equal to the numbers of births by the female population. Because single-year age and parity specific fertility rates for male population are rarely available, males‘ birth rates are estimated on the base of females‘ birth rates and the average age difference between the male and female partners.

(4) Of course death and separations in partnership are changing the living arrangements of children. This is ensured by adequate consistency algorithms.

(5) Consistency algorithms are made available also for changes in the living arrangements of children induced by remarriage of parents.

Demographic accounting equations are used to compute the number of persons and the number of changes of their living arrangements. The basic structure is:

L(x+1,t+1,j) = L(x,t,i) + SE(ij) - SE(ji)

SE(.) is the sum of changes from i to j resp. from j to i with {1,...,i;1,...,j} Î T.

To simplify the calculation it is assumed that the events, e.g. the deaths, are independent from each other. The program user might decide on some more assumptions. One of them per example is the default that married siblings are not living together and that children of separated couples are living with their mother.

Population projections according to the component method proceed from a population of base, who in accordance with certain demographic parameters is updated. At least assumptions are used over the birth rate and the number of deaths of this population in the projection period. If international migrations are to be included, appropriate assumptions are necessary. Model calculations of the living arrangements, household and family structures need in addition assumptions over the transitions between living arrangements and over the household formation.

The estimates of the demographic parameters must be plausible - except if consciously extraordinarily extreme assumptions are set. Assumptions of constancy are obvious in most cases in the meaning that the demographic parameters of the base period are held unchanged for the projection period or sections of it. Even if this implies the assumption, past trends would not continue, which might be far from reality, these scenarios meet at best the expectation to show what will change under such circumstances in any case. The momentum of demographic developments has a lasting effect for a long time. If demographic model calculations do not proceed from constant parameters, then the appropriate setting are to be justified if possible.

The assumptions on fertility and mortality were not located in the focal point of the model calculations, and migration was not at all considered. Therefore we refer to other model calculations (outline in the appendix). The total fertility rate is situated in the earlier federal territory for approximately two decades between 1,28 (1985) and 1.45 (1975, 1976 and then also again 1990). There are few reference points for the expectation that this level would change substantially. In the new Länder of the Federal Republic the total fertility rate fell 1993/94 on a historical low of 0,77 and achieved 1997 for the first time again a value of over 1.

For Germany altogether the model calculation proceeded from a total fertility rate about 1,32 as in the base year 1996. The parity-specific fertility had to be estimated because the birth sequence in Germany is analysed only for births in existing marriages (marital parity-specific fertility). The value of 29.47, proven for marital births 1996 (StBA, FS 1, R 1, 1996, table 9.4), was taken over as mean age at birth. The assumption of constancy appears as justified, because increases of the fertility are possible with further promotions of the family, however hardly to be estimated in their extent. A lower fertility would be just as little impossible.

The average life expectancy will continue to increase in the future with high probability. An automatic constant would not appear as meaningfully. On the basis of the abbreviated life table for Germany 1995/97 (StBA, FS 1, R 1, 1997, table 5.19) it was assumed in the model calculation that the life expectancy of the men to 2040 of 73.6 years rises by 5.7 years to 79,3, those of the women of 80 years by 5.8 years on 85,8. This is the low version of the mortality assumptions of Birg et al. 1998. In other versions those authors calculated with an increase of the life expectancy until 2040 of up to 10 years (Birg et al. 1998; Birg 1999c). Thus the model calculation submitted here now can be defined as conservative. If one would proceed from higher life expectancies, all changes of the population structure, which are generally connected with demographic aging and with an increasing proportion of the older generation of the population, would become apparent still more clearly.

In order to simplify the model calculation, it proceeds from a balanced migration balance. In addition the assumption comes that nevertheless possibly real transnational migrations do not have an effect on the population structure (it could be imagined for example that younger persons immigrate whereas older emigrate). Of course the population size of Germany, also the age structure in the next decades will be influenced substantially by the migration surplus which can be expected with high probability. The future population presumably will be larger and "younger" than in the model calculations. That must be taken into consideration with the following predicates for the development of the resident population of base of 1996. It remains open also whether the entire future population has the same structures as the resident population and its descendants.

Marriages and entering non-marital partnerships even as their solutions are of outstanding importance for the household formation and the forms of living together. The marriage can show a tradition for thousands of years, a intercultural spreading and an allegedly indisputable cultural and legal granting privilege. The non-marital partnerships however, in which likewise partners of different sex live in a common household in a long term relationship (other partnerships are to remain here out of consideration), gained public attention only after they had increased rapidly.

