My 18-month detour through second-degree price discrimination terminology

When I was composing this post about price discrimination last month, I was drawn into a discussion with ChatGPT about second-degree price discrimination. ChatGPT, which I mostly use for checking for inconsistencies and grammatical errors in my draft blog posts, told me that I should refer to menu pricing as a form of second-degree price discrimination. I replied that wasn't correct, because second-degree price discrimination, as defined by Arthur Pigou in the early 1920s, involves offering a declining price for each additional unit that the consumer buys. ChatGPT responded that indeed, Pigou had defined second-degree price discrimination that way, but that in much current industrial organisation usage, second-degree price discrimination includes cases where consumers are offered different options and sort themselves into groups that have different price elasticities of demand (or different willingness to pay) for the good.

That discussion made clear that I had been on an 18-month detour in how I described the degrees of price discrimination. Only last year, I changed the definitions of the degrees of price discrimination in my ECONS101 class to match those that Pigou uses, and therefore moved menu pricing into the definition of third-degree price discrimination (or group pricing). I've held off on posting about my exchange with ChatGPT until now, because I didn't want to confuse my students in this trimester's class about what did, and did not, fall under the different degrees of price discrimination before they were tested on it (and, as it turns out, I didn't test them on that specific aspect of the topic in any case). [*]

This appears to be one of those situations where terminology changes meaning over time, and is a cautionary lesson in making sudden changes to definitions on the basis of reading about the history of economic thought. The issue here is that I had come across Pigou's definitions in one source, and initially dismissed it as it was inconsistent with the way we taught that topic. But then I read The Economics Book by Niall Kishtainy and co-authors (which I reviewed here), which made me more certain about Pigou's definitions. To be clear, I'm not blaming Kishtainy et al. They were perfectly correct in terms of Pigou's definitions. I should have checked some other sources for more current usage. One example is the excellent book Information Rules, by Carl Shapiro and Hal Varian (which I read in 2023 and reviewed here), which made the definitions used in industrial organisation clear (although Shapiro and Varian preferred to use the term 'versioning', rather than second-degree price discrimination).

Now I'm left with the task of combing through my past posts, to ensure that I update my terminology, or revert it to the original text in the few cases where I went back and made changes. I don't want to risk confusing future students, which is a risk given that I refer them to my posts for further detail and examples on topics that we discuss in class.

*****

[*] I didn't perfectly achieve this goal, because one very alert student picked up the error through her own conversations with Harriet, our ECONS101 AI tutor, an irony that was not lost on me.

Exposure to conflict and support for democracy in Africa

Does being exposed to conflict, whether violent conflict or protests, increase or decrease people's support for democracy? That is the question that is addressed in this new article by Nicole Stoelinga (Max Planck Institute for Behavioral Economics) and Tuuli Tähtinen (University of Munich), published in the journal World Development (open access).

Stoelinga and Tähtinen have a quite innovative approach to identifying the answer to that question. They exploit the timing of data collection from the Afrobarometer survey, comparing research participants who were surveyed in the two weeks before a conflict event in their region with research participants who were surveyed in the two weeks after the event, in terms of their support for democracy and perceptions of the extent of democracy in their country. Controlling for other characteristics of the research participants, in theory any difference in support for democracy or perceived extent of democracy between those surveyed before, and those surveyed after, the conflict event should arise as a result of the conflict. This of course relies on survey timing around the conflict event being 'as good as random' (more on that later).

Their data covers up to 38 countries (the exact number varies across waves of the Afrobarometer survey) over the period from 2002 to 2021. Their conflict data comes from ACLED, and distinguishes between 'violent conflict events', which include "battles, explosions/remote violence, and violence against civilians", and 'demonstrations', which include protests and riots. Their dataset includes over 24,000 observations for violent conflict events, and over 42,000 observations for demonstrations. In their main results, they find that:

...conflict has a positive impact on support for democracy. The point estimates indicate that on average, a conflict event in one’s region increases the probability of support for democracy over other forms of governance by 2 percentage points. Exposure to violent events and exposure to protests have very similar impacts. In contrast, individuals’ perceptions of governance in their own country are not significantly affected by conflict exposure.

