Autism vs. the sunk cost fallacy

One of the many ways in which 'real world' decision-makers fall short of the purely rational ideal is that decision-makers in the real world are subject to the sunk cost fallacy. Sunk costs are costs that have already occurred and that cannot be recovered. In his book Misbehaving: The Making of Behavioral Economics (which I reviewed here), the 2017 Nobel Prize winner Richard Thaler argues that the sunk cost fallacy arises because of a combination of loss aversion and mental accounting.

In general, people are loss averse because we value losses more than we value equivalent gains. Gaining $10 makes us happier, but losing $10 makes us unhappier to a greater extent than gaining $10 makes us happier. So, we generally try to avoid losses.

Mental accounting suggests that we keep 'mental accounts' associated with different activities. We put all of the costs and benefits associated with the activity into that mental account, and when we stop that activity, we close the mental account associated with it. At that point, if the mental account has more costs in it than benefits, it counts as a loss. And because we are loss averse, we try to avoid closing the account.

Part of the issue with susceptibility to the sunk cost fallacy is that real world decision-makers are thinking emotionally. If they were dispassionate logical-thinking robots, they wouldn't take sunk costs into account in their decisions. But although decision-makers are not all equally susceptible to the sunk cost fallacy, the costs in terms of sub-optimal decision-making may be substantial. So, studies of the sunk cost fallacy in different population groups are important.

One interesting new study by Nicky Rogge (KU Leuven), published in the Journal of Economic Psychology (sorry I don't see an ungated version online), looks at the difference between people with autism spectrum disorder (ASD) and neurotypical people. Rogge first reminds us of 'Dual Process Theory', which posits that:

...reasoning and decision making can be described as a function of two processing or reasoning systems: the intuitive reasoning system and the deliberative analytic-logical reasoning system. The intuitive reasoning system involves an implicit, unconscious reasoning process that is independent of cognitive ability and working memory, and that is rapid and automatic. The deliberative analytic-logical reasoning system involves an explicit (controlled), conscious reasoning process that depends strongly on cognitive ability and working memory, and is slower and more effortful.

Some of you may recognise this as the 'System 1' and 'System 2' thinking processes that 2002 Nobel Prize winner Daniel Kahneman outlined in his book Thinking, Fast and Slow. Rogge then outlines some of the literature on thinking processes among people with ASD, and notes that:

Brosnan et al. (2016, 2017) and Lewton et al. (2019) argued that the pattern of reasoning and decision-making styles adopted by individuals with ASD is more biased away from intuitive reasoning and more towards deliberative reasoning styles, as compared to what is observed in neurotypicals.

That suggests that, to the extent that the sunk cost fallacy arises from decision processes occurring within the intuitive 'System 1', that people with ASD may be less susceptible to the sunk cost fallacy (as well as potentially other heuristics and biases that together define 'quasi-rational' behaviour).

Rogge then tests a number of hypotheses related to this, using data collected from an online survey of 332 people from Belgium, 187 of whom self-reported as having been diagnosed with ASD, while 34 reported a strong suspicion of ASD but no diagnosis, and 111 'neurotypicals', who reported no ASD. Rogge doesn't just take the research participants' word for it - he administers to AQ-Short test to derive a quantitative measure of where each research participant (self-reported ASD or neurotypical) fits on a scale (the AQ-10 scale). The survey asked participants about six problems, where:

Each of the six sunk-cost decision tasks presents a hypothetical decision scenario which involves a sunk cost...

For each decision task, research participants rated how relatively likely they were to choose between two options, one of which involved accepting a sunk cost. Rogge then uses the responses from the six decision tasks to derive a score for susceptibility to the sunk cost fallacy. He also knows how long each research participant spent on each decision task, which he uses to proxy for how thoughtfully they considered the options (i.e. how much 'System 2' thinking was involved). Then, he uses propensity score matching to create a matched sample of research participants with ASD and neurotypicals, and analyses the differences in sunk cost score between the groups. He finds that:

...(a) the sunk cost did impact the decision made by the average participant across the six sunk-cost decision tasks... (b) participants with ASD were generally less subject to the sunk-cost bias as compared to neurotypical participants... (c) participants with ASD and more autistic traits (as measured by the AQ10-score) were generally less subject to the sunk-cost bias as compared to individuals with ASD and less autistic traits (and neurotypical individuals)... (d) the time to complete a sunk-cost decision task related negatively to the sunk-cost bias for participants with ASD... and (e) this negative relation between time spent in the decision task and the sunk-cost bias was more pronounced for individuals with more autistic traits as compared to their counterparts with less autistic traits (both ASD and neurotypical)...

So, score a win for people with ASD. They are less susceptible to the sunk cost fallacy, and consistent with that, they spend more time on the decision tasks than neurotypicals do. That doesn't answer the bigger question of 'why', but it does demonstrate that people with ASD have more rational decision-making processes in this context.

Of course, there are some problems with this study, and it needs replication in other samples. The biggest issue is the nature of the hypothetical scenarios. It would be interesting to see if similar results would be found in an experimental setting, rather than in an online survey. Also, this type of study could easily be extended, as Rogge notes in his conclusion:

It would be interesting for future studies to measure and compare the sunk-cost bias of individuals with ASD and neurotypical individuals in real-world decision scenarios involving sunk costs and explore the role of, for instance, social and communication skills in sunk-cost bias. Another research question consists in exploring whether individuals with ASD take hypothetical and experimental tasks more seriously and, if so, whether this explains for why they make more consistent and less biased decisions than neurotypical individuals.

Both of those options would be worthwhile additions to the growing research literature on decision-making among people with ASD.

Church attendance and crime

Does church attendance reduce crime? A simple causal argument could be made that church attendance affects people's moral judgements and therefore their behaviour, reducing crime. However, attempting to establish whether this is the case is going to be difficult empirically, because it isn't easy to conduct an experiment and people aren't randomly allocated to attend church.

This 2020 paper by Jonathan Moreno-Medina (Duke University) takes a slightly different approach. Moreno-Medina relies on the quasi-experimental variation in church attendance that arises because of the weather. Specifically, people are less likely to attend church when it rains. By looking at the extent to which it rains during the specific time window associated with church services (which Moreno-Medina sets as 9am-1pm on Sundays) and how that affects church attendance, Moreno-Medina has an instrument that he can use to extract the causal impact of church attendance on crime. The rain data comes from hourly observations by NOAA, and the county-level crime data comes from the Uniform Crime Reports, and both datasets cover the period from 1980 to 2016.

In a standard instrumental variables analysis, all of the data belongs to the same dataset and both the first stage and second stage regression models can be run jointly. However, that isn't the case for this paper, which is based on the more-rarely-used two-sample two-stage least squares method. Essentially, Moreno-Medina first runs a regression that predicts church attendance at the county level, based on precipitation at the time of church (PTC) and other control variables. The church attendance data comes from the American Time Use Survey, and Google's Popular Times. He then estimates the reduced form regression model for crime, with PTC as an explanatory variable (and including other control variables). The combination of the reduced form model and the first stage can be used to extract the causal effect of church attendance on crime (which I won't go into detail on here, but it is explained in the paper, as well as in econometrics textbooks such as Angrist and Pischke's Mostly Harmless Econometrics, which I reviewed here). You might be concerned that PTC is correlated with precipitation at other times of the week. Moreno-Medina therefore controls for precipitation at other times as well.

In terms of the key results, he finds that:

...having one more Sunday with precipitation at the time of religious services increases substance-related and ‘white-collar’ yearly crimes by around 0.25%, but I find no effect on more serious offenses such as violent crimes (murder, rape, aggravated assault, and robbery) and property crimes. The implied effect of church attendance shows that an increase of 1% in the attendance rate would reduce drug-related crimes by 0.8%, alcohol-related crimes by 0.66% and white-collar crimes by 0.67%.