There are conceptions over the future of the marriage and also the non-marital partnership almost in great quantities. That is to be expected in view of the meaning of the heterosexual partnerships for all individuals and easy popularization. Assumptions of only small changes surely are most common. The fast change of the last decades does not support it however.

There are conceptions over the future of the marriage and also the non-conjugal/marital partnership almost in great quantities. That is not to be expected in view of the meaning of the heterosexual partnerships for all individuals and easy popularization differently. A common assumption is that the changes will be small. The fast change of the last decades does not support it however.

On the basis of the marriage rates of never married persons for Germany 1996 and of (especially calculated) age-specific remarrying probabilities it is assumed that the marriage frequency remains unchanged in the entire prognosis period. The marriage propensity of the men was assumed - due to calculations inside the program of ProFamy - with 62 per cent, their first marriage age with 29.7 years. For women a marriage propensity of 74 per cent and a first marriage age of 27.3 years were determined (first marriage age as per StBA, FS 1, R 1, 1996, table 8.4 and 8.9: men 30.0 years, women 27.6 years). The remarrying became estimated on annually 1.7 per cent - for lack of a sufficient database without distinction between divorced ones and widowed and, apart from an slight decrease in the higher ages, without age differentiation.

Transition probabilities into non-marital partnerships were taken from the Family and Fertility Survey (viz. to the FFS: Pohl 1995; Hullen 1998; Roloff, Dorbritz 1999), whereby a minimum duration of three months was set. It amounted to 20 to 24 per cent for women at age 20 to 24, to 8 to 20 per cent for the men of this age. It dropped to 18 to 13 per cent for women at age 25 to 29, resp. to 19 to 12 per cent for men. For age groups over 39, the maximum age of the respondents of the FFS, the parameters had to be estimated.

Modifications of the nuptiality of course have also modifications with divorces to the consequence, possibly also for the dissolutions of non-marital partnerships. In West Germany the proportion of the marriages (starting from marriage year 1990), which end in a divorce or will end, was estimated on 36 per cent, at the marriage cohort 1950 it was 10.4 per cent (Dorbritz, Gärtner 1998: 423). The divorce propensity in the long-term view increased strongly, on the other hand the acceleration of the increase in the last years was no longer as large.

In times of smaller nuptiality, then a thesis is, the proportion of stable marriages of all marriages is higher than in times of earlier nuptiality. One proceeds from a stronger selection of those pairs, which are safe of their decision and which can master better the conflicts of a marriage. This "weeding thesis" is surely applicable, if one remembers that both absolutely and relatively the number of the "early marriages" decreased, which are subject to an above average high divorce risk. Besides the assumption is represented that divorces will further increase in a society, in which the marriage lost its exclusiveness for partnership living arrangements. Even if some will marry for a second or further time, the number of the married persons will decrease (Furstenberg 1987: 30; Beck-Gernsheim 1989a, P. 116; Hall 1997: 289).

It is consistent to argue similarly as for the divorce also for the separations in non-marital partnerships: If one assumes partnerships decrease altogether in favour of a further „singularization“, then the conclusion is narrow that the remaining ones will have a at least continuous, if not even higher stability than today. If one assumes however that living in different living arrangements - successively or also repeated - in the so-called patchwork biographies increases, then a smaller stability also the non-marital partnerships has to be assumed (Prinz 1995: 171).

In the constant scenario it is assumed that the divorce rate remains unchanged in the entire projection period. The age-specific probabilities of divorce were inferred from the official statistics. From thousand married men and women each at the age from 20 to 34 1995/1996 annually nine to ten divorce. In the younger and at the higher age the divorce rates are naturally lower. For men at age 50 for example they amounted to 4.5 per thousand, for women to 3.4.

Four to six per cent of the non-marital partnerships annually lead into a marriage according to the results of the FFS. In contrast to this the probability that the partners of the non-marital partnership separate, has maxima with 27 per cent for women at age 21-22 and of 30 per cent both for men at age 21-22 as starting from 30.

