Look I’m from New Zealand – so when everyone around me started talking about ATARs I just smiled and nodded.

In fact I probably couldn’t talk to most people in New Zealand about education. I’m from the “pre-NCEA” era – where a single end of year exam for five courses, scaled to fit a within-course normal distribution, determined most of anything. As a result, these more complex design criteria are well outside of my lived experience.

But it turns out ATARs are a very important part of an individuals assessment in Australia – providing a measure of how well a student performed relative to their peers, and determining their university admission.

This raises a question – how is a good performance on your ATARs associated with future earnings? Luckily for us Elyse Dwyer and Silvia Griselda at e61 decided to find out.

tl;dr a higher ATAR is associated with higher average earnings – but there is significant variability in income by ATAR. As a result, even though we’d expect the type of person who receives a higher ATAR to end up with higher earnings at 30, there is a lot more going on under the hood.

ATARs and earnings

The earning profiles of individuals between 19 and 33 are shown below, with each ATAR bucket also conditional on the individual going to university (well except the No Uni one). Lets chat about them

As we can see, the earnings of those who do not go to university is higher between the ages of 19 and 23. Anyone else who went to university when most of their high school mates went to work can attest to this truth – at this stage people love to go on about how they are not earning cash while your still at school.

Nonetheless, continuing education is a form of investment – so you should expect a payoff in future earnings. And that is what you see, all the other lines move above the no university line, showing that the median individual who went to university did end up earning more than the median individual that didn’t – irrespective of their ATAR.

Then it comes time to compare all the other lines. As we can see, the median individual with a higher ATAR earns more as well. And all individuals that used their ATAR for university entry earn more than those who entered through alternative pathways. So what does this tell us?

1. Individuals with higher ATARs may have characteristics that lead to higher income – harder work ethos, greater capability, better family and friend networks.
2. Higher ATARs may allow individuals to select into higher value educational opportunities.
3. There may be a direct return from having a high ATAR on your CV above and beyond everything else.

The note digs into a few more mechanisms, the main thing I want to note is that no-one is saying the higher ATAR itself causes higher earnings – just that there is an association worth further investigation.

And figure 2 shows that these things get more complex – there is significant dispersion in the earnings of individuals at 30 on the basis of their ATAR.

For a specific example, someone in the bottom 25% of earnings for the top ATAR bucket (the 25th percentile) earns less than someone in the 25% of earnings who didn’t go to university. And they earn about the same as the median “below 70” ATAR individual who did go to university!

I remember that I had friends from high school who went straight into work – and now own and run the business they were working at. I had friends who were clued on whose grades ended up a bit below mine, who went to university and earn significantly more than I ever will.

Unlike how high schools make us feel, that final grade doesn’t determine our full life path and opportunities – it influences it, but even if high school went poorly there are opportunities out there.

Not just that – never confuse a grade or your income with your self worth. You matter just as much as someone everyone calls smart, or someone who seems to have limitless resources – and the person who you view as unsuccessful matters just as much as well. Your power comes from being able to make your own choices, and own the outcomes – so make these choices with kindness.

ATARs and the SAT debate

Can this information about ATARs tell us anything about the standardised testing debate in the US at the moment? Probably not, but its worth thinking about.

The ATAR takes an average of course performance – like a GPA does. However, it scales the scores based on measures of how difficult that subject is – which is more similar to the weighting used for an SAT.

Concerns about implicit discrimination and the interuptions from COVID led to a large number of US univeristies dropping SAT requirements. However, the NY Times has come out stating that – relative to other metrics universities can use – standardised testing does appear discriminatory. Specifically, the advantage of advantaged students is starker when comparing GPAs to SATs.

To understand this issue we’d want to think through two issues:

1. Is it important: What is the return associated with gaining university entrance.
2. What drives differential selection by group: What is the reason why individuals from certain backgrounds are not being selected into study – and does that choice influence the benefit or opportunity cost of such study.

To understand the importance of univeristy entrance we’d want some idea of what would have happened to people who gained entrance if they didn’t – or people who didn’t gain entrance if they did.

To understand selection we would want to look into not just university selection criteria, but the nature of applications and the decision to study. If individuals from disadvantaged backgrounds are not studying because of other family responsibilities, then the policy choice of “removing SATs” isn’t really dealing with the lack of opportunity faced by that person – it is instead a lack of material resources to be able to make worthwhile investments in their human capital.

I think the e61 note informs this by showing just how heterogeneous these returns can be – hinting that such investments in the self may be very uncertain. SAT bans don’t do anything about that, and if they do restrict access to scholarships (say due to a shift to sport and GPA based support) they may make access worse for truly disadvantaged individuals.

So I guess the work did provide some insights to this debate after all!

COVID was a pretty big life shock for all of us – with uncertainty about the virus, and the magnitude of government interventions associated with it, leading people to change their way of doing things across a number of dimensions of their life.

One of the clearest changes was the decision of where to live. In a fascinating note from e61’s Aaron Wong from mid-2023 he shows people did change their behaviour regarding renting property following COVID – and that this shows up clearly in the rental premium paid for living in cities between 2020 and 2022.

The decision on where to live: Rental premiums

The decision about where to live depends on a number of factors. Where are the jobs? Where is my job? Where is a good school? How good is public transport? Am I near a nice park? Where can I get a coffee? How can I see my friends on the weekend?

The varying amenities associated with a location mean that different individuals sort into very different locations. However, the bundle of amenities available near city centres tends to be of higher value than those that are a long way away for the average person. This is a logical result given that, for most, it will tend to be closer to work, and provides more opportunities to hang out with friends and go shopping.

This is what Aaron finds – but something changed in mid-2020!

After COVID kicked off the discount associated with living further away from the a capital city/centre of a city dropped.

So why would the relative value of living in the city decline? This could come from the value of being distant rising (i.e. if I’m away from everyone I’m less likely to get COVID!) or the value of being central falling (i.e. I’m stuck in my house so I can’t make use of restaurants anymore).