The headline results in terms of crime overall are mostly plausible. The effects on alcohol-related or drug-related crimes occur within a short time window, but the effects on white-collar crimes (fraud, forgery, etc.) occur with a lag. A lack of statistically significant effects on violent crimes is consistent with a mechanism where church affects crimes to a much weaker extent for violent crimes than for substance-related or white-collar crimes. Moreno-Medina isn't able to specifically identify the mechanisms by which church attendance reduces crime. However, interestingly:

I characterize the population of compliers (individuals who would not attend church because it precipitated at that time) along some observable characteristics. I find that compliers are much more likely to have at least some post-secondary education, and are more likely to be young adults and male. While precipitation decreases average attendance to church, I show that it also increases the probability of engaging in leisure activities at home, but has no effect on other potential confounding activities, such as going to restaurants, outdoors or to the mall.

Given that the results appear to be driven by young males, and 'engaging in leisure activities at home' appears to be what church attendance is substituted for, is this evidence that video game use reduces crime? I'll leave that thought for some further consideration in the future.

[HT: Marginal Revolution]

Inequality in New Zealand over the long run

The mainstream idea that income inequality is increasing in New Zealand seems difficult to counter. Despite the fact that the data doesn't any support for increasing inequality (see this post), people still prefer to believe it. That may be because concern about inequality has increased - people are more worried about inequality now than in previous decades, which makes it seem like it's a bigger problem than before, even if inequality itself has remained relatively unchanged since the mid-1990s (although as noted in this post, New Zealanders are less in favour of redistribution than they used to be).

Coming back to the point about how income inequality has changed over time, it is useful to look at the longer term. This post has a picture of inequality going back to 1984 (see also this post). However, this 2018 article by John Creedy, Norman Gemmell and Loc Nguyen (all Victoria University of Wellington), published in the journal Australian Economic Review (ungated earlier version here), outlines the longer time series back to 1935, based on the Gini coefficient measure.

Creedy et al. use data from the Official New Zealand Yearbooks (which I've noted before is an underutilised treasure trove of New Zealand historical data). Of course, they have to deal with changes in the definition of income and how it is measured, as well as who is included in the measurement. The biggest 'problem' is the introduction of the PAYE tax system in 1957/58, which leads to a huge increase in the number of people with low or no measured income being included in the data (the consequence is a large increase in measured inequality). They adjust their inequality measure back to 1935 by first calculating the ratio of the Gini coefficient for people with greater than £400 of income to the Gini coefficient for all taxpayers, based on data for 1959. They then applied that constant ratio to 1958 to get a measure of income inequality for all people for 1958. They then adjusted backwards from there by assuming the year-to-year proportional change in the Gini coefficient is the same for all taxpayers as it is for taxpayers with incomes greater than £400. [*]

Here's the headline graph of income inequality for New Zealand over the long term (Figure 4 from the article):

The change in measured inequality as a result of the introduction of PAYE in 1957/58 is clear. The higher of the two series prior to that is Creedy et al.'s adjusted series. There are several things to note from this graph:

Inequality is seen to be relatively stable from the early 1960s to the late 1980s, after which it rose to 0.47 in 1994... It is hard to escape the view that the increase was associated with the reforms that took place during the 1980s... Relevant changes included the gradual ‘flattening’ of the marginal income tax rate structure, with the top marginal rate falling from 66, to 48 and then to 33 per cent, along with benefit reductions and the end to centralised wage setting...

After 1994 the Gini is relatively constant again, except for a spike in the tax year 1999–2000. This spike is associated with the major income tax changes that raised the top marginal rate from 33 per cent to 39 per cent in 2000. This led to a certain amount of income shifting after the announcement of the change...

Also in relation to the spike in 1951:

This appears to be associated with particularly rapid increases in national income associated with a huge temporary rise in wool prices in 1950–1951... This spike in wool prices and farmers’ incomes would undoubtedly have had a large and disproportionate impact on income levels within the income distribution in 1951, but especially generating temporarily high incomes for wool farmers and related activities.

Overall, the general trends are clear though. Inequality was relatively high (perhaps similar to inequality today) in the 1930s and up to the early 1950s, then fell from the 1950s to the early 1980s. Inequality then rapidly increased back to its prior levels during the reforms of the late 1980s and early 1990s, and then has been relatively flat ever since.

The longer time series raises some interesting questions about current perceptions of inequality. Are people who are concerned about high (and 'increasing') inequality in New Zealand today remembering the 'idyllic' times of the 1970s and early 1980s, when inequality was historically low? Or, are people simply taken in by rhetoric from the U.S., where inequality has been increasing (albeit slower than you might think), without considering that New Zealand might be different?

*****

[*] It would have been interesting to see what it would have looked like if Creedy et al. had applied the method they used for 1958 to all years, rather than their assumed identical rates of change method. They note that:

...examination of the income distributions for early years reveals that the vast bulk of the distribution (around 50–75 per cent) of all income earners during the 1930s and 1940s had incomes below £400; with the percentage generally falling over this period as general income growth occurred. As a result, these early distributions generate unreliable Gini estimates for the sub-sample with x>£400... By the 1950s this proportion had dropped to 30 per cent in 1950 and to 12 per cent in 1957...

The persistent effects of the Spanish Inquisition

Continuing the economic history theme from the yesterday's post, this new article by Mauricio Drelichman (University of British Columbia), Jordi Vidal-Robert (University of Sydney), and Hans-Joachim Voth (University of Zurich), published in the Proceedings of the National Academy of Sciences (ungated earlier version here) looks at the long-run consequences of the Spanish Inquisition. As they explain:

...we investigate the long-run impact of religious persecution on economic performance, education, and trust. The Spanish Inquisition is among the most iconic examples of a state-sponsored apparatus enforcing religious homogeneity... Histories of Spain’s decline and fall as an economic power frequently emphasize the role of the Inquisition... and sociological studies have argued for a “persistence of the inquisitorial mind” in modern-day Spanish thought...

Obviously, given the time period involved (the Spanish Inquisition ran from 1478 to 1834), this relates to the Little Divergence, as discussed in yesterday's post (and this earlier post). Drelichman et al. have data on the number of people persecuted in the Inquisition for many Spanish municipalities (based on around 67,000 records of Inquisition trials), and relate that to modern-day differences in incomes (measured using nightlight intensity), and survey-based measured of religiosity (measured by frequency of church attendance), education (proportion of the population with a high school diploma) and generalised trust. In terms of income, they find that:

Municipalities with no recorded inquisitorial activity as well those in the lowest tercile of persecution have the highest GDP per capita today. Those affected but in the middle tercile already have markedly lower income. Where the Inquisition struck with highest intensity (top tercile), the level of economic activity in Spain today is sharply lower. Magnitudes are large: In places with no persecution, median GDP per capita was 19,450€; where the Inquisition was active in more than 3 y out of 4, it is below 18,000€... Our estimates imply that had Spain not suffered from the Inquisition, its annual national production today would be 4.1% higher...

In terms of their other outcome measures, they find that:

A one-SD increase in inquisition intensity is associated with an increase of between 1.3% and 3.7% in religious service attendance...

We find a consistently negative relationship between persecution and educational attainment... going from no exposure to the Inquisition to half of all years being affected by persecutions would reduce the share of the population receiving higher education today by 2.7 percentage points, relative to a mean of 47.5 percentage points—a 5.6% relative reduction...

A one-SD increase in inquisitorial intensity reduces average municipality-level trust by 0.03 to 0.05 SDs.