Assumptions of future trends of the founding of new households and the dissolution of households are to be described in the following. The leaving of the parental home and the entrance into a community or an institute accommodation (children's home, nursing homes, old people's homes etc.) or into a private household on dissolution of the own household are quantitatively most important. It has to be noted in advance that these demographic behaviours are indispensably linked with a certain flexibility of the housing market or the institution households. The move to another apartment presupposes that a new dwelling was offered, the accommodation into a home is bound to free places. The demographic decisions are contributed by the housing market and also the national policy, as far as they concern this market and the supply of homes.

One of the most irritating findings of the last years was that the youngsters remain ever longer in the parents' house. It surprised that in a time, in which in the business life larger mobility and independence are expected, private decisions over the living conditions are shifted to a later age. The move from the parents' house is not based obviously simply on the "naked" decision for a change of residence. The marriage or the birth of an (own) child precondition traditionally the leaving of the parents' house, or the changes to an educational institution, which is in a larger distance. Taking notice of this, it becomes recognisable that the move is just as dependent on other biographic events as of the availability over (additional) dwelling (viz. up-to-date Lauterbach, Lüscher 2000). Compared between nations there are obviously different patterns of the living arrangements of young adults. In Southeast-European countries the children remain longer in the parents' house, whereas it is more usual in the Scandinavian, west and Central European countries to live in an own one-person household (Manting, Alders 1998:15-18).

Into the scenarios the assumption on the propensity to leave the parents' house and the mean age were held constantly (98 per cent leave the parents' house, men at average with 25 years, women with 23 years; after values of the Family and Fertility Surveys).

Almost one per cent of the entire population and approximately five per cent at age 60+ in Germany live in institutions. Approximately 460,000 of them are 65 years old and older (1995), most of them living in old and nursing homes.

Projections of the population in institutions understandably expose themselves to the question whether the entrances and leaving children's homes, prisons, homes for the elderly etc. - apart from the lethal issue – are to be understood as stochastic processes, or whether they are determined to a high extent by exogenous setting, above all the policy. An example is given by the modification of the principles to the stationary home accommodation in the course of the discussion around the finally implemented nursing care insurance.

If it is at all considered, it is usual to hold the proportion of these subpopulations to a fixed amount. Thus, with unchanged age-specific proportions, Nelissen calculated a substantial both absolute and relative increase of the population in institutions in the Netherlands (Nelissen 1997). Manting and Alders also showed this effect of demographic aging (Manting, Alders 1998) in their scenarios of the households in the European union.

New aspects for model calculations of the living forms might itself open by so-called communitary living forms of older persons. Saup and Reichert distinguish between old person group houses, the owner-and-tenant communities and " integrated living " (Saup, Reichert 1997). With the latter young and alto live together. One may be sceptically, whether the new living forms gain quantitatively significance beside the traditional households and the homes, but it might be possible.

The model calculations for Germany, which can be presented here, proceed from 1996. For this year the most current Scientific Use file of the micro censuses is available, an actually anonymized, 70 per cent sample of the micro censuses (Köhler, Schmipl‑Neimanns, Schwarz 2000). It is as well known an annual one-per cent sample of the population in Germany. The data record consisted of over one half million person data records (509,243) with specification over age (0 to 95+), gender, marital status and - particularly importantly - the relations of the household members. Model calculations with more recent years will be possible, as soon as their Scientific Use files are present.

As stated above the necessary prognosis parameters were taken from the official statistics and also the Family and Fertility Survey. Some were calculated by ProFamy inside the program, for example the proportions of the population in institutions were calculated. The user of ProFamy can determine some adjustments. Here the following was decided:

· The number of the married men and women should be alike, as well those of the men and women in non-marital partnerships. Methodological with this consistency assumption a closed marriage market is assumed

· It is assumed that children do live all with the mother after divorces and after separations from non-marital partnerships

· The option to model also three-generation households more in detail was disregarded due to their small number

· The possibility of differentiating with the number of deaths according to marital statuses is disregarded

· Rechutes in the parents' house are not considered.

· The possible rural/urban distinction of ProFamy is understandably not considered because of missing data (it shows the possibilities of the program of differentiating after subpopulations).

· If inhabitants of one-person households were indicated with the marital status "married" indicate, this was changed to „divorced“ in the projection period. Thus the state of married persons who live separated who is given in reality and even fundamental by the German divorce law is unconsidered. Otherwise one would have to define it as a further marital status with appropriate transition probabilities to other statuses.

In the following two scenarios are presented. They differ by the fact that in the first one the population is divided after four marital statuses (single, married, widowed, divorced), while in the second additionally the non-marital partnerships were considered like a fifth status. The base year of all model calculations is 1996. The end of the projection period was set to 2040.