Examples of this are:

1. Working from home (travel time): With working from home required, the benefit of lower travel time is gone – removing a major benefit of central living.
2. Working from home (house size): Central city properties tend to be smaller than properties available further away – and it is nice to have an extra room when working from home. [Note: The differences shown above control for property characteristics like the number of rooms]
3. Working from home (internet quality): Internet quality is very variable in Australia – but is generally a lot better in cities than in regional areas. This will have increased the relative value of living centrally.
4. Same stuff as above – but about enjoying your leisure time!
5. Fear of COVID and population density: The centre of the city has more people in it – more people, more COVID. Plus one to living regionally.
6. Freedom during lockdowns: Lockdowns tended to be more severe in central city areas than in regional areas. Another relative benefit in the regions.
7. Availability of groceries: COVID was a period where supermarket deliveries picked up. In central cities this was very available – but was oversubscribed meaning that you would spend time lined up at supermarkets. In regional areas it was variable. It is unclear if the relative value changed on this margin.

Is this a permanent change?

Intuitively, the premium is likely to have started returning over 2023/24 as COVID has moved into the past. But if there is a step-change in working from home the city premium may be permanently gone.

Is there anything in the analysis that allows us to work out what is going on? Well this is where the appendix has our back.

House prices reflect the expected rental/housing service yield associated with a property over time. If the change in the rental premium was expected to be permanent then we would expect to see relative house prices also change.

But they haven’t really (apart from a small tick up for fairly distant properties – but not to the same degree as the rent change).

This suggests that the current reduction in the regional rent discount is expected to be a temporary phenomenon – and we should see the gap between city and regional rents reopen in the coming years.

Does this mean that households don’t expect a long-term “Nike Swosh” in working from home (contrary to Nicholas Bloom’s expectations)? Maybe, but that is a topic for another time.

I keep receiving emails that Star Wars Squadron is only $2.50. Wild! I love Star Wars, and when I was a kid I loved flying around all the planes and stuff.

But why such a big special? And why are lots of computer games on a big sale now?

I mean surely this is the time when demand is high (a lot of people are on holiday) and computer games are a durable good (you pay up front, and then consume them over time).

As a result, this must be the time when people are most willing to pay for it – and the incentive to discount must be pretty weak because if you sell a game to someone now, you can’t sell it to them again in the future. For more fun, try link slot games where you can enjoy exciting gameplay and chances to win.

So let’s have a think about why this may be going on – and I promise not to talk about tacit collusion this time. Instead lets talk price discrimination.

A tale of three consumers

Imagine we have three people who all get excited when they see the following image.

One of these people is a 16 year old from a high income family, and a social network of friends in a similar position. Access to a new game depends largely on payment by your parents due to the need of a credit card – even if you’ve gone off to earn the money yourself.

In terms of time you have plenty of time to play games where you would otherwise be practicing hand stands and breaking windows. As a result, such purchases tend to occur when you are able to access them as an external gift – such as during a birthday or at a game launch when you can point out your mates have gotten it.

In this situation, your demand does not really vary with prices, and you are exceptionally impatient about having the game at release.

Another person is a 21 year old university student with fairly strong credit constraints. Although your time is a bit more constrained than at 16, you still do have a variable amount of leisure time – focused specifically at study breaks. Furthermore, others that you will play with are available at similar times – so individuals will coordinate on games.

This combination of factors means that demand can vary quite sharply with price, as groups of students coordinate on different games on the basis of their price.

Finally you have a 35 year old professional with a family, significant work commitments, and some disposable income. Their network is also similar people. Work and child requirements are not so coordinated, and as a result there is less coordination between people about playing together.

Furthermore, you have very limited free time – which means that you don’t really expect to get too much value from a game, as you doubt you’ll have much time to play it. However, you do have significantly more income.

This type of consumer has relatively low price elasticity … but for a game that is multiplayer focused also has a much much lower value for the game.

All three of these individuals would enjoy playing the game – but they differ in the opportunity costs they face from doing so.

• 16 yos: Material resources are available at fixed times, and plenty of time. So the opportunity cost is associated with the specific time of purchase (a form of limited material resources).
• 21 yos: Have limited material resources, but plenty of time. As a result, the key opportunity cost is other things they could spend money on.
• 35 yos: Have material resources, but no time. As a result, the opportunity cost is more associated with not having time available.

Price discrimination and our Steam sale

Our firm is selling a piece of software online. The marginal cost of doing so is close to zero.

As a result, in a perfectly competitive industry the price would collapse to zero. However, such an industry wouldn’t be sustainable – as no-one would make these games. So some amount of market power exists which generates the revenue to cover fixed costs as well.

We know that the game seller is setting marginal revenue equal to marginal cost – to sell an extra unit they have to cut the price to everyone, and they do this until the “extra dollars” they get become zero.

If they had to set the price and leave it, then we can look at our three groups to figure out the price:

1. If we only sell to 16 year olds we can set the price pretty high!
2. If we include the 21 year olds we run into the issue they can’t pay as much – so we have to sell at a lower price for both.
3. If we include the 35 year olds, then they aren’t willing to pay much at all – so we have to set the price very low to cover all of them.

So the seller will compare marginal revenue and cost, and determine what group is the marginal group. They will only sell to that group and the groups willing to pay more, and the groups with a lower willingness to pay will not receive the product.

Lets picture that the price was $50, and it was only sold to 16 year olds in this instance.

This is “inefficient”. Why? Because it costs $0 to provide the game, people like the 35 year olds do have a small positive value from the game, and yet there is no transaction between the game seller and the 35 year old!

Ohhh no, lost gains from trade!!

So why not charge them less based on the age on their passport? Well we can’t age discriminate. Steam offering different prices based on a persons age would lead to manipulation and is illegal.

But instead we can achieve this by changing prices through TIME – based on when the game is released, and when it is discounted.

A $2.50 sale after Christmas – when the 35 year old is on leave from work, everyone else is getting gifts, and they have had some fun playing with their kids toys – is pretty appealing. If the retailer has already saturated the 16 year old and 21 year old markets, then they are really just servicing 35 year olds when they do this sale.

When the game was released it was $50. The game was advertised, and in some places sold with, a joystick. A lot of 16 year olds love spinning their ship around and around to annoy their friends when playing with them (speaking from experience) – so there is significant value from buying at release when there are a lot of people to do this to, friends and people online. This was in October 2020.

