Sunday, May 13, 2007

The Fertility Trap Hypothesis ... Revisited

As the postings, and most notably and impressively the commenting, here at DM are once again ticking along nicely, I thought that I would revisit an oldie but goldie topic, namely the concept of the fertility trap. Waaay back in 2005 :) Edward offered us a very nice summary of the fertility trap hypothesis as it has been conceptualized by the Austrian demographer Wolfgang Lutz. Back then Edward ended his post asking whether new evidence of the hypothesis would be found in the Eurobarometer survey which was due in 2006. Now it seems that we are ready answer that question as Wolfgang Lutz and co-authors inform us in a paper given at this year's PAA 2007. What follows is a cross post from Alpha.Sources of this new contribution together with yet another account of the hypothesis itself;


In order to situate the idea of the fertility trap we need to think first about the demographic transition. As is well known that one of the consequences of the demographic transition is a drop in fertility rates as a country develops economically and in particular the drop below replacement as the country enters the final stages of the transition.

What was not so obvious to those who originally formulated the transition thesis was the nature and extent of this decline in fertility, and crucially, as we can see now see clearly, the fact that fertility rates do not stabilize at replacement level (2.1). This brings us to the core of the The Fertility Trap Hypothesis. The hypothesis does not seek to explain why and how far fertility might potentially drop below replacement levels but rather attempts to outline the mechanisms which may serve to keep fertility rates in a perpetual low state once they reach a certain lower limit (1.5). As such the hypothesis tries to demonstrate how some countries are at risk of entering a trap, as it were, which perpetually depresses their population momentum. My impetus for this entry comes from the recent PAA2007 meeting where a slew of papers and articles on demographics were reviewed, among them an article by Wolfgang Lutz, Vegard Skirbekk and Maria Rita Testa entitled New Empirical Evidence on The Low Fertility Trap Hypothesis (PDF - extended abstract). Reading through the abstract we also get the theoretical foundation (elaborated more in this paper) for the fertility trap hypothesis which as outlined consists of three mechanisms;

(i) a demographic mechanism (LFT1),
(ii) a sociological mechanism (LFT2)
(iii) an economic mechanism (LFT3).

These mechanisms are of course explored in great detail in the sources presented in this entry (see list of references below) and I will then only briefly summarize them here.
  • The first mechanism of the fertility trap hypothesis is described as the demographic mechanism and essentially it refers to the potential effect of a situation where fertility drops consistently below 1.5 live births per woman per year. In essence, the demographic mechanism shows us why y-o-y fluctuations in TFR really mean very little in the general perspective and that a much broader perspective is needed. The demographic mechanism works through the impact of negative population momentum. The main point is that the age distribution of the population acts as an additional variable (i.e. independent) to influence fertility apart from actual realized fertility in that period. As such this mechanism acts independently from current fertility on the basis of past fertility, mortality and migration (Lutz et al 2005). What we consequently need to think about here is that a period of very low fertility exerts a structural damage on the pyramid which affects future fertility from a relative point of view to bring future absolute numbers of live births down even if TFR suddenly (and hypothetically) jumps to replacement levels (Lutz et al. 2005).
  • The second mechanism is derived from sociology and is based on the concept of personal ideal family size which is said to exert influence on total cohort fertility (Lutz et al. 2005). In terms of the actual of ideal family size it is said to lag actual fertility by some margin but more importantly, empirical data suggests that ideal family size is declining in several European countries. In terms of the formal mechanism the ideal family size today is said to be a social norm as a function of previous generations' ideal family size. As such, the process of adapting social norms and learning from generation to generation might work in such a way to accentuate a process of fertility decline because the decrease in ideal family size is socialized from one generation to another (Lutz et al 2005). Of course this could also be somewhat tied together with Becker and Barro's idea of the quantum effect of fertility decline essentially derived from the realms of economics but I will leave that question aside here.
  • The third and last mechanism is derived from economics and more specifically from Easterlin's relative income hypothesis. In essence, Easterlin's relative income hypothesis states that it is not the absolute expected income which is important but rather income relative to expectations formed in youth. The conceptualization of this mechanism thus assumes that fertility is dependent on the relationship (gap) between by aspirations of income and expected income and as the gap increases we will see both a increase in both the quantum and tempo effect of fertility.