All of the effects are statistically significant. Of course, it would be reasonable to worry that the Inquisition targeted its activities on areas that differed in relevant ways, such as areas that were more religious, or areas that were poorer. However, Drelichman et al. find the opposite:

To address the possibility that the Inquisition might have favored locales with high levels of pre-existing religiosity, we collect data on pre-Inquisition religious figures from the Spanish Biographical Dictionary... We then estimate the local density of famous people with strong links to the church and use it to examine whether inquisitorial intensity was systematically higher in places that were more religious pre- 1480. Our data suggests the opposite—places with greater inquisitorial intensity had a lower density of famous religious individuals...

A second concern is that the Inquisition could have been attracted to poorer areas. Standard histories of the Inquisition suggest this is unlikely. The Inquisition was self-financing. It had to confiscate property and impose fines to pay for its expenses and the salaries of inquisitors. While its mission was to persecute heresy, it had strong incentives to look for it in richer places...

In relation to the question of whether targeted areas were poorer, they use the location of hospitals, which were predominantly located in rural areas and:

Largely geared to care for the poor and for out-of-towners who fell ill and had no other place to stay, hospitals were credited with reducing the number of destitute people living and dying on the streets...

So, hospitals are indicative of areas that were, in general, wealthier and with higher social capital. They find that:

Inquisitorial intensity was five times higher in places with hospitals, and the difference is highly significant...

They move on to repeat many of their analyses using coarsened exact matching (which I discussed in this recent post), and find very similar results. Again, as with much of the literature in economic history, the results do not demonstrate causality, but we are getting closer to understanding the Little Divergence. And the results of this paper might not be limited to Spain. As Drelichman et al. conclude:

...the Inquisition also operated in Southern Italy and throughout Spanish America. Inquisitorial practices were also instituted throughout the Portuguese Empire and the Papal States. All of these areas today show relatively low education, lower incomes, and less generalized trust.

[HT: Marginal Revolution]

Read more:

• Economic institutions and the 'Little Divergence'
• Portugal, gold, and the resource curse in the Little Divergence

Portugal, gold, and the resource curse in the Little Divergence

Back in June, I wrote a post about the 'Little Divergence', based on a paper that showed that the divergence in economic fortunes between England and the Netherlands (which experienced an increase in prosperity), and Spain and Portugal (which experienced a relative decline) was not explained by differences in institutions. So, what does explain the difference in fortunes between northwestern Europe and southwestern Europe in the Little Divergence?

This new paper by Davis Kedrosky (University of California, Berkeley) and Nuno Palma (University of Manchester) may provide an answer. Kedrosky and Palma look at the case of Portugal, and investigate whether Portugal experienced a 'resource curse' due to large production of gold in the Brazilian colony. Gold was discovered in Minas Gerais province in Brazil in the 1690s, and by the 1730s there were 140,000kg of gold mined there, of which 80 percent was sent to Portugal (mostly into private hands). The resource curse (sometimes referred to as 'Dutch disease') occurs when a country's resource wealth leads to an exchange rate appreciation, making its other export sectors uncompetitive internationally, and paradoxically making the country worse off. As Kedrosky and Palma explain in the case of Portugal:

The economy, inclusive of Brazil, can be divided into the expected three sectors: gold is the booming sector, directly augmenting incomes and providing a high marginal product of labor; manufacturing, viticulture, and cereal agriculture constitute the lagging traded sector; and animal and forest production form the land-intensive non-traded sector. Increasing gold production in Brazil enriched Portuguese nationals in the colonies, who either remitted their funds home or exchanged them directly for the durable goods arriving on the thrice-annual treasure fleets. Newly-expanded incomes would necessarily increase demand in the non-traded sector, causing a real appreciation and a consequent withdrawal of resources from the lagging sector through the spending effect... With the collapse of the traded export sector and the increased purchasing power of the currency, exports would decline and imports surge, increasing the trade deficit. Gold re-exportation would then be required to pay the outstanding balances, as the proceeds of domestic industry no longer earn foreign exchange.

In the long run, the decades-long influx of gold would tend to continue the compression of the traded sector (or at least arrest its expansion). Skilled-based productivity advances, which depend on industry and cereal agriculture, would be persistently stifled, removing the principal driver of expansion. After an initial boom, therefore, GDP growth rates would tend to slow, or even reverse, with critical sectors of the economy being replaced by imports.

Using data on prices in three Portuguese cities from 1650 to 1825, Kedrosky and Palma calculate the ratio of traded good prices (including wheat/maize, wine, olive oil, linen, and candles) to non-traded good prices (meat, hens, eggs, soap, and charcoal). If Portugal suffered from the resource curse, then the price of non-traded goods should increase relative to the price of traded goods. Indeed, that is what they find, as shown in their Figure 4:

Of course, an increase in relative prices doesn't constitute on its own evidence that the resource curse was the cause of Portugal's relative decline. However, Kedrosky and Palma collate some additional evidence, concluding that:

While GDP rose in the short run, the longer-term effects of the gold shock were negative — contractions in industry and cereal production slowed the accumulation of technical progress, causing stagnant growth in successive years. Even if manufacturing’s share of employment converged to pre-shock levels, income was permanently reduced. Over the eighteenth century, the country de-industrialized, and in fact the percentage of the population working outside of agriculture declined from around 46% in 1750 to only 33% a year later.

Clearly, there is more work to do in this area, but the case against institutions as being determinative of the 'Little Divergence' is becoming clearer.

[HT: Marginal Revolution]

Read more:

• Economic institutions and the 'Little Divergence'

Top5itis in economics

Back in 2018, I wrote a post about the emphasis in economics on publishing in the 'top five' journals (the American Economic Review, Econometrica, the Journal of Political Economy, the Quarterly Journal of Economics, and the Review of Economic Studies). That post was based on this NBER Working Paper by James Heckman and Sidharth Moktan. They found that there was high value (in terms of the probability of achieving tenure) associated with publishing in the top five, although there were troubling gender differences.

So, perhaps the emphasis on the top five is not entirely bad. However, not everyone agrees. This short 2018 paper by Roberto Serrano (Brown University), title "Top5itis" presents a contrary view:

Top5itis is a disease that currently affects the economics discipline. It refers to the obsession of the profession of academic economists with the so-called “top5 journals.” These are, alphabetically, the American Economic Review, Econometrica, the Journal of Political Economy, the Quarterly Journal of Economics, and the Review of Economic Studies. In its most extreme forms, top5itis reduces the evaluation of a paper to the following test: a paper has any value if and only if it was published by one of the journals in this list. Therefore, in order to evaluate the scientific production of a scholar, a person affected by top5itis simply counts the scholar’s “top5’s.” Since the disease is a simple application of the counting measure (typically a person learns to count in primary school), through a process of contagion, top5itis spreads quickly to affect people outside economics, including schoolchildren offspring of economists who get together in the playground to make disparaging remarks about each other’s parents. With similar patterns, the disease has also spread to competent university administrators and influential granting agencies.

Ok, Serrano's paper is not entirely serious (but worth reading for a lighter take on the issue, especially the 'true stories'). However, there clearly is an overt focus on these top journals in economics, and that has negative unintended consequences (including problems of reporting bias and publication bias, and bias towards 'surprisingness', as mentioned in my previous post). Would Maximilian Kasy's suggested alternative publishing structure solve the problem of top5itis? Unfortunately, it is doubtful (both whether it would be a cure, and whether the alternative is workable).

[HT: Marginal Revolution, back in 2018]

Read more:

• The emphasis on publishing in the top five economics journals

The past and future of statistical significance

The latest issue of the Journal of Economic Perspectives had a symposium on statistical significance, which included three articles (all ungated). In the first article, Guido Imbens (Stanford University) outlines three concerns with the use of statistical significance and the use of p-values:

The first concern is that often p-values and statistical significance do not answer the question of interest. In many cases, researchers are interested in a point estimate and the degree of uncertainty associated with that point estimate as the precursor to making a decision or recommendation to implement a new policy. In such cases, the absence or presence of statistical significance (in the sense of being able to reject the null hypothesis of zero effect at conventional levels) is not relevant, and the all-too-common singular focus on that indicator is inappropriate...