In the three years inbetween, there were plenty of sales – as well as a lot of negative reviews regarding bugs and a lacklustre story. Whether these things matter I don’t know, as I’ve never played or watched the game, but for our price sensitive 21 year olds it will have taken a number of school holiday sales to convince them to give it a go. $20 or $30 sales may have worked for this group.

If this is a correct description of the market, through time based price discrimination the retailer can charge 16 year olds who want the game at launch $50, our 21 year olds who are willing to delay $20, and the relatively disinterested 35 year olds $2.50 – providing the game to a set of different groups who value it, and making the highest profit.

Now we can’t just randomly set the prices lower at different dates as it is a durable good. If everyone knows it will be $2.50 on a certain day, then why would they pay $50!

• It is uncertain if, and when, the game will become $2.50 – hell it took 3 years!
• People may value having the game early – or more specifically, coordinating when they have it to play with other people. The people who value games in this way will pay more for receiving the game at an earlier date.

As a result, the ability to price discriminate is limited by the ability for a high paying type (16 year olds) to tag themselves as a different type (i.e. a 35 year old). If the $2.50 sale was one month after release, the 16 and 21 year olds would likely delay the purchase by a month.

But look at when the $2.50 sale has occurred – three years after game release, at the start of January, after Christmas when middle aged people who don’t commonly buy games are sitting around thinking about their youth.

Great time to convince them to drop half the price of a coffee on something they will never open or use … after customers that were willing to pay a bit more have already paid.

At the close of 2022, Gulnara and I thought it would be fun to pop up some forecasts for 2023. So what were they and how did we do?

The forecasts

It appears we made four forecasts:

1. That the “rebalancing towards consumption” in China would dominate the discussion post-COVID economic adjustment – making it harder to bring inflation down in a low cost manner and keeping tradable inflation high.
2. Automation would be a major theme – with a clear visual number of worker tasks across a number of jobs now being clearly automated.
3. Industry policy would be the go to policy lever pointed at given the two above economic themes.
4. However, there would also be increasing discussion of unemployment benefits – and the importance of the safety net for supporting these transitions.

Right, nice consistent story – two changes in the underlying economy, two policy responses.

How did we do? I’m going to self-evaluate – but super keen for everyone elses thoughts in the comments!

Self-evaluation

The importance of rebalancing

4/10

The only reason this gets close to a passing grade is because China and the IMF have been talking about rebalancing a lot and terming 2023 the year of rebalancing.

However, here are the ways this has failed as a forecast:

1. Consumption shares had been falling in the lead up to 2023, and it isn’t clear they’ve risen.
2. Instead of rebalancing – growth and inflation in China have slumped.
3. Exports have held up – even if growth has been weak.
4. The Chinese housing market is imploding.

And then it appears the Federal Reserve has enginnered a soft landing – with inflation coming down and unemployment staying low – without any interference from Chinese rebalancing.

Maybe 4 was too generous – but I am willing to take that it is clumsy internal rebalancing that is creating woes for the Chinese economy, as discussed in our podcast. If you want to be harsher feel free to be in comments

Automation as a key theme

9/10

The first signs of ChatGPT being effective for mass consumer use came in late November last year.

Since then it has improved in power massively, can generate images, can read PDFs, and is actually able to make readable latex tables!

However, this is not just a story of ChatGPT. On Twitch automated streamers are producing content. In the local retail stores autonomous lawnmowers are taking over more household tasks. And I’ve heard that GitHub copilot is a game changer for coding – something I should really look into given how much of my work is coding nowadays!

The influence of automation on service work has become increasingly clear – and 2023 was a watershed year for this.

Industry policy

3/10

The terrible decision to pay movie and game producers have continued, and industry policy has remain an important topic among the chattering class.

But with supply chain disruption easing, any talk of industrial policy went with it. Instead – at least in the Australian context – there has been a lot more chat about mobility and job opportunities, rather than a determination for government to determine what those jobs should be.

I had anticipated a lot more friction here.

Unemployment benefits

7/10

In the first half of the year this looked like a steller forecast – over here in Australia there were significant changes in the Federal Budget, an Employment White Paper that talked strongly about benefits, and wide discussions about how the safety net could support transitions.

Our own work on the income support system, age-based targeting of payments, and employmemt responses to the financial incentives from benefits led to a lot of interesting and important discussion about these systems.

However, a hot labour market does start to push the unemployment benefit system into the background of peoples thoughts. And in the last quarter of 2023 it has become clear that people aren’t really thinking about these systems now.

Once a recession happens, or once we see a sharper increase in unemployment, then the discussion will reopen – but the view that we need to think about whether the unemployment benefit system is fit for purpose for looking at labour market transitions is quite “there” yet as a public discussion item.

However, the significant increase in discussion about how the welfare system influences work choice and the return of chats about EMTRs (effective marginal tax rates) definitely made this an ok forecast

So what are the themes for 2024

We’ve talked about some of them. But still have to finish our projections off – let me have a think and I’ll be back to you in early January . Happy New Year!!!

It’s Boxing Day – and I’ve already used up the idea of discussing the now largely missing Boxing Day sales. And the idea that Boxing Day is increasingly lame is common across years.

So instead the focus of today will be on a different industrial economics topic – the state of competition in Australia. Specifically, does the lack of Boxing Day sales tell us something about the level of competition in Aussie?

For this I’ll be borrowing heavily from a Research Note on the topic by my stellar e61 colleagues Dan Andrews, Elyse Dwyer, and Adam Triggs (now of Mandala). I’ll also be relying on this paper by Jonathan Hambur from the RBA.

However, all misunderstandings and inappropriate comments are my own.

Competition in Australia

Yelling that businesses aren’t competitive enough is a great way to get attention, and push for “getting something for nothing”.

What do I mean? If firms are very uncompetitive then we know they are generating an economic rent – and that this rent could be redistributed to someone (consumers, workers) without undermining economic activity.