So, in terms of a general introduction of the fertility trap hypothesis this is pretty much what you need to have in your head at a first glance. Moving on to the actual operationalization of these mechanisms we of course need to be more specific. In essence, we could also say that the first mechanism (the demographic one) in itself represents the fertility trap and its dynamics whereby the two other mechanisms become explanatory variables in terms of what potentially may sustain and exacerbate this negative demographic momentum. Another point is of course the idea of a trap below a TFR of 1.5 which is crucial for the understanding the idea of this theory. As such, the threshold of a TFR of 1.5 represents the level where anything below risks causing a cumulative acceleration of the three mechanisms noted which then makes it exactly a trap from which recovery will be very difficult.

Turning now to the empirical foundations for this hypothesis brings us to the recent contribution presented at the PAA 2007. So let's take a look at the recent data presented in the paper abstract from Lutz, Skirbekk, and Testa at the PAA 2007. Specifically, we have the very recent evolution of ideal personal family size which is shown in the two figures below based on data (see the abstract) from Southern Europe and France. As we can see the decline might look ever so slight but once we get the figure in percentage form it shows that the decline in some countries is pretty hefty given the fact that we are only talking about a period of 5 years. Moreover, note that Italy's personal ideal family size is fast approaching sub-replacement levels and seeing that this is a lagging factor of actual fertility decline the evidence for a traps seems particular sinister here. Of course the drop in Spain's ideal family size of about 9% is also something to watch.

Regarding the data on Easterlin's relative income hypothesis I will refer to the tables and graphs in Lutz et al's extended abstract. The stylised facts are clear, income for women in their peak child bearing age have declined relative to older age groups which may serve to exacerbate the tempo effect of fertility decline as it materializes in terms of increased birth postponement (Lutz et al 2007).

Summing up, I see this post as a kind of intro to the fertility trap hypothesis as it has been conceptualized by Wolfgang Lutz, which is in part based on an elaboration of Peter Macdonald 's 2005 - Fertility and the State: the efficacy of policy. And concluding with some general questions and remarks I would like to note the following:

To what extent is the idea of a fertility trap applicable in a universal sense across socities in general? Traditionally, the idea of a fertility trap has been conceptualized in connection with southern European countries. Having said that the mechanisms on which the fertility trap hypothesis rests seem to be universally valid to be examined across many societal environments.

If the fertility trap really exists, how might a country escape? What follows from this is of course the immediate point that the extent to which the fertility trap hypothesis is true paints a very grim outlook for population dynamics in some countries with no imminent way out of the misery. Of course, a positive demographic shock through immigration as experienced in Spain for example might postpone the effects but given the first mechanism noted above merely postpone the inevitable given the cumulative effects of fertility decline.

Lastly, are the mechanisms valid? In terms of the first one there is little to be argued I think since this is essentially pure arithmetic and as such this effect also occurs at fertility levels above 1.5 but may only exhibit a particular 'trap-like' behavior once fertility is sustained below 1.5. As for the other mechanisms they are of course hypotheses in terms of explanatory variables of total actual fertility and may hold different explanatory power pending on the country of question.

List of References

Lutz, W., V. Skirbekk, M. R. Testa forthcoming (2005) "The Low Fertility Trap Hypothesis"
Vienna yearbook of population research. Vienna: Austrian Academy of Sciences Press.

Lutz, W. and V. Skirbekk. (2005). "Policies Addressing the Tempo Effect in Low Fertility Countries", Population and Development Review. December: 703-725.

McDonald, P. (2006). “Low Fertility and the State: The Efficacy of Policy”, Population and Development Review. 32, no. 3 (Sep 06): 485–510

Testa, M. R., and L. Grilli. (2006). The influence of childbearing regional contexts on ideal family size in Europe. Population 61(1-2), 109-138.

Lutz, W., V. Skirbekk, M. R. Testa (2007) New Empirical Evidence on The Low Fertility Trap Hypothesis PAA 2007


Anonymous said...

Can you reject the hypothesis that the results from the two surveys are equal?

Similarly, are the questionnaires identical, or might there be a systematic bias between the two?

CV said...

Hi ...

You mean between the 2001 and 2006 surveys?

As far as I know they were both Eurobarometer which uses solid and consistent methdology.

Anonymous said...

I think anon is asking about statistical significance. For that we'd need to know more than just that graph, namely the number of people surveyed and the spread of their answers.

@cv: Are these data available on the Europa website somewhere?

Edward Hugh said...