The second concern arises if a researcher is legitimately interested in assessing a null hypothesis versus an alternative hypothesis... Questions have been raised whether p-values and statistical significance are useful measures for making the comparison between the null and alternative hypotheses...

The third concern is the abuse of p-values... To put it bluntly, researchers are incentivized to find p-values below 0.05.

These are all concerns that are not new, and relate to the case made in the book The Cult of Statistical Significance by Stephen Ziliak and Dierdre McCloskey (which I reviewed here). Imbens argues that the first concern is the most important. Interestingly, he takes a more moderate view than others have done in recent years:

In my view, banning p-values is inappropriate. As I have tried to argue in this essay, I think there are many settings where the reporting of point estimates and confidence (or Bayesian) intervals is natural, but there are also other circumstances, perhaps fewer, where the calculation of p-values is in fact the appropriate way to answer the question of interest.

Confidence intervals do make a lot of sense. However, to me they are still not so much different to a p-value (the 95% confidence interval is just as arbitrary as a p-value of 0.05).

In the second article, Maximilian Kasy (University of Oxford) discusses the problems arising from the 'forking path'. The forking path is a metaphor drawn from Jorge Luis Borges, who:

...wrote a short story in 1941 called “The Garden of Forking Paths.” The plot involves (among other elements) a journey in which the road keeps forking...

Statisticians have used the metaphor from Borges to convey how empirical research also involves a garden of forking paths: how data is chosen and prepared for use, what variables are the focus of inquiry, what statistical methods are used, what results are emphasized in writing up the study, and what decisions are made by journal editors about publication.

Essentially, this article is about the selection bias in published research, arising from reporting bias (where only some, but not all, statistical results are reported in published studies) and publication bias (where only some, but not all, studies are published). Kasy outlines the problems (which again, are well known to researchers), and then some potential solutions, including: (1) pre-analysis plans, where the analyses are pre-specified and deviations along the forking paths can easily be identified by editors, journal reviewers, and readers of research; (2) pre-results journal review (or 'registered reports'), where journal articles are accepted on the basis of proposed analyses, before any analysis is conducted or results are available; and (3) journals for null results and replication studies, which could reduce the publication bias and help to identify studies with fragile results.

Kasy finishes by making an alternative proposal for the structure of publishing:

There might be a set of top outlets focused on publishing surprising (“relevant”) findings, subject to careful quality vetting by referees. These outlets would have the role of communicating relevant findings to attention-constrained readers (researchers and decision-makers). A key feature of these outlets would be that their results are biased by virtue of being selected based on surprisingness. In fact, this is likely to be true for prominent outlets today, as well. Readers should be aware that this is the case: “Don’t take findings published in top outlets at face value.”

There might then be another wider set of outlets that are not supposed to select on findings but have similar quality vetting as the top outlets, thus focusing on validity and replicability. For experimental studies, pre-analysis plans and registered reports (results-blind review) might serve as institutional safeguards to ensure the absence of selectivity by both researchers and journals. Journals that explicitly invite submission of “null results” might be an important part of this tier of outlets. This wider set of outlets would serve as a repository of available vetted research and would not be subject to the biases induced by the selectivity of top outlets...

To make the findings from this wider set of publications available to attention-constrained decision-makers, systematic efforts at aggregating findings in review articles and meta-studies by independent researchers would be of great value... Lastly, systematic replication studies can serve as a corrective for the biases of top publications and as a further safeguard to check for the presence of selectivity among non-top publications.

I'm not sure how workable that system is, or how we would get to there from where we are today. Some of the elements are already in place, and replications and systematic meta-analyses are becoming more common. However, there would be substantial reluctance on top journals to be seen as publishing research that is 'biased by surprisingness'.

The third article, by Edward Miguel (University of California, Berkeley) focuses on open science and research transparency. This covers some of the same ground as Kasy's article (pre-analysis plans, and registered reports) but also covers the availability of statistical code and data to be used for replication. Miguel notes that there has been an increase over time in the sharing of data and code, but he also notes that it is not without cost:

Across 65 project datasets, the average amount of time to prepare replication materials for public sharing was 31.5 hours, with an interquartile range of 10.0 to 40.5 hours (and a 10th to 90th percentile range of 5.8 to 80.2 hours). This is non-trivial for most projects: still, remember that this estimate of preparation time applies to field experiments that often require multiple years of work on collecting data, so it remains a very small share of overall project work time.

There are offsetting benefits though:

The most immediate private benefit that I and many other scholars have personally experienced from new open data norms is the fact that our own research data is better organized for ourselves and thus easier to reuse for other analyses and papers as a result of the effort that is put into improved documentation (like the README files and other replication materials).

There are a couple of salient issues here, both of which Miguel touches upon. The first issue is equity - producing replication materials is likely to be lower cost for researchers who employ a small army of research assistants, whereas many researchers would have to do this work themselves. The second issue relates to replication more generally, where:

...researchers’ growing ability to access data and code from previous studies has led to some controversy... there may be “overturn bias,” in which reanalysis and replications that contradict an originally published paper are seen as more publishable.

This is related to the 'surprisingness' that Kasy notes. The last thing we would want is that the journals that are devoted to replication have a bias towards negative findings (so, then there would be publication biases in both directions).

Overall, there is a lot of interest in the three articles in this symposium. It is not all bad news - an optimistic view would be that many of the problems with statistical significance, publication bias, etc. are already acknowledged, and steps are already being undertaken to address these. The biggest thing that researchers can do going forward, though, is to be a bit more sensible in relation to interpreting statistical significance. The difference between a p-value of 0.049 and a p-value of 0.051 is not in itself statistically significant. As Ziliak and McCloskey noted in their book (and a point that was only raised by Imbens of the three authors in this symposium), it is economic significance, rather than statistical significance, that is most important.

Migration to Australia and New Zealand vs. North America during the Age of Mass Migration

Historians refer to the period between about 1850 and the start of the First World War as 'The Age of Mass Migration' - a period when migration out of Europe expanded greatly, with over 40 million emigrants, going particularly to North America, but also to Australia and New Zealand and elsewhere. These huge migration flows substantially changed the demographics of the destinations, influencing these societies for generations to come. It is interesting, then, to look at the differences in migration flows going to different countries.

In a recent article published in the journal Australian Economic History Review (open access), Timothy Hatton (University of Essex) compares the migration flows from the UK (including Ireland) to the U.S., Canada and Australia/New Zealand (combined). He notes that:

From 1853 to 1913 the outflow of British citizens amounted to 12.9 million, 10.1 million of whom departed after 1870. Most of these passengers were emigrants travelling in the steerage compartments of emigrant ships. About three fifths of the emigrants were male, about three fifths were single and, among the adults, more than four fifths were aged 18-45 years. This represents a substantial labour migration of prime age adults, some of whom were travelling as families with children. They came from all parts of the United Kingdom with a particularly heavy concentration from Ireland, especially in the 1850s and 1860s. But there was also a substantial return flow. From 1871 to 1913 the cumulative net outflow amounted to 5.9 million, equivalent to 13.1% of the UK population in 1913...

Most of the UK emigrants travelled to four main destinations, the United States, Canada, Australia and New Zealand. From 1871 to 1913 these countries account for 88% for the gross outflow of UK citizens to extra-European ports and 93% of the net outflow.

