The myth of the Great Resignation (in New Zealand)

I was interested to read this article in the New Zealand Herald this morning:

Companies grappling with labour shortages and discerning job seekers are turning to hefty cash incentives to recruit staff - offering several-thousand dollar sign-on bonuses or finder's fees.

Recruiters say employers facing the "perfect storm" of wage inflation, increased overheads and talent shortages are getting increasingly innovative to catch the attention of potential candidates...

[People & Culture director at Simpson Grierson, Jo Stevenson] explained the pandemic had also made prospective employees more picky.

"People are becoming more discerning about their employment choices and they're wanting to move to a place because it's got the right culture, and the right level of flexibility," she said...

Dunedin's Platinum Recruitment director Dean Delaney said employers were facing the perfect storm of wage inflation, increased overheads and talent shortages.

He said companies were finding sign-on offers a better use of their money, than widespread job advertising.

His clients were reporting very little interest in their job ads.

I found it interesting because it is hard to reconcile labour shortages and difficulty in recruiting with a rhetoric of the Great Resignation, like this from an AUT press release:

NZ employers take note: the “great resignation” is happening here...

Internationally, and especially within the United States, there is a lot of talk about the “great resignation” – the informal name for the widespread trend of a significant number of workers leaving their jobs during the COVID-19 pandemic. 

Ok, let's take a step back and think through what it would mean if we had both of these things (difficulties hiring and a Great Resignation) at the same time. To do so, we need to this through a pretty simple model of how we categorise people of working age (which we use in my ECONS101 class).

People of working age [*] can be classified as being in the labour force, or not in the labour force. Pretty simple, really. People of working age are in one of these categories, or the other. People in the labour force are either employed (they have a job), or unemployed (they don't have a job, but they are looking for one and they are willing to work). Again, this is pretty simple. People in the labour force are one, or the other. People of working age who are not in the labour force are those who both don't currently have a job, and aren't looking for one.

Now think about the 'Great Resignation'. If people of working age are resigning from their jobs in large numbers, then there are only a few things that can happen to them in our model. They start as employed, and then they can either: (1) become employed somewhere else; (2) become unemployed; or (3) leave the labour force. Those are the only things that can happen, unless they are erased from existence.

Now, assuming that Thanos isn't operating in our labour market, let's look at the other three options. If people were resigning and being employed somewhere else in large numbers, then we wouldn't see problems with recruiting, as in the Herald article quoted above. So, clearly, that isn't happening. If people were resigning and simply becoming unemployed, we'd see an increase in the unemployment rate. But, at 3.2 percent, the unemployment rate is the lowest it has been since 2007.

The decreasing unemployment rate also tells us that people aren't simply leaving the labour force either, but to see why it requires a bit more explanation (and a tiny bit of maths). The unemployment rate is the proportion of the labour force who are unemployed. So, because the labour force (LF) is made up of employed (E) and unemployed (U), then unemployment rate (UR) can be calculated as:

UR = U / LF = U / [U + E]

Now consider what happens when the number of employed people decreases, but without an increase in the number of unemployed (as would happen if people leave their jobs and exit the labour force). Unemployment (U) doesn't change, but the labour force (U+E) decreases (because employment (E) decreases). [**] The unemployment rate would increase. But it hasn't. It's at a long-term low, at 3.2 percent.

What does that leave us with? If people aren't resigning and going into new jobs, becoming unemployed, or exiting the labour force entirely, then they must be staying in their current jobs. There is no Great Resignation in New Zealand. Or at least, no evidence of it yet.

[Update]: A couple of keen former students of mine pointed out that there is another possibility: people of working age resign and then move overseas. However, that is effectively the same as erasing people from existence, so if that were the case the unemployment rate should be increasing.

However, if people were moving overseas (lower E), but then being replaced from the ranks of the unemployed (lower U, higher E), then resignations combined with new hires could reduce the unemployment rate. But that wouldn't necessarily be consistent with the difficulty that employers are facing in finding people. While net international migration is negative, it has been around 8000 people per month over the last few months, which is about 0.2% of the labour force per month. Not all migrants are of working age, of course, but it could still be a contributing factor to a lower unemployment rate. Maybe there is something to the Great Resignation, after all? 

*****

[*] Let's not get bogged down in how 'people of working age' is defined. It actually makes no difference to the rest of the story whether working age is defined as 15 to 64 years, or 15 years and over (the two most common definitions), or whatever other definition you want to apply to it.

[**] Interestingly, the effect on the unemployment rate of people leaving the labour force is exactly the same as if Thanos erased employed people (but not unemployed people) from existence.

More evidence that life satisfaction may be dead

Back in 2020, I wrote a post that questioned the foundations of subjective wellbeing, based on this article by Bond and Lang. They essentially showed that the conclusions that are drawn from studies of subjective wellbeing (or happiness) depend on how the ordinal variable (subjective wellbeing, being measured on a 0-10 scale) is converted to a cardinal variable. They concluded that:

It is essentially impossible to rank two groups on the basis of their mean happiness using the types of survey questions prevalent in the literature.