That seems to me to be a quite surprising result - exposure to conflict increases the support for democracy. Stoelinga and Tähtinen then dig a little deeper, looking at the difference in results between members of an ethnic in-group (which is an ethnic group that is 'represented in government', meaning that it governs alone or shares power), and members of an ethnic out-group. In this analysis, they find that:

The point estimates are imprecisely estimated, but suggest that exposure to violence increases support for democracy among out-group members, while the point estimate of similar magnitude and opposite sign suggests no effect among in-group members.

When they say that the estimates are 'imprecisely estimated', they mean that they are statistically insignificant. However, in terms of perceptions of democracy in the country, Stoelinga and Tähtinen find that:

...exposure to violence has a significant positive impact on individuals’ perceptions of governance in their own country.

That result holds for both the in-group and the out-group. Demonstrations have no effect on perceptions of democracy for either group. Stoelinga and Tähtinen interpret these results as showing a 'rally around the flag' effect, where people exposed to violent conflict show greater support for their leaders. Stoelinga and Tähtinen provide further evidence in support of this 'rally around the flag' effect by looking at trust in institutions, finding that:

...exposure to violent events significantly increases trust in president and ruling party, as well as in the police and army—but significantly less for the in-group than for the out-group.

Stoelinga and Tähtinen then turn to looking at differences in results between more autocratic and more democratic regimes, finding that:

In autocracies, exposure to either violence or protests increases both support for democracy and the perceived extent of democracy. Conversely, in democratic countries, conflict exposure does not have a significant impact on democratic preferences, although the point estimates suggest that the perceived extent of democracy is negatively affected...

In autocracies, exposure to violence generally increases trust in key political institutions, particularly in president and ruling party, as well as in the army... In democracies, the effects of conflict exposure on trust are less pronounced. Although mostly lacking statistical significance, the estimates point towards both violence and protests having negative effects on trust.

In other words, the 'rally around the flag' effect is a feature of autocracies and not of democracies. That should worry advocates for democracy who might hope that popular movements, which may be characterised by protests or riots, and may occasionally escalate into violent conflict events, might lead to regime change. Instead, the results in this paper suggest that conflict events increase support for democracy, but also increase trust in autocratic governments, and so may be counterproductive. Stoelinga and Tähtinen suggest that this may explain the lack of democratic transitions in Africa. To that, I would add that a similar mechanism may also explain Tunisia slipping back towards authoritarianism in recent years, despite the Arab Spring uprising.

Now, it should be noted that these results could be sensitive to the research design. If conflict events change the way that data collection activities are conducted, or change the locations of data collection, then that would confound the effects of conflict on support for democracy. For instance, if the Afrobarometer survey was originally scheduled for Region A, but that region is affected by conflict and instead they shift to Region B, which is less affected, that might bias the results in exactly the direction that Stoelinga and Tähtinen observe - towards more positive views towards democracy, if those who are less affected by conflict have greater support for democracy. Stoelinga and Tähtinen are upfront about the potential invalidity of their results arising from endogenous change in the timing or location of survey data collection. However, they argue that, because there are no observable differences in research participants before and after the conflict event, there is no problem. That doesn't address whether there are unobservable differences between research participants before and after the conflict event though.

Another issue with the research design is that conflict events don't suddenly arise out of nowhere, so the 'control group' of research participants surveyed before the conflict event may already have experienced some of the antecedents to the conflict event, such as growing tensions and uncertainty. That would have an ambiguous impact on the results of this study.

So, we should not treat these results as the last word on this research question. The paper provides a surprising and counterintuitive finding, that conflict may increase support for democracy, but in autocracies it may also strengthen trust in government institutions. That complicates the appealing idea that protest or conflict will naturally push autocratic countries in Africa towards democratic transition.

The modest impact of Australia's baby bonus on fertility timing

The challenge of returning low-fertility countries to a higher-fertility state has become especially clear in recent years. As noted in this post, aside from the post-WWII Baby Boom, there have been no significant episodes of increasing fertility. And that's not for want of trying. Some governments have become increasingly generous over time in their attempts to encourage higher fertility. Others have flailed around looking for a solution. One notable example is the Australian 'baby bonus', which initially paid each mother a lump sum of $3000 for each child born after 30 June 2004. The amount was increased to $4000 in July 2006, then to $5000 in July 2008, before being reduced to $3000, and eventually removed (and replaced with changes to the Family Tax Benefit) in March 2014.