As most of us are consumers, the unfairness associated with being charged what we view as too much makes this narrative pretty appealing – and a variety of thought leaders who want to persuade us love nothing more than packaging up ideas we find appealing, and repeating them to us ad nauseam.

However, that doesn’t mean there isn’t some truth to concerns about competition. In fact, it is in the interest of people who are earning economic rents to pretend that daring to tax these rents, or reduce barriers to entry, will undermine some magical value they create – and will use the language of productivity and innovation to make us scared of doing anything.

So given that public narratives are tarred with ideology, lets step back and work things out for ourselves – by looking at some data.

When looking at the State of Competition, the e61 authors focused on the degree of concentration in Australian industries – specifically they focus on the market share of the four largest firms in every Industry Group (3-digit industry code).

There are two graphs I’d like to focus on – noting that the Research Note adds a lot more detail if you want to give it a read!

The graph above shows that these concentration ratios are higher in Australia than the US – so Australian industry groups tend to have a greater share of sales that occur among only four firms. Now this could be partially the result of Australia’s smaller markets.

But then we come to the graph below – concentration ratios have risen across a number of industry groups ESPECIALLY retail sales and mining. As our focus today is on retail sales, this tells us that retail sales activity has become more concentrated among a small number of firms.

The other paper tells us that this rising concentration has also been associated with rising margins earned by firms. However, when compared to other countries that increase in margins has been a bit smaller. Margins in Australia are high – but it isn’t clear that this has been a worsening issue over the past 20 years.

What do bad sales tell us about competition?

So Australian industries are concentrated, margins in Australia are fairly high, and judging by my shopping attempts today Boxing day sales are lame. But does this actually tell us anything?

Concentration alone is actually not telling us as much as we hope. It is possible to have 100s of reasonably sized companies in an “industry” but for each to have significant market power – due to the spatial or product dimensions varying between firms. On the flip side it is possible to have a single firm but very little market power (effective competition).

Margins are also tricky, as the necessary return for entry to the industry depends on risk and the needed return to also meet fixed costs – post Global Financial Crisis there has been in lift in both uncertainty and regulation, both factors that require higher margins to incentivise entry.

But does firms reluctance to provided good boxing day sales mean there is a competition problem?

Intuitively this makes sense – businesses look like they are colluding (even if tacitly) to avoid giving me a cheap Xbox.

However, what is boxing day? It is a period when everyone is off work and goes out shopping – there is a huge surge in demand, and people have the time to shop around.

Given this description, this is when a firm with market power would price discriminate by cutting prices to attract these customers who are relatively more willing to substitute between firms. While more competitive firms would simply face a lift in demand and thereby charge higher prices.

So what is it – are the bad boxing day deals a sign that competition in Australia is stronger than in the past, or is it a sign of a competition issue?

Tacit collusion – a tale of two models

OK so we are going back to our description of Boxing Day in 2019, and one of our favourite descriptions of Oligopoly collusion traditionally.

The difference is that this time we are focusing on whether Boxing Day is telling us anything about competition – specifically, does it tell us if retailers are tacitly colluding in setting their prices during Boxing Day?

If we accept the Green and Porter (1984) model of tacit collusion and competition, then firms collude until they think other firms are cheating (a trigger strategy). During Boxing Day the stores are full up, and so there is no need for sales – but as soon as demand drops, firms think that this is because other businesses have “defected” and stolen customers. As a result, the sales will kick off when the stores are not so busy.

In this world, the disappearance of Boxing Day sales is consistent with collusive behaviour – but we’d expect to see some big New Year sales.

If we buy the Rotemberg and Saloner (1986) model of tacit collusion and competition, then tacit collusion breaks down during high demand states – as firms face an out-sized incentive to try to steal customers. In this model, a collusive equilibrium would involve situations where there are big sales following observable increases in demand – like Boxing Day.

We all know what Boxing Day is – as a result, the second model appears more appropriate for understanding whether we generally have collusion. But if we buy this model, it would suggest that one of two things must be true:

1. Competition has increased and as a result there is no long tacit collusion that falls apart on Boxing Day.
2. Competition has declined, and as a result retailers can maintain tacit collusion even when the reward for reneging is huge.

Teasing out the difference between these is pretty tough. When it comes to the service station industry, David Byrne from University of Melbourne has shown that tacit collusion has increased among petrol retailers, with a related e61 event study reinforcing this.

But without more information we just can’t say which way things have gone among retailers in general – such is the nature of industrial economics!

Sorry, no conclusions for today. If it makes you feel better take this post and demand your local retailer reads it – in order to avoid doing so they’ll probably give you a discount

I’ve been seeing a lot of the following around the place:

Looks pretty scary right – red columns go lower than during the Global Financial Crisis :O

However, these big red columns are a bit misleading. The data is true – financial institutions holding government bonds have made large unrealised losses on government bonds due to the surge in interest rates. But there is a reason that financial regulations don’t make banks recognise these losses.

Now I don’t know much about much, so I’ll talk through some of the basics of how these things fit together and how I understand it. When there is something important I’ve missed feel free to point it out – but I think after the discussion we’ll at least be in a less misleading place than “giant red bar on Twitter”.

tl;dr: Unrealised losses on government bonds for banks are not really a concern. The real concern would be if there was a genuine increase in bad debt/default. However, the existence of these unrealised losses may make financial markets more sensitive to shocks.

A stylised bond

Imagine a world where you are deciding what to do with a spare $100 you have sitting in your pocket – and you’ve made a decision you want to save it to spend in a year.

In this make pretend world you have four options:

1. Change your mind and spend it – if the product is durable you could sell it in the future.
2. Hold the cash.
3. Pop the cash in a bank and earn interest.
4. Purchase a fixed-income instrument that matures in a year – i.e. a government bond.

To keep things simple, lets focus on your choice among the last two – say because the things you can buy are really expensive to store or resell, and cash is expected to fall in value through time due to inflation.

If I pop the money in the bank, then a year later I get it back with interest. Lets assume an interest rate of 5%, that means that in 1 year I have $105. Nice.

Now say that there is a bond sitting there that is completely riskless. It costs $100, pays me a $5 coupon during the year, and at maturity will pay me $100 (the face value). Well buying this bond involves saving $100 now and receiving $105 in the future – so it looks like the same thing, with a 5% interest rate.

