Thursday, November 30, 2006

Sweden, Pro-Natalism, Baby Booms and Demographic Transitions, Continued

by Edward Hugh

This is a continuation of my last post, since looking again at what I said, and examining the Swedish experience a bit more closely, I think I may have overdone things a little. This point was made clear to me by a comment from Stefan Geens who drew attention to the rather complicated picture which the Swedish case offers us. The core of the problem is to be found in this statement from Mark Ammerman:

"Despite all the effort Sweden has never since then risen over or even near replacement fertility."

Now as Stefan points out, this statement, as it stands is too strong, and undoubtedly an oversimplification. In support of his assertion Stefan points us to an earlier post on his blog which examined the recent evolution of Sweden's population, and he draws attention to the fact that the number of live births in Sweden has been rising steadily since 1999. In fact detailed information on live births and deaths in Sweden can be found on this page.

Now as I point out to Stefan in comments, one of the problems in looking at this data is to get an accurate measure of what it actually means in terms of the issue which concerns us here, replacement rate fertility. This is normally considered to be a fertility rate of 2.1tfr, but it is just here that we find the problem since TFRs may not be an accurate enough measure of completed fertility, and in particular during periods of rapid birth postponement when many women displace having children to ever later ages.

The total fertility rate is a measure of periodic fertility and uses the sum of the number of children born by women in a specific age (between 16-49) during a given time period in relation to the total number of women in the age group to make an estimate of the fertility rate. The cohort fertility rate (CFR) is a much more accurate measure since it is based on the average number of children which women give birth to during their entire lifetime fertility cycle. Now CFR is important since it measures whether the completed fertility for women is above the replacement level or not, but it has the important disadvantage that it becomes available only after women in any given cohort have completed their fertile life, and in periods of significant fertility fluctuation may reflect more what WAS happening, rather than what the current situation is.

In Sweden, for example, the latest available data is for the 1955-1959 cohort, but we may be rather interested in how fertilty behaviour may have changed when this group is compared with the behaviour of those born in the years 1965-69. Since data for this cohort only becomes available around 2015 this means that it is hard to take policy decisions based on CFRs, or, at least, it is possible, but important details may be missed, and hence the fairly widespread use of TFRs to try and get a handle on the process of fertility change.

To cut a long story short the reading on CFRs for Sweden to date have been not far short of replacement level, since they have been around the 2.0 mark. See, for example, this paper:

Why does Sweden have such high fertility?


By current European standards, Sweden has had a relatively high fertility in recent decades. During the 1980s and 1990s, the annual Total Fertility Rate (TFR) for Sweden undulated considerably around a level just under 1.8, which is a bit lower than the corresponding level in France and well above the level in West Germany. (In 2004 the Swedish TFR reached 1.76 on an upward trend.) The Swedish completed Cohort Fertility Rate (CFR) was rather constant at 2 for the cohorts that produced children in the same period; for France it stayed around 2.1 while the West-German CFR was lower and declined regularly to around 1.6.

So if the question we want to answer is: does Sweden currently have replacement (or near replacement) fertility, the only honest answer is that we don't really know. Having said that it does appear that Mark's original statement is a bit too strong, even if the current cohorts do not turn in replacement level fertility, they may well be not too far short of it.

So, in completing my mea culpa, I would make the following points:

1/ The important issue which I wanted to stress was that what we are dealing with here is not a simple boom-bust phenomenon, but rather a steady and continuous decline in fertility, which (if we strip out the US and Israel) seems to have taken place across the entire developed world, and which is now also rapidly occuring in the parts of the third world which are developing economically. In this context many countries which understandably seek to place an emphasis on reducing their fertility to replacement levels will almost certainly find it difficult to stop the decline at that point, and will in all probability move to fertility levels which are well below replacement level.

2/ Pro-natalism is important, but even in the society where most effort has been placed on a systematic pro-natal welfare system sustaining fertility at replacement levels has not proved easy (if indeed it has in fact been achieved).