If - as assumed in the constant scenario - fertility remains in such a way like it is, the life expectancy continues to rise and if there is no migration surplus, then the population in Germany will naturally decrease. From approximately 82 million it goes back to 2020 to 74 million, to 2040 then on 63 million. The regenerating cohorts are in terms of figures narrower than the older, the point of the population pyramid - which corresponds to the picture of a pyramid thereby ever fewer - more broadly than their base.

In the following now the totals of the population and of its subpopulations are not in the focal point, but the population structure. In the subdivision of the population after the marital statuses the married ones remained the biggest subpopulation. The smaller marriage frequency let it decrease in the last years however, for men from 51 per cent in the year 1991 on 47 per cent until 1997, for women from 47 per cent in the year 1991 on under 45 per cent in 1997. The single proportion rose for men to over 45 per cent, for women to almost 37 per cent. Understandably it gives fewer widower than widows (under 3 per cent or 13 per cent). The relative proportion of the divorced ones is approximately alike with well 4 or 5 per cent with both sexes.

Table 1: Population by gender and marital status (in percent)

Source: Federal Statistical Office

The actual population structure will substantially change. It will give ever more cohorts with more single than married persons up to the fourth life. Entirely seen however, ever more strongly, the cohorts of the older ones with higher married than single proportions outweigh.

· In the population at all the married proportion compared with the last years is again rising, for men from 43 per cent in the year 1996 to 2040 to 50 per cent, with the/for women from 34 to 38 per cent. This is the result of the relative increase of those (older) cohorts, whose members by the majority are married.

· In contrast to this the proportion of the single ones could decline for men from 51per cent in the year 1996 to 2040 to 42 per cent, for women from 48 to 39 per cent.

· The widowed persons take off in absolute numbers, relatively however slightly too, for men at approximately three per cent, for women to approximately 16 per cent.

· The proportion of the divorced persons increases for men already until 2010 to over five per cent, for women 2030 seven per cent are achieved.

When the large cohorts, born 1950 to
1970, will become extinct the aging of the population will presumably slow
down, which lets the single proportion of the entire population rise again. A
period going beyond 2040 appears however as too far, in order to dare calculations
of their marital status structure.**
**

At present there are approximately 37 million households in Germany. Each about one third of them are one-person resp. two-person households. The one-person households increased in the last decades absolutely and relatively considerably. 1970 for example there have been in the earlier federal territory 5.5 million one-person households, that means a quarter of all households. Even if one regards only the last years, an increase of the small and smallest households becomes visible. In ever more households the "reference person" is single or divorced, also the number of the households of separated living married persons has increased. The model calculations are limited to the main places of residence of the population. Second domiciles, which nevertheless constituted 1996 over 800000 households according to official estimation, thus remained unconsidered. Institutions however are included.

Apart from a altogether good matching, the counting and projection of the households by ProFamy unfortunately shows lower figures in particular for the one-person households (31.7 instead of more expected 34.8 %). This on the one hand is due to the fact that weighting could only be done with the "person factor", on the other hand to complications with the initial counting of those households, which contain also non-family members (cf. Zeng, Vaupel, Wang 1999).

The model calculations confirm the expectation that the total number of the households will continue to rise, from 38 million in the year 2000 to 39 million in the year 2020. Afterwards the housing number diminishes again on 35 million in 2040 according to the shrinking population with a zero migration balance. In relation to earlier decades the small propensity to build partnerships, more frequent separations and the small child number as well as the increase of older ones without partners lead to more one-person households. Their proportion might increase from 2000 to 2040 from 37 per cent to 51 per cent, i.e. in each second household then only one person lives.

The proportion of two-person households rises from 33 per cent to over 36 per cent in the year 2020 and is falling thereafter however again, although slightly, on 34 per cent in the year 2040. In the last decades of the projection period the Baby Boomer of the 60's with their nuptiality still higher in relation to the following cohorts reach the age, at where death and widowing increase. Households with three and more persons decrease continuously. From 2000 to 2040 both the absolute number and its relative proportion bisect themselves. The average size of the households reduces itself thereby to 2040 to 1,8 persons.

If one groups the households thereafter, how many generations are living together, the effects of longer life and small child number show up. Households with two generations (parents and child, also single parents with child) remove, households with only one generation (single living persons, partnerships without children) increase absolutely and relatively strongly. The number of three-generation households remains small. The information about generations’ living together is still detailed in the following for certain subpopulations.