Hatton then goes on to investigate the factors associated with the size of migration flows to the three destinations, using data from 1872 to 1913. He finds that:

the key forces are real wage rates, business cycle shocks as represented by deviations from trend GDP and network effects as captured by the stock of previous emigrants at a destination... The costs of passage seem to have little influence, perhaps because of measurement issues. But nearly half of the emigrants to Australia and New Zealand travelled on assisted passages, which brought the costs of travel to the antipodes closer to those for crossing the Atlantic. This seems to have been important in sustaining the flow to much more distant shores.

One of the more interesting (and perhaps surprising) aspects of the data is that it appears that the migrants to Australia and New Zealand were more skilled on average than migrants to North America. On this point, Hatton notes that:

Over the whole period, the percentage skilled is 64.0% for Australia and New Zealand, 38.7% for the United States and only 26.1% for Canada.

That is based on emigrants whose occupation was classified by the U.K. Board of Trade as "skilled trades". The results are similar if commercial and professional occupations are also included. As for the reason why, it might be attractive to suggest that assisted passage helped. However, using a model of the skilled share of migration, Hatton concludes that:

It is also likely that the larger network of previous migrants in the United States provided more opportunities for emigration among the unskilled, which would be consistent with the negative coefficients on the emigrant stock. Weaker network effects, due to the smaller emigrant stock, would help to explain the higher skill content of emigration to Australia/New Zealand compared with the United States (but not with Canada). But one thing that does not explain the difference in migrant skills is assisted migration, which, if anything, reduced the skill content of migration from the United Kingdom to Australia and New Zealand.

Migration involves a cost to the migrant. Unskilled workers are generally less able to pay that cost. However, part of the cost is integrating into the destination society (getting a job, housing, etc.), and that component of the cost is reduced if the migrant has a larger social network in the destination to take advantage of. The larger social network of unskilled (particularly Irish) migrants in the U.S. appears to have made North America more attractive as a destination for unskilled migrants than Australia or New Zealand, leading to more skilled migration flows to the antipodes (on average). It also helps to explain the relative lack of Irish heritage in Australia and New Zealand (in comparison with the U.S).

Teenage daughters and divorce

Sociological research has shown that couples with daughters are more likely to divorce than couples with sons. However, there are a number of competing (and complementary) theoretical explanations for this observed relationship, and it has been difficult to disentangle the different mechanisms that might be at work. In a recent article published in The Economic Journal (sorry I don't see an ungated version online, but there is this conference presentation), Jan Kabatek (University of Melbourne) and David Ribar (Georgia State University) use comprehensive data from the Netherlands to investigate. Specifically, they use:

...administrative data that cover the near universe of marriages and registered partnerships that began in the country between October 1971 and December 2015. The data include nearly 3 million marriages and partnerships, allowing us to estimate effects precisely and consider how effects vary with children’s ages, parities, and conditional on parents’ backgrounds. The data are highly accurate with the exact dates of weddings, births, and divorces.

This is the sort of research that can only be done with population register data like this, because the sample size is enormous and sample selection is less of an issue. However, first let's take a step back. Kabatek and Ribar outline the various competing theories for why couples with daughters may be more at risk of divorce:

Consider an overarching preference for boys over girls. Such a preference would raise the value of the marriage-specific capital for couples with sons, thereby lowering their incentives to divorce... Similar effects would occur if fathers prefer spending time with sons more than spending time with daughters...

The constraint-based explanations posit that daughters are more costly to raise than sons... Another possibility is that boys are more susceptible to developmental problems if parents divorce, which would lower the value of parents’ alternatives to marriage... It is also possible that parenting interactions are more strained with girls than boys and that these strains lower the match-specific quality.

...it is also possible that the association is not causal. Hamoudi and Nobles (2014) described how girls in utero have survival advantages under conditions of stress relative to boys. They found that mothers who reported high levels of relationship conflict prior to their children’s births were more likely to give birth to girls. This means that the gender-related divorce disparities may result from sex-selection into live birth.

Kabatek and Ribar use what is referred to as a complementary log-log (cloglog) discrete-time hazard model. When what you are modelling is really a type of survival analysis, cloglog provides a better fit than a logistic regression, and it deals with low-probability events better. Anyway, enough of that. In the headline results, they find that:

...having a daughter increases the risks of divorce among Dutch couples - the first robust finding of this association for a European country. An even more novel and intriguing result is that the increased risks of divorce only appear when daughters are teenagers (aged 13-18) - there is no detectable gender difference at earlier or later ages. We observe this pattern among both firstborn and subsequent children. We also find the same age pattern in analyses of the 1980, 1985, 1990, and 1995 US Current Population Survey Marriage and Fertility Supplements (CPS-MFS).

The divorce risks for couples with firstborn daughters aged between 13 and 18 are about 5.5% higher, on average, than the risks for couples with firstborn sons. However, because there are no differences at younger or older ages, the cumulative effect of the child’s gender on parental divorce is modest. For instance, the cumulative divorce rates of couples with firstborn sons and daughters aged 19 only differ by 1.8% (0.36 percentage points).

So far, so consistent with the sociological literature, except for the fact that the result is largest for teenage daughters. Kabatek and Ribar then go further:

The ‘teenage daughter effect’ is at odds with many possible explanations, including: (i) overarching, time-invariant preferences for sons, (ii) sex-selection into live birth, and (iii) rational forward-looking behaviours based on age-specific differences in preferences for or costs of raising sons and daughters. This is because each of these explanations predicts that gender disparities would also appear earlier in the child’s life. Instead, the precisely-estimated pattern of zero effects during childhood and non-zero effects during the teenage years leads us to consider explanations based on unexpected changes in parents’ valuation of marriage during children’s adolescence. We give special consideration to family conflicts that might arise from differences in family members’ gender-role attitudes. These differences can become a more salient source of conflict as girls mature, and conflict in this dimension of family life may spill over to the parents’ marital relationship.

Additional analyses of Dutch administrative data support this explanation. We find that the excess divorce risks associated with having teenage daughters are higher for couples whose gender-role attitudes are likely to differ from those of their daughters (for example, for less-educated couples and immigrant couples), and that the risks are exacerbated further for couples in which the two spouses are likely to hold different gender-role attitudes (for example, for parents with different immigration backgrounds or different levels of education). In a further analysis, we examine the gender composition of the parents’ siblings and find that the teenage daughter effect only appears among fathers who grew up without sisters. In contrast, we find no differences with regard to the gender composition of the mother’s siblings, which suggests that the father’s experiences are critical.

It appears that the effect of daughters on divorce arises from conflict between parents in terms of their beliefs in gender roles. This only becomes problematic when the daughter is older and presumably better able to assert themselves. It is particularly interesting that there is a difference here between fathers with sisters and fathers without sisters, suggesting that having sisters may affect fathers' beliefs about gender roles, and reduce conflict with their daughters. Kabatek and Ribar use additional survey data to support these explanations.

Some would not doubt argue that these results fall short of a definitive causal explanation. The gender of children is mostly distributed randomly, but there may be some concerns about sex selection, especially for more recent children. Kabatek and Ribar test for differences between couples with first-born sons and couples with first-born daughters, and there are few statistically significant differences. However, couples who are both immigrants are more likely to have a first-born son, and parents wait slightly longer before having a second child after a first-born son. The number of siblings is also higher for couples with a first-born son, but the difference is tiny (0.003 children, on average). Those variables are all controlled for in the analysis, but there might still be unobservable differences between those couples that are correlated with the chance of divorce.

However, despite that caveat, this is the best analysis we have to date on why couples with daughters are more likely to divorce than couples with sons.

Global inequality and the Subnational Human Development Index

The Human Development Index (HDI) is a widely used summary measure of the level of development of countries. It improves on a simple ranking of GDP per capita or income per capita, because it takes into account health and education. Specifically, it is made up of four indicators: (1) life expectancy at birth; (2) mean years of schooling of adults (aged 25 years and over); (3) expected years of schooling of children aged 6 (which is based on current age-specific enrolment rates at each level of schooling); and (4) gross national income per capita (adjusted for purchasing power parity).