3/ The quantity of resources necessary for the kind of policy Sweden has pursued are significant, and whilst they may be feasible in a comparatively rich society like Sweden, with a high level of awareness of the issues involved, introducing such a policy and devoting the quantity of resources required in less economically affluent societies, and in particular in the context of the situation that below replacement fertility and rapid increases in life expectancy both tend to occur simultaneously, will not be easy. Simple lump-sum cash payments - which of course are not unwelcome - are unlikely to make a substantial impact and in particular not in those societies which have already fallen below the 1.5tfr level.

4/ The Swedish (and Scandinavian in general) case also draws our attention to the critical difference between those societies with "near replacement" fertility (where pyramid sustainability is not resolved, but at least the relatively higher fertility makes the transition easier) and those who have fallen into the lowest-low fertility category. The need to avoid the critical drop below a TFR of 1.5 cannot be over-emphasised. Many societies already find themselves in this uncomfortable position (and the list of these is a lot longer than many imagine, including most of Souther Europe, Eastern Europe, the Asian Tigers, Japan and now possibly China), but it is important to try and prevent others meeting the same fate, which is why there is an urgent need in the developing third world to draw attention to the urgency of putting pro-natalist policies in place as these societies approach replacement level. The lesson of Southern Europe and the Tigers seems to be that if you do not do this you will almost certainly overshoot and find yourself in the lowest-low category. Bottom line: in trying to raise this awareness the Swedish model is, in fact, not without some importance.

11 comments:

Mark Amerman said...

Stefan (and Edward),

I'm a bit disappointed with your responses (although at the same
time the misreading is intriguing for what it suggests about how
the human mind works).

Both of you are really bright guys and very decent people. I've
been reading Stefan Geens blog for years including occasional
forays into his commentary on Google Earth and I do an even job
of tracking what Edward is up to in his many blogs.

In my comment to Edward that he responsed to (and thanks for the
response, Edward) I started off by saying,

"Sweden seems a good empical test of what laws favoring mothers
can do."

That was the subject.

What kind of impact does a whole slew of different laws giving
assistance to mothers have on fertility. Sweden is an unique
example. For seventy years, they've been doing this, and in
a big way.

Lots of countries are now starting to do this, although not to
the same expense as Sweden. So the obvious and natural question
is what kind of fertility boost should we expect to see from
all this? Sweden gives an answer. Or it should be able to.

Obviously we do not want to include immigrant births in trying
to figure out the fertility boost. Or it should be obvious.
It's easy to import fertility (or at least it will be for a
while.)

A few years ago I had neighbor, an immigrant from El Salvador,
and in the space of nine years, she had four children. There
is nothing unusual about her situation. Most immigrants are
relatively young and at the right age to have children. Plus
they are coming from cultures that don't have Social Security
and value children highly.

I'd be amazed if it weren't the same situation in Sweden.

So to evaluate the long-term effect of all these state incentives
for women to have children on the birth rate we need to tease
apart the births from immigrants from those from native ethnic
swedes. Otherwise what we're measuring is to a considerable
extent the fertility of immigrants which would be high regardless
of the incentives.

The Swedish government refuses to collect this data.

The ethnic swede fertility might be estimated by estimating
the immigrant birthrate and subtracting from the government's
statistics or another approach might be to look at the political
subdivisions of Sweden that are still almost entirely ethnic
swede and looking at their fertility.

We all know it's going to be below replacement and in fact looking
at Stefan's graphs and thinking about the immigrant birthrate I see
in my community, you know, it's not implausible that Sweden's
ethnic fertility rate might be down around Italy's.

Which would be amazing and very discouraging. If all these
efforts to promote births through state assistance don't work,
what are we going to do?

Pål Sletten said...

"We all know it's going to be below replacement and in fact looking at Stefan's graphs and thinking about the immigrant birthrate I see in my community, you know, it's not implausible that Sweden's ethnic fertility rate might be down around Italy's."