The increase of the in one-generation households is caused by the increase of the one-person households. From 2000 to 2040 it increases from 14 on over 17 million. The number of households of couples (without children in the household) rises only in the first two decades of the projection period to approximately ten million, when large cohorts reach the age, where children the parents' house already left, both parents however are alive. One-generation households of non-family related persons remain in the entire projection period with two to three million. These predominantly consist of non- marital partnerships (1996 1.7 million). A second larger group are the institutional households. With the ratios calculated for the present and held constantly for the future, this total number increases according with the age structure up to 2.7 million.

At present 43 per cent of the women at age 20 to 34 live together with children in their household (21 per cent with one child, 23 per cent with two or more children). Starting from 2010 this will presumably only apply to 26 per cent of the women. In other words: The percentage of the women at the age of 20 to 34 years, which do not live with children, increases in relatively short time to 2010 of 56 per cent to approximately 74 per cent. The serious modification of the households concerns naturally not only the women, but likewise the men living in partnerships. Causes of this development are both the high age of the mothers at birth and the small number. 1980 the mothers in the former federal territory at average were 25.19 years old at the birth of their marital living-born first children, 1990 however 26.93 years and 1995 even 28.15 years.

Fig. 5

Fig. 6

Effects of the decline of the birth rate show itself to clearest with women at age 35 to 54, that phase, where the establishment of family already took place, the children’s move out of the parents' house often in addition. Today the proportion of the women, who live either without child (0) or with more than one child together (2+), is equivalent largely. The proportion of the women with exactly one child in the household is somewhat smaller. These proportions move itself in favour of the households without children. Almost each second woman at age 35 to 54will live without child, three tenths with more than one child, two tenths with a child.

The youngest women and the oldest women understandably live together more rarely with children. Also in the future 99 per cent of women below 20 will be without children. The proportion of women at age 55 to 64 living together with children will decrease from 16 per cent to eleven per cent. This is an echo of the fast decline in the birth-rate starting from 1964, after which decades with a comparatively constant period fertility followed.

Three quarters of the older men, i.e. men at age 65+, today live together today with a female partner, mostly as married couple, fewer than a fifth alone. This will change in the next years due to the small marriage frequency. The proportion with a partner of the living together could be reduced to under 60 per cent. Accordingly the proportion of those men living in a couple would rise to a third. The proportion of the men, who live together with “others”, would remain small. The majority of them are inhabitants of institutional households.

Comparatively a greater proportion of the older women are living alone, caused also by their higher life expectancy. The model calculation shows a further increase to 2040 to 56 per cent living as a single. Accordingly the proportion of the women living together with a partner, falls after an easy intermediate increase to 2040 to 34 per cent. Approximately twice as many women as men, i.e. over 10 per cent, live together with others, above all probably with their children or in institutions.

The number of persons aged 80+ will approximately treble in the projection period. The conception that this will coincide with large modifications of their living arrangements, would be however false. Two thirds of the women, a third of the men will live alone. The proportion of persons living with a partner is about 59 per cent for men, for women it is rising from eleven to eighteen per cent.

The increase of the average life expectancy to 80 years and more affects more clearly the living arrangements of the persons at age 65 to 79. The proportion of men living alone increases from 14 to 35 per cent, the proportion living with a female partner declines from 82 to 58 per cent. 46 per cent of the women at that age are living together with a partner. This could rise to approximately 53 per cent until 2010, then however again decrease. The increase is due to the increasing life expectancy of the men, the biologically rising "availability" on the marriage market, the decrease to the small nuptiality and therefore actually decline of real availability.

In the years starting from 2010, when more women as in former times live together with partners, each fourth private household in Germany will be the household of one couple at age 65+.

In the second scenario the base population is differentiated by age, sex, marital statuses and household composition, and additionally non-marital partnerships are considered. In the micro censuses 1996 this living arrangement explicitly was asked for. The answer of the questions "Sind Sie Lebenspartner(in) der ersten Person“ [Are you engaged in a partnership/union with the first person?] and „Falls ein(e) Lebenspartner(in) der ersten Person im Haushalt lebt: In welcher Beziehung stehen Sie zum/zur Lebenspartner(in) der ersten Person?“ [If there is a partner of the first person in the household: Which relation do you have to the partner of the first person?] was voluntary, it does promise however reliable information because of the response readiness which can be subordinated, the size of the sample and because of the data preparation.