If interest rates suddenly doubled the equation changes. The doubling in interest rates mean that, if I put the money in the bank I will end up with $110. However, in the bond example I have $105. As a result, I’m not buying the bond – I’m going to save the money.

In fact, a lot of other people will do the same thing – reducing demand for bonds. People will be willing to sell the bond until the yield on the bond (amount received in a year divided by the amount paid now) is the same as the interest rate. As the coupon payment is a fixed amount of income the price of the bond must adjust.

This implies that the price of the bond must decline to around $95.45 (since this amount plus 10% is approximately the $105 received in a year).

Now assume that the interest rate occurred after I had purchased the bond. In this case I am holding a bond that I paid $100 for, but the price has fallen by 4.5%!

In one sense I have lost money – if I went and sold the bond right now to consume right now, I would have less than the $100 I started with.

In another sense, nothing has happened – if I was always going to wait to maturity I am going to end up with $105 in the future, just like I was anticipating before the interest rate change. In this way I haven’t lost anything and my financial situation is purely unchanged.

If bank are just holding these bonds to maturity – in the way they are with other loans – then these losses don’t effect anything, they are just paper losses. However, if they intended to buy and sell the bonds as a means of liquidity management in the here and now this wouldn’t be so good.

Is this a concern?

According to people on twitter the sharp drop in the market value of bonds is a concern. Why? Big red line mate – time to load up specifically on my crytpocurrency.

To be more specific, the inference is that this unrealised loss is still a loss – just like individuals defaulting on loans, this should be seen as a reduction in the value of assets of the firm.

A reduction in the value of assets implies that – at a point in time – a bank will have less available to meet its liabilities (i.e. our deposits at a bank). This could play out in a couple of ways:

1. Since assets = liabilities + equity a sharp drop in the value of assets may make the net equity in the bank negative – making it insolvent!
2. Even without insolvency, if concerns about these unrealised losses led an increasing number of individuals to withdraw this would run down banks liquid reserves, and increase the chance they have to sell bonds and realise the loss – leading to a self-fufilling bank run.

The first case is a pretty much non-issue. As noted above, if the bond is held to maturity an unrealised gain or loss is just a “paper” change in value – it does not change the real financial situation of the bank.

The issue only occurs if the bank is forced to realise this loss.

So say there is a sharp increase in demand for funds from depositors, how does this play out.

• Seeing that there is an increasing demand for funds, retail banks need to meet this.
• If inter-bank lending is also limited (due to other banks feeling similar pressure) the bank will feel tempted to use their bonds to gain access to liquid funds to meet this demand from depositors.
• They can do this by borrowing from the central bank, using the government bonds they hold as collateral.

Recently an RBA assistant governor even gave a speech about how all of this works – give it a look!

Summing everything up – these unrealised losses on government bonds for banks that are the topic of twitter are not really a concern. The real concern would be if there was a genuine increase in bad debt/default.

However, in a world of uncertainty where there are certain triggers for default (rising unemployment, structural reform in China and their housing bubble) the concern about unrealised losses can make financial markets a bit more fragile to these shocks – something that financial regulators will be keeping an eye on.

A few years ago Gulnara and I made a video on unemployment. Although this was a particularly unpopular video, it was actually one of our favourites – and we keep thinking about ways we want to expand the story of our friend JM.

In this video we give examples of how JM, a lawn mower man, could end up unemployed. A key example we gave was the creation of autonomous lawnmowers. We were then shocked when we walked into a store to see the following:

So what does the existence of automated lawnmowers tell us about JM, and how does it help us think about labour markets?

All the ways to mow a lawn

As you can see from this Frankenstein image from Dall-E, there are many ways to mow a lawn.

• Use scissors.
• Use a hand mower.
• Use an electric push mower.
• Use a ride on mower.

As we move down this list we are using a technology that can cut an amount of grass more quickly, and with less human effort.

These tools, or capital equipment, allows a person to cut more grass using one hour of their time. As a result, an individual can complete a task more quickly – and them (or their employer) will invest in these tools if the cost of capital items is sufficiently low that the saved labour cost is attractive enough.

However, more complex tools also require greater knowledge. It’s pretty easy to use a pair of scissors – but a ride on lawnmower requires maintence, knowledge of how to drive, and knowing when and where to change mowing settings. This implies that individuals have skills that allow them to do tasks using specific capital items/productive processes.

But now a new technology has appeared.

The automated lawnmower and our lawnmower man

The lawnmower man from our video – JM – had invested in a good lawnmower (physical capital), relationships with customers (intangible capital), and the skills to utilise these (human capital).

The relatively cheap consumer lawnmower strands all of this capital – JM’s customers can now replace JM with a robot, and within a year they have saved money.

JM could cut how much he charges, or only serve the small part of the market that doesn’t trust robot lawnmowers. But either way, the return he receives on his capital is now much lower.

If JM was renting the machine, and if people simply reached out to him on a platform without relationships, then his income would have solely been a result of his skill at performing tasks. As the new technology can perform the task at a lower cost, his return on this skill will now be lower.

And if its low enough he’ll be forced to drop out of lawnmowing and find something else to do.

Next steps for JM

The art of lawnmowing may have lost some of its allure, but that does not mean that JM does not maintain any skills from his background with the rotary blade.

He dealt with customers, he managed his own accounts, he learned about repairing machinery, and he understood how to deal with difficult situations.

The routine task of walking around with a lawnmower may have been replaced, but the non-routine task of managing relationships with customers is something JM will need to do in many other jobs.

More importantly, JM has a number of general skills that can be shared across jobs – skills that he will still earn income for when finding his next job. Specific skills associated with lawnmowing have lost value – but how much of his earnings was really a function of being a good lawnmower, rather than being good at running a business and managing relationships?

This question about the skill content of tasks, and the ability for people to substitute when there is a shock to the return to skills, is a key one in labour economics.

In fact this is a key motivating factor for much of the wage scarring work that was mentioned last week. If people are facing shocks that lead to a sharp decline on the return to their skills, we would expect to see long and persistent earnings scars for these individuals.