How (un)successful have policies like Australia's baby bonus been? This new article by Sarah Sinclair (RMIT University) and co-authors, published in the journal Economic Modelling (open access), takes an unusual approach to answering that question. Rather than identifying policies and then testing directly for whether fertility changes happened at those points in time, Sinclair et al. first use time series models to identify structural breaks in the time series of fertility for 31 maternal-age-by-birth-order series. A structural break occurs when the time trend for the series changes meaningfully at a particular point in time. Using their approach, Sinclair et al. look for points in time where many of the time series have meaningful changes. What they find is not much of anything, with:

...the clearest and most consistently identified turning points for second births, with breaks in 2005 and 2015 detected across dates that plausibly align with major changes in family transfer settings. Other shifts, such as in selected age groups and some higher-order births, are less robust, and we detect no structural break in the aggregate fertility rate.

The timing of the 2005 and 2015 changes is consistent with the timing of the major changes to the baby bonus (or, at least, consistent with nine months after the major changes to the baby bonus). However, notice that they found effects only for second births, and not for births overall (or the aggregate fertility rate). That suggests, as they conclude, that the baby bonus affected the tempo of fertility, but not fertility overall. In other words, women brought forward the birth of a second child as a result of the baby bonus, but did not have more children overall.

Of course, identifying structural breaks is not the same as estimating a causal policy effect, but the timing of the breaks provides suggestive evidence about whether policy changes may have mattered. However, one aspect of this paper in particular is kind of unusual. When Sinclair et al. outline the 2005 and 2015 structural breaks in second births, their results are shown in Figure 5 in the paper:

The grey line tracks the second birth rate each month, while the red line shows the overall trend. Notice that in the top panel of the figure, there is a clear change in the trend in 2005. The pre-2005 trend is downwards, and then there is a big jump upwards in the second birth rate in 2005, before it returns to its previous downward trend. The oddity occurs in the lower panel of the figure, where Sinclair et al. show a somewhat less downward sloping trend in the second birth rate up to 2015, before there is a big jump up, and then a much steeper decline. The second figure ignores that there was already a structural break in 2005, where the trend jumped upwards. It seems to suggest that the end of the baby bonus induced a big increase in second birth rate. Now, that could be true, if couples anticipated the removal of the baby bonus, and tried to have a baby before the bonus was removed. However, Sinclair et al. don't really discuss this.

A more interesting interpretation occurs if you squint at the top panel of Figure 5, and imagine one structural break in 2005, and then a second at 2015. In between those two years, the trend in the second birth rate might be mildly upwards. Of course, we don't know this for sure, as Sinclair et al. didn't test for multiple breaks in their time series. But perhaps their results overstate the case against the fertility impacts of the baby bonus. To be clear, these would still be impacts on the tempo of fertility, not on total fertility, but perhaps the baby bonus did have an enduring effect on bringing forward second births. This would be something for future researchers to follow up on.

Nevertheless, this research adds to the evidence that relatively generous cash payments like Australia's baby bonus are unlikely, on their own, to reverse declining total fertility rates. It is becoming abundantly clear that low fertility will be an enduring feature of future population change.

Read more:

• The pandemic 'baby bust' in other countries
• Cohort effects and the downturn in US fertility since the Great Recession
• The new economics of fertility
• The economics of the falling total fertility rate in New Zealand
• The economics of fertility in high-income countries
• Can fertility return to replacement levels?
• You can make future population decline disappear just by changing the way you categorise people and fertility

This week in research #126

Here's what caught my eye in research over the past week:

• Klebl, Jetten, and Kirkland (open access) find that citizens living in nations with greater (vs. lower) levels of economic inequality were more likely to attribute responsibility for climate change mitigation to individual people rather than to governments (and businesses)
• Andersen, Grimsrud, and Lindhjem find that residential property prices in Norway decline in proximity to operating wind farms, with prices lower by 4 to 14 percent for properties within 2 km of a wind farm, and with effects declining to zero by 7 km from a windfarm