In ProFamy the affiliation to a (heterosexual) partnership is treated like a fifth marital status. Of course model calculations require then parameters of the transitions also between this marital status and the others. For simplification it was assumed that only single, widowed or divorced persons enter non-marital unions and that in the case of its termination they can be married or single again.

1996 four per cent of all men and of all women lived in non-marital partnerships. This increases to 2040 to nine per cent. The increase corresponds to the expectation that this living arrangement will quantitatively win importance, however not replace marriage. To that extent the assumption used in the model calculation lead to plausible numbers.

Naturally these proportions are very much higher for younger adults than altogether for the population. In 1996 twelve per cent of the men at age 25 to 29 and 16 per cent of the women of the same age belonged to such partnerships. In the year 2000 these proportions amount to 35 per cent for men and 31 per cent for women, 2010 then 38 or 34 per cent. These values result on use of the transition probabilities for the beginning and the termination of non-marital partnerships, calculated on the basis of the Family and Fertility Surveys of 1992. Even if there some living together reported as non-marital partnership remained unconsidered because their minimum duration was set to three months, this living arrangement was obviously more frequently reported by the respondents of the FFS than of those of the microcensuses.

The consistency criterion of equal numbers of men and women living in partnerships is only approximately achieved. Probably this would require a further differentiation of the married ones whether those live together or separately and likewise a consideration of the fact that there are also married persons in the non-marital partnerships.

Discussion

The living arrangements, household and family structures changed in the last three decades extraordinarily strongly. With low fertility and with increasing life expectancy demographic aging strengthened, the decline of the nuptiality and increase of the divorces reduced quantitatively the meaning of the marriage and the traditional family, although these living arrangements from married persons and their children are still lived by the majority. If one assumes that the current demographic behaviours (leaving the parental home with 23 or 25 years, probability of marriage 60 or 70 per cent, probability of divorce with 30 per cent, on the average 1.33 children) remain, then ever more regenerating cohorts will be shaped by high proportions of unmarried persons, childless and inhabitants of one-person households. Frequently all three features will accompany together.

If one sees the total population, this change in the medium run however is covered of the fact that the older cohorts - also because of the assumed increase of the life expectancy on 80 resp. 86 years - become comparatively weightier. Because of their high marrying proportion this marital status in the entire population will even again increase in the next decades. This might occur also if one differentiates the non married more exactly according to whether they live in consensual unions. In the model calculations it was assumed that the today's behaviours – in demographic terms: the transition probabilities - to enter non-marital partnerships and terminate them in favour of a marriage or a separation remain constant. Therefore most of these partnerships remain a prelude of the marriage, which is lived particularly at younger ages. Scenarios with other assumptions, for instance with a further increase of consensual unions and their institutionalisation as an alternate to the marriage scenarios of course would show other results.

Taking account of the non-marital partnerships also high proportion of single persons in the future population are projected. Clear modifications obviously only could be expected when, if the behaviour concerning marriages and unions or also the household relations change drastically, to more partnerships at higher age years or away from the one-person household.

Within the older population all forms of living together increase quantitatively. With the fact so pleasing that older married couples presumably longer live together and they protect themselves the capacities, to care each other (Dinkel 1996), it should not been covered that the numbers of older non married and widowed persons likewise rise.

At the end - over the above addressed uncertainties beyond the database and the estimated parameters - problematic results and doubt-worth methodical steps are to be listed to become again conscious on the one hand of the dependency on the assumptions met first and to clarify on the other hand the necessity for further work.

The model calculations proceed from the population at the main place of residence. This is mandatory necessary, because otherwise persons with several places (main and further) were counted several times. The consequence is that here again only the numbers of the main places can be projected. It would require separate estimations, which might be rather far from the scenarios of living arrangements, to make projections for the second homes.

For the simplification of the calculation it was assumed that the migration balance is zero and that the migration, which nevertheless can take place does not have influence on the population structure. With all results it is to be considered that they refer only to the resident population of the base year and its descendants.

No differences between Germans and foreigners were considered in the model calculation.

The regional differences of the population were disregarded.. It can be assumed that the demographic development is characterised by clearly lower birth rates and marriage frequency in the new Länder still longer. Because of the difficulties to number the assumptions in detail over the future, procedures of appropriate model calculations at present appear as too difficult.