However, one of the problems with the HDI is that it aggregates across each country as a whole. If you want to know anything about the relative levels of development in rural and urban areas of a country, or coastal and landlocked areas, or between different states, the HDI doesn't provide much assistance. However, help is at hand. There is now a Subnational Human Development Index (SHDI) available, and published by the Global Data Lab. The SHDI covers 1625 regions in 161 countries going from 1990 to 2019. Interestingly, for New Zealand it provides index values for all 16 regions, which might be useful for research (because there seems to be a reasonable amount of variation both between regions and over time).

I was alerted to the SHDI's existence by this 2020 article by Inaki Permanyer (Centre d'Estudis Demogràfics) and Jeroen Smits (Radboud University), published in the journal Population and Development Review (ungated version here, and useful summary here). Permanyer and Smits use the SHDI to characterise changes global human development inequality since 2000, which marks a change from considering inequality purely in terms of income (or wealth). Here's the 2018 distribution of the index (Figure 1 from the paper):

Permanyer and Smits note that:

...one can observe clear geographic patterns within countries (e.g., north–south divides in Belgium, Germany, Italy, and Spain). Some countries exhibit large regional variations (e.g., China, India, or Colombia) while others are quite homogeneous (e.g., Australia). Very often, the region where the capital city is located exhibits the highest human development levels and remote rural regions the lowest.

Not all of those are visible in the figure of course, due to the scale. Then, looking at inequality as measured by the Gini coefficient, they find that:

...inequality in the global SHDI distribution has monotonically decreased from 0.14 in 2000 to 0.11 fifteen years later. 

That's consistent with the overall trend observed in income (see here, for example). There are similar trends in the components of the SHDI (health, education, and income). However:

...we observe substantial differences in the magnitudes and speed of the decline. According to the Gini index, differences in the life expectancy index across world regions are smaller than differences in the education index.

Countries are converging much quicker in terms of health than in terms of either education or income. Looking at whether global inequality is mostly within or between countries, they show:

...the very high contribution of within country inequality to total inequality in the groups of countries at low- and intermediate levels of development (where as much as 70 percent of the world population lives). In these groups of countries, about half of inequality in SHDI is within-country inequality.

That is quite a different result from the analysis of Branko Milanovic, who showed that only 10-20 percent of global income inequality was within-country inequality (see this post). Permanyer and Smits explore their results a little further, finding that for the least developed countries:

...within-country SHDI inequality is mostly due to variation in education... In the high developed countries, standard of living surpasses education as the most important explanatory factor for within country SHDI variation...

So, if you only consider variation in per capita income, as Milanovic does, you potentially miss a large contributor to within-country inequality in the least developed countries, which is the variation in education.

All of this helps to paint a more complete picture of global inequality in living standards. Looking forward, if we want greater equality in human development, there clearly needs to be a greater focus on education in developing countries.

Read more:

• The great Chinese inequality turnaround?
• Several research articles on inequality
• Chinese growth and global inequality
• More evidence of declining global inequality
• Changing global inequality, migration and open borders
• The last two centuries of global inequality
• Inequality was similar, or higher, in ancient times than it is today

The great Chinese inequality turnaround?

Despite any news you have heard to the contrary, inequality globally has been decreasing over the last several decades (see the links at the bottom of this post for more on this point). Largely though, the reduction in global inequality has been driven by the remarkable growth in incomes in China and India, especially in China. That has, paradoxically, led to an increase in inequality within China, at the same time that inequality globally has been reducing (for more on this point, see here).

However, now it appears that China may be turning the corner. At least, that is the conclusion of this new article by Ravi Kanbur and Yue Wang (both Cornell University), and Xiaobo Zhang (Peking University), published in the Journal of Comparative Economics (ungated earlier version here). They use data from several waves of two household surveys (the Chinese Household Income Project (CHIP) and the China Family Panel Studies (CFPS)), covering the period from 1995 to 2018, and first demonstrate that:

...the time path of inequality, after increasing sharply before 2010, has hit a plateau. Inequality declined after 2010 till 2016, and despite the rebound in 2018 over the last decade it did not go much or at all above the peak of 2010.

In other words, after several decades of rising inequality over time, China has recently experienced a decrease in inequality. Kanbur et al. then dig into the sources of inequality, and find that the trends in inequality are driven by underlying trends in inequality in wage income. They also show that:

The long-term trend of regional consumption inequality shows that there are three peaks for the rural-urban between component, in 1995, 2000, and 2004. After the third peak, the rural-urban between component maintained a declining trend. The trend of regional income inequality since 2002 also shows the same pattern with the peak at 2005. Notice that both regional inequality and rural-urban between components as measured with income and consumption turned downward the mid-2000s.

In other words, the recent decline in country-level inequality in China arises from a decrease in inequality between rural and urban areas. Rural areas are 'catching up' with urban areas in terms of income, which might arise as rural-urban migration increases the supply of labour in urban areas, depressing wages there, while decreasing the supply of labour in rural areas, resulting in higher wages. In the development economics literature, this is referred to as the "Lewis turning point", after the 1979 Nobel Prize winner W. Arthur Lewis.

Kanbur et al. then go on to describe some of the policy changes that might have helped to depress rural-urban inequality, including:

...in 2004 the Ministry of Labor and Social Security issued a “Minimum Wage Regulations” law and the next decade saw rising minimum wage standards coupled with substantial improvements in compliance... Further, a number of social programs were introduced and strengthened from the 2000s onward. Since 2004, for example, China has introduced new rural cooperative medical insurance, currently covering more than 95 percent of the rural population. Rural social security has also been rolled out since 2009.

However, this is clearly not the last word on Chinese inequality. The data that Kanbur et al. use doesn't cover all provinces in China, and for a lot of their analysis (including the rural-urban differentials) they rely on province-level comparisons, which omits any consideration of inequality within provinces. Their results are also at odds with the predictions of other migration researchers, including Branko Milanovic, who predicted that by 2020 China would be contributing to an increase in global inequality. Of course, the slight uptick in inequality that Kanbur et al. note at the end of their time series might just be the start of that. It will be interesting to see how things progress from here.

Finally, and ironically, Kanbur argued in an earlier article (which I discussed in this post) that we should focus on global inequality, rather than inequality on a country-by-country basis. And yet, here he presents further research on country-level inequality (albeit a very important country in a global context, and in the context of global inequality in particular). Clearly there is more to come on this topic.

[HT: Marginal Revolution, last year]

Read more:

• Several research articles on inequality
• Chinese growth and global inequality
• More evidence of declining global inequality
• Changing global inequality, migration and open borders
• The last two centuries of global inequality
• Inequality was similar, or higher, in ancient times than it is today

Inequality and the race between education and technology

A key contributor to income inequality is the wage premium for a university education. Workers with a bachelor's degree or higher earn much more than workers with higher school or less education. This wage premium partially arises from increased human capital (better skills and training), and partly from signalling (for more on this point, you should read Bryan Caplan's The Case Against Education (which I reviewed here; see also this 2019 post on the returns to dropping out of university).

How much does the university wage premium contribute to inequality? This 2020 article published in the American Economic Review Papers and Proceedings issue (ungated earlier version here) by David Autor (MIT), Claudia Goldin and Lawrence Katz (both Harvard University) provides some answer for the U.S.