I doubt this very much, although I haven't seen the Swedish figures. However, Norway is in a very similar situation, and here is a comparison of TFRs for immigrant and non-immigrant women in Norway in 1997-1998:

http://www.ssb.no/vis/english/magazine/art-2002-09-06-02-en.html

As can be seen, the Norwegian TFR does not depend on whether or not immigrants are included. I think the share of immigrants in the total population is slightly higher in Sweden, but I would not think it is enough to make a difference.

As to the discussion on the impact of policies, a couple of articles (in English) can be found in the 2004 issue of Demographic Research (here: http://www.demographic-research.org/volumes/vol10/default.htm) - perhaps the article "Fertility and Public Policies - Evidence from Norway and Finland" is of interest?

Pål Sletten said...

And the same canards is refuted here, for the case of France:

Cinq idées reçues sur l'immigration
http://www.ined.fr/fr/ressources_documentation/publications/pop_soc/bdd/publication/482/

"Dans la période 1991-1998, le nombre moyen d’enfants par femme était de 1,72 en France métropolitaine pour l’ensemble des femmes, et de 1,65 pour les seules françaises natives." As pointed out, the immigrant population is simply to small to have a large impact.

I have not seen any such figures for the USA, would they be any different?

Edward said...

"I doubt this very much"

Yes Pal, I'm afraid I have to go with you here. It all depends on fertility in the country of origin of the migrants. For Sweden (we do have numbers knocking around somewhere but I can't put my hands on them right now) a large part of the migration is from Eastern Europe and in particular from the ex-Yugoslavia where fertility is all near the lowest-low level already.

"And the same canards is refuted here, for the case of France"

Again ditto. Since France has received a lot of migrants from the Maghreb, and since fertility there is all rapidly closing on the reproduction level, the same picture operates.

Spain is getting a lift in fertility precisely because many migrants come from Latin America where fertility is still comparatively high (especially Ecuador, Peru, Bolivia).

"I have not seen any such figures for the USA, would they be any different"

Oh very much so, maybe you should look at this post here. US fertility is different *in part* since they have had a large number of migrants over recent years (around 11 million illegals) who have come from countries of comparatively high fertility. So - givenm that the US doesn't have the kind of pro-natality programme in place which we are talking about in Sweden, they are vulnerable on fertility should the Latinos ever seriously get into birth postponement.

I worry about this, but I think you will find that relatively few people inside the US share my concern, although it should be said that Mark (more or less) does.

Edward said...

Here is the data I had on the composition of the Swedish migrant population:

As of 2004, 13.3% of the population was foreign-born.

Rest of Scandinavia 277,100
Western Europe 106,400
Eastern Europe 326,100
Former Soviet Union 37,800
Middle East 223,700
Rest of Asia 91,700
Sub-Saharan Africa 51,000
North America 26,500
South America 55,500
Oceania 3,500
Total Foreign-Born 1,199,300
Data as of December 2004

As Pal suggests the size of the migrant population from high fertility regions is not large enough to make an important statistical difference.

Pål Sletten said...

And, as pointed out in the article from Statistics Norway, the fertility rates of immigrants converges with the rest of the population over time. (I.e., the longer immigrants stay in Norway, the more their fertility rates are similar to that of the non-immigrant population.) The same holds true for France.

Incidentally, the fertility rate of women immigrating to Norway from Western Europe is 2.0, which is higher than what is found in most countries in Wester Europe - in other words, the TFR increases when moving to Norway. I guess the most important explanation is that the reason to immigrate to Norway from Western Europe is in many cases marriage. I don't know whether it is also an answer to Mark's original question: "a good empical test of what laws favoring mothers can do."

Mark Amerman said...

Pal Sletten,

Thanks for your reference to:

http://www.ssb.no/vis/english/magazine/art-2002-09-06-02-en.html

Reading it I'm amazed to discover that according to the "Statistics
Norway" magazine immigrants have had no perceptible impact on
norwegian fertility. This would seem different from the U.S.
experience.