JMs journey will depend on how well he can substitute to a new job in a world of technological change. In fact, it will depend on how well any of us can adjust – even economists.

Previously we’ve talked about the cost of job loss in terms of revealed consumption responses – i.e. how do individuals cut their spending following job loss. This is a useful concept, and there is going to be more to say on this in the next year.

But a more basic point people may ask with respect to the cost of job loss – how much lower is someone’s lifetime income if they end up getting shoved out of their job!

Let’s chat about that while investigating some initial work my colleagues and I have undertaken at e61. For those who know things already, or don’t want to read, here is the picture you are after:

What is wage scarring?

Imagine you are sitting there doing your thing, working for your generic employer, doing your generic job, thinking about the great things you’ll do on the weekend.

Suddenly, the government says that everyone who has a tax file number that ends in three has to leave their job and cannot be reemployed at the same employer.

That would sort of suck – but it would be a purely exogenous job loss. After that job loss you would go around applying for jobs, trying to understand what might suit your skills, figuring out if you should move country or study etc etc.

Eventually you would find a job, and over time you would progress in that job and earn an income.

Wage scarring reflects the difference in wage earnings between this job, and what you would have earned if you had not lost that initial job (the counterfactual).

We can work out this counterfactual you by considering what happened to individuals who didn’t have a tax file number that ended in 3, but otherwise look very similar. If you would have earned income in a similar way to them in the absense of job loss then the difference between their earnings and your earnings is as estimate of the wage scar you experienced.

Who cares?

It would be easy to say “this is an important measure of the harm from job loss”, but this is actually a question that requires a bit more.

By just framing this as a measure of labour market costs from job loss, it is easy to just view this as a realisation of labour market risk.

If this is the case, then these job losses are just a product of the world changing and a product of the risk we have in life – in that case any arguments about this relate back to how we feel about self vs social insurance for labour market risk.

However, lets think about some of the reasons why income may have declined for the person:

1. Time out of work see’s the workers skills become less valuable (loss in human capital).
2. Technological change has reduced the value of individuals skills (loss in human capital).
3. The firm was a high productivity/high pay firm (firm value/rent).
4. The match between the worker and the firm was better than a new match (match value).
5. The worker was very knowledgeable about using systems at that firm to their best use (match specific capital).

And what about the firm?

1. The value of the worker had declined, but not the tasks involved in the job (reduced match value).
2. The value of the tasks declined (reduced job value).
3. External factors (i.e. credit constraints) forced the firm to end a productive match.

Each of these combinations of reasons why the firm ended the job and reasons why the worker ended up with lower income has different interpretations. Some are just transfers from the worker to other people, while others show loss of surplus that imply this is a net loss in income over everyone.

If we want to fully think through the consequences of shocks that lead to job loss we have a lot of other things to think about. However, understanding the income loss associated with an exogenous job loss helps us start to tease out this problem – and also understand the cost for people who face job loss.

Why do you keep saying exogenous

Fair.

Job matches can end for a whole bunch of reasons. There can be self-selection (i.e. a job-to-job transition or “voluntary” unemployment where the worker quits) or there can be firm selection (being fired).

Self-selection isn’t really our focus, as it is likely people are leaving jobs to get a better job!

Firm-selection also has an issue – as the firm is likely to remove their “least productive” or “worst matched” workers. As a result, they select a type of worker that they were not going to give a pay rise – and so comparing them to other workers will overstate the income loss from this event.

We are trying to isolate the effect due to job loss – not due to selection. As a result, we want to make a case why the job losses we’re looking at do not involve much self-selection or firm selection. Now it is unlikely that we can find a perfect experiment here, but we just need to be clear about this!

What does prior evidence say?

International studies

I’ve already written a lot of words, so I’m going to share a table from Bertheau et al. (2023).

So the five year scars estimated tend to be between 6% and 36%.

What about New Zealand – Hyslop and Townsend (2017) found scars of 13-22%.

The Australian case

There have been a few recent studies in Australia, lets list them out:

1. Lancaster (2021): Using survey data there was no additional wage scar found after 5 years – involved looking at all unemployment.
2. Ballantyne and Coates (2022): Using survey data a scar of around 10% was found after 5 years – again looking at all unemployment.
3. Andrews et al. (2023): A recent e61 piece using administrative data found a 29% scar – focuses on those who are made redundant in ATO data.

Well, what do you all do?

Selection and triggering events

So looking at our piece – specifically the technical appendix – we are looking at a very specific event, influencing a very specific set of firms and workers.

• The Global Financial Crisis (GFC) year.
• Firms who have let go over 30% of their employees for a persistent period of time (mass layoff).
• Workers that leave a mass layoff firm during the GFC.
• Workers who are in their sixth year of work at the same firm (long tenure).

The comparison is then between long-tenure workers who left a mass layoff firm vs long-tenure workers who did not experience this.

What does this have to do with selection? Well a mass layoff helps to mitigate firm selection – as they have many of their workers at the same time.

However, who is to say that these aren’t still the least productive? To help with that we look at long-tenure workers, who have formed a relationship with the firm and where the worker and the firm really know each other.

Results

For most of the results I’d suggest reading the note – it is only 2 pages long

But lets list some highlights here.

What is our five-year scar? 16% of earnings – so the individuals who lost their job in a mass-layoff have earnings 16% lower than expected.

How much is this just about people having trouble getting and maintaining work, as opposed to employed individuals earning less?

Turns out a lot. Those who are re-employed after 1 year (who stay re-employed) have no scar after 5 years. Those re-employed within 5 years have 5% scarring.

My answer is “no idea” – I just wanted to use the same title as the Australian Budget post. You yell inflation at me and I say “monetary policy offset through higher interest rates” – and the monetary and fiscal authorities can argue about that. New Zealand based economists can describe that for us.

Instead I’m going to have a think about a couple of the policies: game subsidies and the higher top tax rate.

I also see that they are ramming from the Tax Principles Act. I’ll save thoughts on that for its own post later on.

Are they making my Steam games cheaper?