The underlying framework of their analysis is interesting in its own right (and was the subject of this 2010 book by Goldin and Katz, which I hope to write about once I finally get a second-hand copy of it next month). The framework is the 'race between education and technology', which was first described by Nobel Prize winner Jan Tinbergen in the 1970s. Essentially, on the one hand skills-biased technological change increases the demand for highly-skilled workers (i.e. university graduates), and tends to push up wages and the university wage premium. On the other hand, increasing numbers of university graduates increase the supply of highly-skilled workers, and tend to push down wages and the university wage premium. The evolution of the university wage premium over time then reflects the race between higher education pushing down the premium, and technological change pushing up the premium. Looking back in time, you can make similar arguments in relation to the high school wage premium (when a high school education was far less prevalent than it is today).

The 2020 article by Autor et al. uses data from various sources covering the period from 1825 to 2017. They first show that:

Rising educational wage differentials characterized the period from the 1820s to 1914, not unlike that in more recent history.

The high school movement, starting around 1910, then produced a large decline in the high school wage premium from 1914 to 1960. The college wage premium also narrowed from 1914 to 1950. But then, in a real roller coaster ride, the college wage premium rebounded in the 1950s and 1960s, narrowed in the 1970s, and then soared post-1980. The college wage premium today exceeds its high level of 1914.

Their model of the race between education and technology seems to fit the data on the evolution of the college wage premium reasonably well. They then turn to looking at the contribution to inequality, restricting their attention to the period since 1980, which is of interest because of the consistent increase in income inequality in the U.S. over that time. They find that:

Wage inequality increased at about the same rate from 1980 to 2000 as from 2000 to 2017. But the college wage premium increased far more rapidly in the first period than in the second. The rise in the returns to college education explains a far larger share of the increased log hourly wage variance from 1980 to 2000 than it does from 2000 to 2017, accounting for 75 percent in the first period but just 38 percent more recently.

The canonical two-skill model of the RBET explains the lion’s share of the enormous increase in wage inequality from 1980 to 2000, when the slowdown in the growth of the relative supply of college workers produced a sharp rise in the college wage premium. But most of the recent rise in wage inequality has occurred within, rather than between, education groups.

There are many contributing factors to the underlying level of inequality in society (as I noted in this recent post). The university wage premium (and its underlying driver, skills-biased technological change) is only one contributing factor. It is really interesting though, that it's contribution to U.S. inequality has fallen over time. That points to other structural changes in the labour market, globalisation, and government policy changes such as deregulation, etc. as having contributed more in recent years.

[HT: Marginal Revolution, last year]

Chinese and world demand, and the price of beef in New Zealand

When a country is open to international trade, the prices in the domestic economy don't only reflect domestic factors, but are also affected by changes in the international market as well. Consider this recent story from the New Zealand Herald:

If you feel like red meat is more expensive than it used to be, you're right.

Around 95 percent of New Zealand's sheep meat and 87 percent of beef is exported, and what's left behind for locals is being sold at a premium.

In January 2007, 1kg of beef mince would have cost $9 according to Stats NZ's food price index. If you put a pack in your shopping trolley in January this year, it would have cost $16.39...

Beef + Lamb NZ chief executive Rod Slater said the cost of meat in New Zealand reflected what markets overseas were willing to pay...

One of the emerging buyers for our red meat is China.

African swine flu decimated the country's pig numbers in 2018 and former trade negotiator and founder of consultancy Sanders Unsworth Charles Finny said this was why China had been importing more beef and lamb.

Consider the domestic market for beef, as shown in the diagram below. New Zealand is an exporting country, which means that New Zealand has a comparative advantage producing beef. That means that New Zealand can produce beef at a lower opportunity cost than other countries. On a supply-and-demand diagram like the one below, it means that the domestic market equilibrium price of beef (PD) would be below the price of beef on the world market (PW). Because the domestic price is lower than the world price, if New Zealand is open to trade there are opportunities for traders to buy beef in the domestic market (at the price PD), and sell it on the world market (at the price PW) and make a profit (or maybe the suppliers themselves sell directly to the world market for the price PW). In other words, there are incentives to export beef. The domestic consumers would end up having to pay the price PW for beef as well, since they would be competing with the world price (and who would sell at the lower price PD when they could sell on the world market for PW instead?). At this higher price, the domestic consumers choose to purchase Qd0 beef, while the domestic beef farmers sell Qs0 beef (assuming that the world market could absorb any quantity of beef that was produced). The difference (Qs0 - Qd0) is the quantity of beef that is exported. Essentially the demand curve with exports follows the red line in the diagram.

In terms of economic welfare, if there was no international trade in beef, the market would operate at the domestic equilibrium, with price PD and quantity Q0. Consumer surplus (the gains to domestic timber consumers) would be the area AEPD, the producer surplus (the gains to domestic beef farmers) would be the area PDEF, and total welfare (the sum of consumer surplus and producer surplus, or the gains to society overall) would be the area AEF. With trade, the consumer surplus decreases to ABPW, the producer surplus increases to PWCF, and total welfare increases to ABCF. Since total welfare is larger (by the area BCE), this represents the gains from trade. So to summarise, exporting beef makes domestic beef consumers worse off (lower consumer surplus), domestic beef farmers better off (higher producer surplus), and society overall better off (higher total welfare).

Now consider how the change in demand from China affects the world market for beef, as shown in the diagram below. World demand has increased from DW0 to DW1, and that increases the equilibrium world price from PW to PW1.

Now, let's go back to the New Zealand domestic market for beef. The world price has increased from PW to PW1, as shown in the diagram below. Now, the domestic consumers have to pay the higher price PW1 for beef, since they are still competing with the world price (and the world price is now higher). At this higher world price, the domestic consumers now choose to purchase Qd1 beef, while the domestic beef farmers now sell Qs1 beef (still assuming that the world market could absorb any quantity of beef that was produced). The quantity of exports is now (Qs1 - Qd1). That means that more beef is now being exported.

What does that mean for economic welfare? With the higher world price, the consumer surplus decreases further to AGPW1, the producer surplus increases further to PW1HF, and total welfare increases further to AGHF. In other words, the increase in the world price of beef makes domestic consumers worse off (which is what the article notes), but it makes domestic beef farmers better off, and society overall better off.

Domestic consumers are affected by events on the world market, when the domestic market is open to international trade. However, openness to international trade is not all bad news for consumers. If international demand falls, the domestic price will fall and consumer surplus will increase (essentially, the opposite of the example above). And, New Zealand is not an exporter in all markets. In markets where New Zealand is a net importer, prices are lower, and consumer surplus is higher, than they would be without trade.

The effect of timber export restrictions on the domestic market for timber

The housing crisis is causing the government to search frantically for solutions. As the New Zealand Herald reported last week:

The Government was warned its efforts to tackle New Zealand's housing affordability issues could be hampered by wood shortages.

The issue has become so significant, Building and Construction Minister Poto Williams is considering limiting timber exports to ensure there is enough in the country.

What happens if the government limits timber exports, by implementing an export quota? Before we can answer that question, we need to consider the effect of exports on the domestic market, without any restrictions on exports. That situation is shown in the diagram below. New Zealand is an exporting country, which means that New Zealand has a comparative advantage producing timber. That means that New Zealand can produce timber at a lower opportunity cost than other countries. On a supply-and-demand diagram like the one below, it means that the domestic market equilibrium price of timber (PD) would be below the price of timber on the world market (PW). Because the domestic price is lower than the world price, if New Zealand is open to trade there are opportunities for traders to buy timber in the domestic market (at the price PD), and sell it on the world market (at the price PW) and make a profit (or maybe the suppliers themselves sell directly to the world market for the price PW). In other words, there are incentives to export timber. The domestic consumers would end up having to pay the price PW for timber as well, since they would be competing with the world price (and who would sell at the lower price PD when they could sell on the world market for PW instead?). At this higher price, the domestic consumers choose to purchase Qd0 timber, while the domestic suppliers sell Qs0 timber (assuming that the world market could absorb any quantity of timber that was produced). The difference (Qs0 - Qd0) is the quantity of timber that is exported. Essentially the demand curve with exports follows the red line in the diagram.