It seems that Bond and Lang's result is not an aberration. I recently read this 2017 article by Carsten Schröder (Free University Berlin) and Shlomo Yitzhaki (Hebrew University), published in the journal European Economic Review (ungated earlier version here), which came to a very similar conclusion. Schröder and Yitzhaki first rightly note that:

Well-being (life satisfaction or happiness) is a latent variable that is impossible to observe directly and that has no natural quantitative measurement unit. Data are collected in scientific surveys that ask questions like, “All in all, how satisfied are you with your life at the moment?” Respondents answer these questions by ranking their satisfaction levels on a pre-defined scale, usually ranging from 4 to 11 points (and sometimes more), with the individual points assigned terms such as “very bad,” “bad,” or “good.”... In a scale like this, we know that “very bad” is lower than “bad” and that “bad” is lower than “good”, but it is not clear whether the distance between “very bad” and “bad” is greater or smaller than the difference between “bad” and “good”. The ordinal nature of this data means that any monotonic increasing transformation of the scale is allowed.

They then apply some simple monotonic transformations to subjective wellbeing data from the German Socio-Economic Panel (SOEP), and the results are not good (at least, for those who hope to use subjective wellbeing and want it to be meaningful). IN a simple comparison of the gender gap in life satisfaction, Schröder and Yitzhaki find that:

...in only one case is a simple comparison of average satisfaction robust to monotonic increasing transformations. In the other 47 cases, an admissible transformation that can reverse the sign of the gender gap in satisfaction exists...

In other words, they can reverse the gender gap by simply transforming the data in a way that shouldn't affect the results. So much for simple comparisons of mean differences. Looking at regression models, they also find that:

...using random effects does not prevent the reversal of coefficients. Using fixed effects avoids reversals but the significance of the coefficients hinges on the transformation. Note, however, that the robustness of the sign of the fixed-effects regression coefficients to monotonic increasing transformations, in general, is not guaranteed. Reversals of signs of coefficients can be shown for this type of model as well...

So, pretty standard linear regression models aren't robust. What about ordinal models? After all, they are designed specifically to deal with ordinal data. Schröder and Yitzhaki first note that:

...they result, by definition, in non-intersecting cumulative distributions. That is, these models assume a dominance relationship of the cumulative distributions that the raw data do not necessarily support...

Then they find that, comparing life satisfaction between German and non-German respondents:

In three of the five significant cases, the empirical distributions intersect. Of course, the importance and frequency of violations of dominance depend on the empirical context.

So, the distributions intersect, and so the ordinal models are finding statistically significant differences when they should not. Overall, the results are not flattering for users of subjective wellbeing data, and should especially be of concern to those who are trying to look beyond GDP as a measure of wellbeing (as noted in my recent book reviews here and here). No doubt there is more to come on this topic.

Read more:

• Happiness is dead?

The relationship between alcohol outlets and crime is not an artefact of retail geography

There is a vast amount of literature that links the location (or density) of alcohol outlets with crime, particularly violent crime. I've contributed to this literature on a number of occasions (see here and here and here, for example). However, one criticism of this literature is that, while it shows that the number or density of alcohol outlets and crime are correlated, it doesn't demonstrate causal relationships between alcohol outlets and crime. These studies are all based on what are termed ecological relationships. We see more outlets where there are more crimes, and we infer that the outlets cause the crimes. There are good theoretical reasons to believe this relationship is causal, but the statistical tools that are used don't demonstrate this definitively.

The relationship between alcohol outlets and crime is something I've been working on during my current study leave period, which is due to finish soon. I'll be presenting some causal analysis (based on instrumental variables regression) at the New Zealand Association of Economists Conference at the end of this month. However, that isn't the only important work I've done in this space over the last year.

One of the key criticisms of the ecological literature on alcohol outlets and crime is that the spatial pattern of alcohol outlets is essentially the same as the spatial pattern of retail activities more generally. Alcohol outlets tend to locate in the same places that retail outlets of all types tend to locate. So, any observed relationship between alcohol outlets and crime might not be due to the alcohol outlets themselves, but rather due to retail outlets generally. This criticism can be backed up by a theory in criminology known as routine activity theory. Routine activity theory says that crime occurs when there is a motivated offender and a potential target that are in the same place at the same time, in the absence of a capable guardian (that is, someone who can prevent the crime either directly or indirectly). It can be argued that areas where there are retail activities are mostly deserted at night, reducing the number of effective guardians, and therefore making crime more likely.