That would be nice – not particularly fair, and definitely not an efficient use of funds – but at least it would benefit more than 1-2 people.

Instead they are doing the film subsidy rort for game companies now.

Rant on – if you don’t like rants skip

So for context New Zealand currently wastes about $200m pa putting that little “supported by NZ” logo at the end of credits on a variety of movies and television shows.

No that isn’t fair, they also support Weta’s well known culture of sexual harrasment and general toxic work culture – you aren’t really from Wellington unless you personally know someone that has been sexually or verbally harassed by senior staff at Weta.

They also make sure Peter Jackson has enough money to influence local council elections and make sure that nothing gets built in Wellington – while the city slowly collapses into a slum. And lets not forget that it also gives politicians the chance to go to movie premiers and maybe get a photo with someone famous!

The fact that New Zealanders are not publicly disgusted at film subsidies, and the corrosive impact they have on Welllington particularly has long frustrated me – and now with lights falling out of the sky, water mains bursting on a daily basis, the central city smelling of sewage, half the buildings still unfit for use (red zoned) since 2016, house prices out of control due to the lack of historic building, and the recent tragedy with fire, you’d think making the capital livable again would be the priority.

Instead, lets also give money to game developers, so they can lobby for their old neighbourhood to never have new building or infrastructure.

And then former Green MPs have the temerity to pretend there is justice in this – I guess the catch phrase of “green growth” just means making sure the green is going to whatever favours them.

Man, honestly screw you (not Economissive – he’s good people calling this trash out).

I’ve never liked the rort that is film subsidies. And Jacinda’s decision to keep them scaling up was a huge disappointment. But the fact this exists alongside the utter degredation of the infrastructure in my old home of “Wellywood” – especially in the last five years – should make a lot of people as angry as I am.

Rant off

Now that is off my chest – and if anyone is insulted by it please contact me in person so I can double-down – lets pretend you like the game subsidy and you are not a corrupt lobbyist. Instead you believe that there is good potential for New Zealand to compete with the world in “weightlesss” technologies – and with a bit of support, we can build an industry together! [dynamic comparative advantage]

If that is you, then the barbs were definitely not at you – I like you, and I like the way you’re thinking about things.

First off I’d warn you to ignore any “economic impact assessments” that come out – anything with that name ain’t a real assessment, and will make up a bunch of benefits and not mention costs. Lets instead think about the nature of this seed funding:

1. A gaming company will take on staff that have certain skills, around coding and creative designs.
2. It will combine those staff and some underlying capital to produce games.
3. The gaming industry is risky – but if they have a good idea there is a chance of making a good return. This return is shared in earnings and wages between the capital owners and the workers.
4. If things go south that’s pretty tough – the capital owner loses that, and the employees need to find more work.

If the gaming company doesn’t happen, people are in other jobs and the “capital” is allocated to other uses. The fact something is made doesn’t mean there is “net value”.

If people privately decided not to do this, but then did with a subsidy, it is either because they were incentivised to take on a risk they wouldn’t otherwise – or they are being shifted to a lower value activity.

You might say STOP, I’VE GOT YOU – imagine we have an industry with valuable skills that are transferable between firms, but each potential firm faces a lot of risk, in that case we may have a coordination failure where no firms move into that industry and invest … but if they did, value added!

Ok nice – that was a much stronger argument than I expected.

But why is this gaming in particular? Why doesn’t the government generally offer a scheme where they “co-own” firms – providing beneficial arrangements on failure, but taking a cut when there is success! You know, like a limited liability bankruptcy regmine with a corporate tax system.

…. Wait. Are you telling me we could just change those settings, rather than picking our friends and giving them handouts!

Specifically picking gaming companies, and giving them support, instead of articulating and understanding a general argument about the insurance and support settings in a country is dumb. New Zealand was unique in how well this was understood by politicians – but modern politics in New Zealand appears to involve a series of people who think they are the smartest person in a room, but only experience it when they’re alone.

Trust taxes at 39%

Ahhh, glad to see this change! The gap between the top income tax rate and the trust rate is problematic – see here – so this makes those settings more sustainable and undermines a bunch of potential tax planning.

What’s the planning? Well previously the trust rate was 33%, but if you earn over $180,000 the top tax rate is 39%. So say you have your own business and you earn $400,000. You could pay yourself $180,000 and have the extra $220,000 distributed to a trust. That $220,000 gets taxed at 33% (instead of 39%) and then the following year you just send yourself that money without additional tax – noice.

Now IRD has fancy rules to try to avoid really bad examples of this, but it is hard to avoid subtle avoidance – even with the new trust information collection rules. So increasing that rate is a surefire way to sort things out!

However, one thing that gets a bit mixed is some thinking that “trust income is always avoidance”. Hold up mate, one of the issues with bumping the rate up to 39% is that some people will now be taxed at a rate higher than they would if it was classed as their income!

Imagine you have zero personal earnings, but your trust earns $14,000. You don’t really pay much attention, and you live meekly in your own cottage eating leeks that you grow for yourself. Nice.

If the $14,000 was taked as your income you’d pay a tax rate of 10.5% on it. If it is trust income, you pay 39%.

As a result, you’ll get some unattentive people who are genuinely using it as a shelter who pay more tax.

What is the solution here? Well you could treat trust income like company income – so the tax is a withholding tax, and when the income is paid to the person imputation credits are attached.

Problem solved!

Well, remember our mate before – they are pretty inattentive. Reporting everything as if it is company tax is quite involved – that might be a bit much to ask them to do. We could get around that by setting an income deminimus … except the issue only exists for those with lower income.

These are issues that are too hard for someone like me – so I’ll leave that to the experts and just say this was interesting to see!

If we are only going to spend a fixed amount on welfare payments, then there is a trade-off between the size of the payment and the incentive to work – where the incentive to work is captured by how much of their new found labour earnings they get to keep. It is even a common point that I make when I’m off lecturing on the topic.

But what if I told you that empirical evidence suggests unemployment benefit recipients who are working don’t appear very responsive to what are essentially huge (50 percentage point) increases in their tax rate?

Well that is indeed what I find here, as discussed by the AFR here.