In my local neighborhood for instance the elementary school has
doubled in size in the last ten years while the percentage of kids
with latin american parents has risen from maybe 15% to over 70%.

Of that's just an anecdote, here's some collected data:

http://www.demography.state.mn.us/PopNotes/Fertility/FertilityPopnotes.pdf

Note that in Minnesota, foreign-born mothers account for 14% of the
births in the year 2000.

And according to this report (in Minnesota, in the year 2000):

white TFR is 1.96
latino TFR is 2.94

but that comparison is strange because the latino births are part of the
white TFR (with the exception of a small percentage of latinos who
are counted as african-american).

To make sense of the numbers above and the demographic impact of
immigrants we need to know the percent of women aged 15 to 44 that
are foreign-born, and also the percent that are latino and white.
Unfortunately this Minnesota state report doesn't give that information.

But searching quickly, I find:

http://quickfacts.census.gov/qfd/states/27000.html

and see that in 2000 for minnesota:

foreign-born persons were 5.3% of the population
non-latino whites were 86.4% of the population
latinos were 3.5% of the poplulation

(Ignoring african-american latinos and assuming perhaps unreasonably that
that foreign-born women ages 15 to 44 are also 5.3% of then women of that age
overall.)

This is a pretty good comparison with Norway. The overall
percentage of immigrants is roughly similar and although the TFRs
are different the relative differences are similar:

1.8/2.4 native versus immigrant TFR for Norway
~1.9/~2.7 native versus immigrant TFR for Minnesota (quick estimates based on above)

If I can figure out how to do the math, I imagine the immigrant
TFR is going to barely budge the overall TFR. Just like Norway.

Strange result really. From the perspective of TFRs, immigrants
apparently don't have much impact, but to a parent in Minnesota
3 out of 20 kids are immigrant while only 1 in 20 of the population
overall is immigrant.

Hmmm...

One other thing that I wonder about is this extrapolation of TRFs.
According to the Minnesota report, the TFRs are calculated on the
assumption is made that women who have not yet passed the age of
childbirth will continue to have children in exactly the same
percentages as the different age groups are now.

I'm not sure but it's possible this means immigrant birth-rates
have been projected in to the future on the assumption that they
will mirror native-born mother birthrates. If so that could be
a dubious assumption.

Pål Sletten said...

This is not really my field, but here is some speculation:

"If I can figure out how to do the math, I imagine the immigrant TFR is going to barely budge the overall TFR."

I think the important thing is the relative size of the immigrant and non-immigrant population. In the case of Norway, there were 857,725 non-immigrant women aged 15-44 in 1997/1998 against 61,682 immigrant women. Even if the TFR of immigrant women was double that of non-immigrant women, the average of the two groups would be fairly close to that of the non-immigrants.

However, the way I understand it, this does not mean that immigrants do not have much of an impact. ("Strange result really. From the perspective of TFRs, immigrants apparently don't have much impact, but to a parent in Minnesota 3 out of 20 kids are immigrant while only 1 in 20 of the population
overall is immigrant.") Immigrants are younger, and therefore have more children. (Many non-immigrant Norwegians are old, not that many among the immigrants. Maybe this is different in the USA, where immigration is an older phenomenon?) For this reason, the size of the immigrant population (defined as persons who immigrated and their direct descendants) is projected to grow in Norway, even in the absence of new immigration.

On the last point: "I'm not sure but it's possible this means immigrant birth-rates have been projected in to the future on the assumption that they
will mirror native-born mother birthrates. If so that could be a dubious assumption" I also think that would be dubious, and I am fairly certain this is not the way it is done by Statistics Norway. The projection is done using birthrates specific to each sub-population.