Some years ago, I had a couple of summer research students look into the relationship between a range of different retail outlets and violent crime in Hamilton. I recently followed that up with some more thorough and detailed analysis, and the results of that have just been published in the influential journal Addiction (sorry, there is no ungated version online). I used data on the number of alcohol outlets (by type) from the Ministry of Justice, data on the number of non-alcohol outlets (by type, separately for licensed clubs (like sports clubs), bars and night clubs, other on-licensed outlets (like restaurants and cafes), and all off-licence outlets (including bottle stores and supermarkets)), and data on violent crime calls-for-service and total calls-for-service from the New Zealand Police, all measured at the area unit level (approximately suburb level) for Hamilton. The non-alcohol outlets data were painstakingly collected by my research students, by visiting every commercially-zoned area in Hamilton and observing the types of outlets located there. They collated data on the number of bakeries, hairdressers, service stations, and takeaway food outlets, for each commercial block, which we then aggregated up to the area unit level.

My analysis was conducted in a couple of steps. First, I find that the number of outlets of each type (separately for each type of alcohol outlet and for each type of non-alcohol outlet) are highly correlated with each other. That suggests that it would be difficult to disentangle the effects of alcohol outlets from the effects of retail more generally, as expected. This demonstrates the problem that a lot of the literature faces.

Second, I use the data on the non-alcohol outlets to create a measure of retail density for each area unit, and show that, after controlling for non-alcohol retail density (as well as population, local demographics, and social deprivation), statistically significant semi-partial correlations remain for each alcohol outlet type and both violence and total calls-for-service. That tells us that the relationship between alcohol outlets and crime is not purely an artefact of retail density, since alcohol outlets can explain some of the remaining variation in crime even after retail density is accounted for.

The contribution of this paper seems small, but it is likely to be important. While this research doesn't quite get us to a demonstration that alcohol outlets cause crime, it does at least eliminate one of the common counter-arguments against the literature. Now we need to follow this up by applying alternative methods to demonstrate causality, and that is what I am working towards. Watch this space.

Book review: For Good Measure

Back in April, I reviewed the book Measuring What Counts, by Joseph Stiglitz, Jean-Paul Fitoussi, and Martine Durand, which summarised the work of the High-Level Expert Group (HLEG) on the Measurement of Economic Performance and Social Progress, established by the OECD in 2013. I made the point that the summary "lacks some of the detail necessary to fully understand some of the points that were made", and noted that there was a companion volume, For Good Measure.

I have now read that book, edited by Joseph Stiglitz, with Amartya Sen and Jean-Paul Fitoussi, and with contributions from members of the HLEG. Indeed, it does contain much of the detail that the summary lacks, but is a bit uneven in what it offers. For example, the chapter by Francois Bourguignon on inequality of opportunity includes a good framework for understanding the different between inequality of opportunity and inequality of outcomes, and how they are interrelated. However, much of the chapter is quite mathematical, and the general reader may be left quite confused. In contrast, I thought I wouldn't get much out of the chapter on measuring sustainability (by Marleen De Smedt, Enrico Gionvannini, and Walter Radermacher), but it turned out to be a highlight of the book. In particular, the discussion of the differences between absorptive capacity, adaptive capacity, and transformative capacity, was quite illuminating.

Other chapters that I found interesting included the chapter on subjective wellbeing (by Arthur Stone and the late Alan Krueger, who passed away before the book was published), and the chapter on trust and social capital (by Yann Algan). The former chapter has an excellent discussion of the challenges and concerns with subjective wellbeing, and future directions for measurement and improvement. The chapter on economic security (by Jacob Hacker) included some data on New Zealand that I found interesting - that 'economic vulnerability' is relatively high (measured as the proportion of the population who are not 'income poor', but have equivalised wealth below 25 percent of the income poverty line - that is, they have less than a three-month buffer of wealth), but income poverty is relatively low (measured as the proportion of people with equivalised income below 50% of the median income). With that in mind, we could definitely do with better understanding of economic insecurity in New Zealand, and perhaps policy settings that are more attuned to reducing insecurity. I wish that Hacker's chapter had been available when I had a PhD student looking at economic insecurity in Timor Leste some years ago, as it would have helped to better frame his work.

A lot of the book is devoted to inequality, and I found those chapters not to contribute much. Readers who are less aware of the current state of the literature and measurement of inequality will no doubt find those chapters much more rewarding than I did. We also didn't get an answer to the question of how many indicators are too many for a dashboard, which was an outstanding issue from my review of the summary book. This book provided less of a general critique of GDP than the summary book, except in the first chapter, and that reflects the difference in audience for this more technical book. I think that made for a better product overall. The only thing the book really lacked was a concluding chapter to bring it all together. However, I guess in some way the summary book fulfils that role. 

Despite those minor gripes, this is a good book for those who want to go a little bit deeper into the research that underlies the HLEG work as outlined in the summary book. And going even further in particular areas can easily be achieved by following through to the well-referenced material in this book.

Despite the subtitle of this book, it does not really provide "an agenda for moving beyond GDP". It would be more accurate to label it an agenda for improving the measurement of wellbeing generally. Maybe I'm being unfair - being better able to measure wellbeing might help us to move beyond a singular focus on GDP. If you're wanting to understand how we might do that, then I strongly recommend this book as a good source for you.