What does this mean? A number of things so lets have a chat.

Financial disincentives to work

When it comes to the Iron Triangle the “incentive to work” is about these fellas:

Note: Quick heads up that the above numbers are a bit buggy and based on unfinished code – and based on “old” payment rates in this alpha version of a tool here. But the pic is useful for talking concepts.

How do I read these without going through this excessively long blog post? The effective marginal tax rate (EMTR) noted here tells us that, if someone worked for an extra hour they’d give back this percentage of their hourly earnings to government – through taxes and reductions in their benefit.

The effective average tax rate (EATR) tells us that, if someone decided to work at a given hours level, they would give that percentage of their gross earnings back to government.

That is where this line comes from:

Currently, a single Australian working part time on the minimum wage at a department store will pay a higher effective tax rate (including both tax and the loss of benefit income) than someone earning in the low six figures at a large investment bank.

This is due to benefit abatement – in order to “keep the scheme cheap” we essentially ramp up the tax rate faced by those who are on the benefit.

Now that feels sort of BS, but one thing we need to remember is that these beneficiaries are still net recipients of financial support (or in other words they receive more in benefits than they pay in taxes).

So if our view was ONLY that the benefit should provide a minimum standard for people to live on, and we should aim to make this scheme as cheap as possible, we may design a scheme that has very high abatement – and therefore very high EMTRs and EATRs.

The trade-off given this poverty allieviation target is then behavioural responses – do people adjust how much they work because of this sharp abatement. And this is the area we want to build a bit of understanding about.

EMTRs and the intensive margin

Often our focus is on very high EMTRs and how that may lead people to restrict how much they work to stay on the benefit, or maximise government support.

This is what we term the intensive margin response – given you are working some hours, what is your incentive to work another one?

Based on fairly standard techniques we can look at bunching in taxable income data to get an idea of how much individuals manipulate their earnings to avoid paying a higher marginal tax rate. We can see that nice and clearly with our buddies “high income self-employed Aussies” right here:

Look at that spike in response to an 8 percentage point increase in the tax rate! Now such a spike can be the result of a bunch of things, actively adjusting who receives the earnings, when you receive the earnings, rounding bias in reporting an income amount, and a decision to change behaviour (such as work less) in order to earn that amount.

Self-employed people have the most ability to do this, but you do see these types of spikes occur all across the income distribution – including for low income earners at their tax brackets.

So what about benefit recipients when faced with the abatement rate going up by 50 percentage points – that is a big tax change, so it should be pretty wild right?

Na.

On this basis it doesn’t look like there is much of an “intensive margin” response by those on benefits. Similar non-responses show up across a variety of periods and also at the second threshold – while responses are quite typical at tax thresholds in Australia.

So tapering is all good right – as long as we are morally comfortable with these high tax rates on this group (which will still be negative tax rates when we think about it in terms of “redistribution”) no behaviour no problems?

Other margins: Mutual obligations and participation

Participation

Not so fast. Lets start with the elephant in the room – what about the 75-80% of JobSeekers who aren’t working? There is the reason the micro note starts with a clear definition of which margin we are looking at – because the response of these individuals also matters, and isn’t included.

Pushing up the income free zone to around $300 would provide a lump sum to everyone who is on the benefit and earning above that amount of $75. It would also pull a few higher low income earners onto the payment. But most importantly, it would reduce the EATR for a minimum wage worker who want to work part time in this range by around 8%.

The potential to bring in an extra $75 a fortnight if you can find a job that pays anywhere between $300 and $1,000 a fortnight could have implications for people’s willingness to participate – and it could be that any labour supply action you see occurs there. This is an important reflection as noted here:

A ‘large’ extensive elasticity at low earnings can ‘turn around’ the impact of declining social weights implying a higher transfer to low earning workers than those out of work, in turn providing an argument for lower tax rates at low earnings and a role for earned income tax credits.

Blundell, Bozio, Laroque (2011)

Information and barriers

For those working there is a separate approach that captures that there may be informational frictions, or other frictions that lead to bunching “shifting the counterfactual”. Chetty etal 2013 discusses this in great detail – strong recommend. There are two ways this can chip in:

• The change in bunching may occur at “salient balances”,
• The broad density associated with the distribution may move – rather than spike.

For a long time I believed the result had to be one of those two – with the difficult thing defining the “counterfactual group” that could be used to construct it. Lets just say that this “extensive-intensive” margin just didn’t end up changing anything – but is something I’m passionate to keep testing results against it.

Regulatory barriers – mutual obligations

However, there is an important reason why the limited intensive margin response may show up in any extensive margin work – the heavy importance of mutual obligations.

The micro note focuses pretty heavily on mutual obligations as a concern.

The Australian benefit system has surprised me with the sheer amount of conditions individuals and supposed to meet to receive payment. Specifically, significant job search requirements – and the threat of payment removal that is credible – make undertaking part time work to avoid these obligations attractive in of themselves. Part-time work requirements to remove mutual obligations vary, and short amounts of work can be mixed with other requirements in order to maintain benefit eligibility.

If this is the case, then individuals will simply be taking up the job offers available – the lack of bunching behaviour makes sense. However, if it is mutual obligations driving this “non-response” on the intensive margin, then it should show up on the extensive margin as well – given the risk of losing your benefit if you don’t take up marginal work, those who are particularly responsive to financial incentives may have already jumped into work to insure themselves against this loss.

The fact that benefit thresholds don’t show bunching while the tax thresholds do – even at the low end – does indicate that is a systematic difference in the choices made. Although it is common for people to make comments about rationality and attentiveness I call this out in the note – it is more plausible this is about the choice set available to people and the genuine ETRs they face when making these choices.

Mutual obligations are a big part of this. And their existence will change how responsive individuals will be to a change in headline financial incentive measures.

So the potential for people to “move into” work from out of work remains a key margin to understand – but the note is indicating that the stark non-bunching in combination with the mutual obligations system suggests that pure financial incentives might not be enough, given the importance of regulation and mutual obligations for low income earners.

So this gives two key research directions for these labour supply responses – extensive margin responses, and the efficacy of mutual obligations. Let me get back to you on that