In terms of economic welfare, if there was no international trade in timber, the market would operate at the domestic equilibrium, with price PD and quantity Q0. Consumer surplus (the gains to domestic timber consumers) would be the area AEPD, the producer surplus (the gains to domestic timber producers) would be the area PDEF, and total welfare (the sum of consumer surplus and producer surplus, or the gains to society overall) would be the area AEF. With trade, the consumer surplus decreases to ABPW, the producer surplus increases to PWCF, and total welfare increases to ABCF. Since total welfare is larger (by the area BCE), this represents the gains from trade.

Now consider what would happen if there is an export quota limiting the quantity of timber exports below (Qs0 - Qd0). This is shown in the diagram below. Let's say that the quantity of exports is reduced to the amount between B and G on the diagram (about half the amount of exports that were previously occurring). Now consider what happens to the demand curve (including exports). The upper part represents the domestic consumers with high willingness-to-pay for timber. Then there is a limited quantity of exports that are allowed under the export quota, at the world price PW. After that, there are still profit opportunities for domestic suppliers (that is, there are still some domestic consumers who are willing to pay more than what it costs the suppliers to produce timber). So, the demand curve (including the export quota) pivots at the point G, and follows a parallel path to the original demand curve (i.e. the demand curve including exports follows the red line in the diagram). The domestic price is the price where supply is equal to demand (P1). The domestic consumers choose to purchase Qd1 timber at the price P1, while the domestic suppliers sell Qs1 timber at that price. The difference (Qs1 - Qd1) is the quantity of exports. Notice that the price of timber that timber consumers pay has fallen, and more timber is purchased domestically - we'll come back to those points shortly.

Now consider the areas of economic welfare. The consumer surplus is larger than it was without the restricted exports (it is now the area AJP1), the producer surplus is smaller than it was without the restricted exports (it is now the area P1HF plus the area KLHJ. The first area (P1HF) is producer surplus as if the farmers sold all of their products to the domestic market, while the second area (KLHJ) is the extra profits the suppliers get from selling the quota of exports. Total welfare is smaller than without the restricted exports (it is now the area AJHF+KLHJ). There is a deadweight loss (a loss of total welfare arising from the restricted exports) equal to the area [BKJ + LCH] - these areas were part of total welfare with trade and no restricted exports, but have now been lost. The reduction in exports makes timber suppliers worse off, as well as society overall (in terms of economic welfare in total). However, timber consumers benefit in terms of higher consumer surplus.

Now consider the goals of the export quota. If the government is worried that domestic timber prices are too high, the export quota will lower the price (from PW to P1). If the government is worried that not enough timber is available and sold locally, the export quota will increase that quantity (from Qd0 to Qd1). It sounds like the export quota will have all the effects that the government might want. However, there is no free lunch here. Domestic timber producers are made worse off, and by more than the amount that domestic timber consumers gain (we know this because total welfare overall declines).

The negative impact on domestic timber producers is going to create a couple of negative incentives. First, at the margin it will dissuade timber growers from planting forests, because the return on investment will be lower (as timber prices are lower). Of course, that's not going to impact the market until 20-25 years into the future, so the current government might not care. Second, timber growers might prefer to leave their forests uncut, hoping that the export quota is lifted after the next change in government. If prices are low now, but there is an anticipated higher price in the future, then holding back supply might be a good strategy for some timber growers. That will have the opposite effect from what the government intends, because a reduced domestic supply of timber raises the domestic price, and decreases the quantity of domestic timber sold. This effect seems very likely to me.

The government needs to tread carefully, lest they create incentives that actually make the problem worse in the long run. Policy alternatives that encourage timber supply, rather than discouraging it, are likely to be more effective overall.

The benefits of the four-day workweek, in the presence or absence of tournament effects

The four-day workweek was in the news again last week. As reported in the New Zealand Herald:

An advocate of the reduced working week says it's time for New Zealanders to re-examine how they're doing business, and that cutting back hours can improve both productivity and mental health.

Charlotte Lockhart is the chief executive of the not-for-profit 4 Day Week Global project, which she established along with fellow New Zealander Andrew Barnes for like-minded people who are interested in the idea of a 32-hour working week.

She spoke with Sunday Morning about the need for change, and said that technology is making more flexibility possible in many workplaces.

"The reality is we know that the way we're going to work in the future isn't how we've worked in the past.

"There's no reason why we need to be working five days a week. It's not a mandatory number."

Lockhart is right that the five-day workweek is not a mandatory number, and neither is the 40-hour week. However, as I noted last month when I blogged about the four-day workweek in Iceland, there is a key point that advocates of the four-day workweek simply don't understand, and that is tournament effects. As I said in that earlier post:

...when there are tournament effects, people are paid a 'prize' for their relative performance (that is, for winning the 'tournament'). The prize may take the form of a bonus, a raise, or a promotion. The point is that each worker only needs to be a little bit better than the second best worker in order to 'win' the tournament. Those incentives would work to undo the decrease in work hours, since if everyone else reduces their work hours from 40 to 32, a worker that keeps working 40 hours will increase their chances of winning the tournament. If you doubt that tournament effects are real, I recommend asking any serious academic how many hours they work each week (since tournament effects are rampant in academia). This is not consistent with the overall goal of the four-day workweek, which is to reduce work. All it would do in these occupations is shift more of the work to outside of the paid workweek.

In occupations that have strong tournament effects (like academia, as well as professional jobs like lawyers, accountants, finance, professional sportspeople, etc.), shifting to a four-day workweek wouldn't change the incentives for the workers. If you work a little bit more than your peers, then you are more likely to 'win' the tournament and get the promotion, the prestige, the bonus, or whatever other reward the tournament is leading to. If your peers choose to drop down to working four days a week, all the better for you. They'll soon realise the cost of their choice, when the rewards accrue more to those who ignored the four-day-a-week call (just as they currently accrue now to those who work excess hours).

However, not all occupations have strong tournament effects. In some jobs, the tournament effects don't depend on productivity as measured on a weekly basis, but depend on hourly productivity. In those jobs, the workers are reasonably interchangeable as well, so it doesn't matter so much who does the work. Reducing from five days to four days a week for the same pay isn't going to reduce these workers' chance of winning the tournament (and if it increases their hourly productivity, as four-day-workweek proponents claim, then it might increase the workers' chances of winning). Consider service workers, or hospitality workers, or manufacturing workers, or construction workers, or administrators. They aren't competing for productivity-based rewards, or if they are, those rewards depend on hourly productivity, not weekly productivity. A four-day workweek for the same pay is a clear win for these workers (although a five-day workweek with a limited number of hours per day might be even better).

In some jobs, the potential tournament effects are reduced to some degree because promotion or salary advancement or prestige is determined by something other than productivity, like tenure. For example, salary advancement for teachers or nurses is based on tenure. Teachers or nurses don't work extra hours to get ahead of their peers in the salary advancement tournament (however, promotion or prestige might still reward those who work more hours). Moving them to a four-day workweek may be beneficial for workers in these jobs, to the extent that the structure of rewards counters the tournament effects.

There are clearly many jobs where the four-day workweek may have positive effects, but only where the tournament effects are weak, or where they are neutered by the way that the tournament is structured. Unfortunately, those are the jobs that the four-day-week advocates are focused on, to the exclusion of jobs where tournament effects are strong and unavoidable. A one-size-fits-all approach to adopting the four-day workweek is likely to be problematic.

Read more:

• Perpetual Guardian, four-day workweeks and the Hawthorne effect
• Iceland and the four-day workweek