I think the point I wanted to make with the first two comments about Norway and France is only that their TFRs is not influenced a lot by immigration. Hence, their level and evolution must be explained by other factors. Of course, immigration by itself has a direct impact on the demography of Norway (simply by adding more young people), but not on the TFR we currently observe.

One question: The Minnesota data was interesting, and I was very surprised to see the large differences by race - as I understand the table on page 3, it was 2.99 for African American women, and 3.23 for Asian/Pacific Islander women, against only 1.84 for White women in 1990. (Incidentally, the Norwegian TFR for 2005 was also 1.84.) Why these large differences, and why are the levels so high for some groups? Are immigrants included?

Edward said...

"One other thing that I wonder about is this extrapolation of TRFs.
According to the Minnesota report, the TFRs are calculated on the
assumption is made that women who have not yet passed the age of
childbirth will continue to have children in exactly the same
percentages as the different age groups are now."

Yes, quite. Tfrs are an estimate. I hope the post which follows this one makes this situation a bit clearer.

Mark Amerman said...

I decided today to make one of those calculations I skipped yesterday.

The point of all this to separate out the impact of immigration on
total fertility rates in Minnesota. It gets kind of complicated.

From the two sources I mentioned yesterday

http://www.demography.state.mn.us/PopNotes/Fertility/FertilityPopnotes.pdf
http://quickfacts.census.gov/qfd/states/27000.html

Here are the basic numbers for Minnesota

total fertility rates by group
------------------------------
1.97 for all the groups together
1.945 white
2.725 black
2.33 american indian
2.29 asian/pacific islander
2.94 latino

The Minnesota state fertility report actually gives two TFR
numbers for each group, a high and low estimate, with very little
explanation of how they were calculated. To simplify I took
the average of these.

Now latinos need to be considered separately because they
are not a disjoint set like the others and instead are
included in the white and black groups. So to makes since of
this mathematically we have to figure in the data on that split.

latino percentages of minnesota
-------------------------------
3.5 latino
3.2 white latino
0.3 black latino
86.7 non-latino white
3.8 non-latino black

and then the racial composition, percentages
--------------------------------------------
89.9 white
4.1 black
1.2 american indian
3.4 asian
0.1 native hawaiian
1.4 ambiguous


So instead of percentages imagine we have one hundred young women
in strict proportion to their percentages. Multiply each group
by its TFR, and,

89.9(1.945) + 4.1(2.725) + 1.2 (2.33) + 3.5 (2.29) = 1.968

now a quick check of multiplying the non-ambiguous groups by the
overall TFR

98.6 (1.97) = 194.2

which should be the same number but isn't quite. The discrepancy
is probably due to round-off error plus more importantly that
earlier step where high and low TFR projections were averaged.

Now

86.7 (x) + 3.2 (2.94) = 89.9 (1.945)

Where x is the unknown TFR of non-latino whites, 86.7 the percentage
of non-latino whites, etcetera. Solving for x:

x = 1.91

Check. And that's not that much different from 1.97. So a small
immigrant group with high TFRs has only a modest impact on overall
TFR even though that small group is rapidly become a major percentage
of the future population because of the disparity in TFRs.

Mark Amerman said...

Making these calculations it becomes clearer and clearer that
small differences in the TFR have enormous impact on demography.
The difference between 1.8 TFR and 2.1 TFR is in many contexts
just overwhelming.

For example what would it take, assuming we have a population
with 1.8 TFR, to raise it's fertility to replacement (in other
words 2.1 TFR) by immigration?

Suppose 5% of the population is immigrant, what would their
fertility need to be?

100 (2.1) = 210

95(1.8) + 5(x) = 210

solving for x

x = 7.8 TFR

Wow! An average 7.8 children per immigrant woman!
Just to go from 1.8 to 2.1.

Now try a more realistic calculation. Restrict ourselves
a high real world TFR, say from central america:

2.95 TFR

What percentage of a population would need to be
immigrant with this level of fertility to raise the total
population up to replacement rate?

26 percent.