Wow, you guys seem game for some really heavy duty reading. General information on the Eurobarometer series can be found here. Details on the standard Eurobarometer are given here, while the EU's more user friendly presentation of the findings is available here. The Eurobarometer summary on Chidbearing Preferences and Family Issues in Europe can be downloaded here. And guess who the author of the report presentation: Maria Rita Testa (co-author of the LFT hypothesis, and close collaborator of Wolfgang Lutz). Small world isn't it.

I imagine these survey are pretty rigourous by normal scientific standards, but if you rummage about a bit I am sure you can find the details on methodology you are looking for, although perhaps any observations about lapses in methodology would be better directed towards the statisticians at Eurostat than towards we humble bloggers here :). In this regard we have the data we have.

CV said...

@Colin ...

I see your point.

The graphs in this post is merely a reproduction of the numbers in Lutz et al's extended abstract. Edward also beat me to it in terms of the actual source data I see :).

Edward Hugh said...

Anon and Colin,

"Similarly, are the questionnaires identical"

"I think anon is asking about statistical significance."

Well, having to some extent poured cold water on your initial reservations - I do think the surveys conform to some sort of general notion of statistical rigour - I do think there is a much bigger issue knocking around somewhere about the interpretation to put on this kind of survey.

What do the answers people give really mean? Are they a reflection of what they think would be generally good for society, or do they reflect their own individual intentions (intentions which, of course, we may find it hard to live up to).

Some grist to my mill here is to be found in the most recent German results. As Rita Testa observes:

"A question on personal ideal family size was also asked in the Eurobarometer 2001 round (standard Eurobarometer 56.2), by using exactly the same wording. A comparison of Eurobarometer 2001 and 2006 shows that the mean ideal family size has increased for both sexes and in all age groups in Germany. This is true both of the country as a whole, and of the East and West taken individually. The increase is evident when we compare the same age groups in the two different surveys (Table 6), but also if we do a cohort analysis - i.e., we follow people born in the same years and compare their ideals as revealed in the 2001 and 2006 surveys (Table 7)."

(page 12 in the above link)

Now.... does this mean that since 2001 people really are keen on having more children, or does it mean that after all the publicity and debate people are aware that it would be "ideal" if everyone in Germany had two children? All Rita Testa says is:

"No other country within the EU15 has experienced such a change in the interwave period between 2001 and 2006, and this finding needs to be further investigate in a more in depth analysis."

Which seems to me like a standard get out clause. Who is going to do this analysis, and what exactly would it consist of?

Interestingly, while everyone here in Spain is aware of the existence of low fertility as an issue, people seem much more prepared to accept immigration as a practical short term solution (I mean everyone is being told everywhere - eg UN median forecast - that fertility will eventually "rebound" so this idea has presumably also sunk in) than they are in Germany. In other words could all this be interpreted as meaning that German society is more aware of the fact that if they don't want immigrants then they really do need to have more children.

But moving from this recognition to reality is a very large leap, as the ongoing fertility numbers show. Claus - in a mail - pointed this out to me:

"Now I don't know about 05 and 06 but look at the fertility rates in Germany since 1996 (i.e. 1996-2004) ... pretty much stable at 1.3-1.4"

1.32, 1.37, 1.36, 1.36, 1.38, 1.35, 1.31, 1.34, 1.36, 1.32, 1.37, 1.36, 1.36, 1.38, 1.35, 1.31, 1.34, 1.36

"essentially ... there seems to be no real effect from 2001+ ..."

I think we need to distinuish here between two things:

1) The possibility that social/normative mechanisms exist whereby the presence of an environment where few children are actually born may mean that young people form ever lower childbearing expectations.

ii) the viability of surveys like the eurobarometer one as a tool for measuring (1).

The former seems emminently possible, but the latter raises important difficulties as an assessment tool.

Basically what we have here is a large methodological and interpretative problem, and my feeling is that Lutz may be attaching rather too much importance to Eurobarometer, but this does not mean that Eurobarometer-type findings are not interesting as a guide, but rather that they can't be taken as firm evidence for anything. OTOH, they would seem to be all we (or anyone else) have.

Anonymous said...

What Colin said. Perusing the sources Edward gave, it seems that the differences between the two surveys are within the margin of error. Admittedly, the fact that all the differing national samples vary downwards is suggestive, but it would be a mistake to draw any conclusions about trends from them.