What happens if we increase benefit payments? Part One
For those who subscribe to the e61 alerts you will have seen a couple of things come out this morning. Normally I’d save discussing these for a little bit - but I want to give you my thoughts before I’ve seen public comments, so I’m giving you my unpolluted feelings about the insights our teams work on :)
If you don’t subscribe, then do. And now you are back, what am I talking about?
Throughout the last couple of years we’ve been releasing research and insights on the benefit system in Australia. You can find these pieces here.
The focus of this week is two-fold - to ask about the labour supply responses to changes in unemployment benefit changes, and to return to our 2023 decision about the “insurance” benefits of a higher payment. [When this post goes out only the labour supply work will be up - but I will edit with links and delete this comment once it is all up :)]
If you are unsure about the Australian system, may I recommend this framework post, this graph post, and this post on absolute vs relative concepts.
Wait wait, please give me a framework
Fair.
So quickly we could just paste this into ChatGPT to ask it what it says. But since I’m too lazy to do even though that, I’ll outline the Baily-Chetty Framework here.
In a situation where people are looking for jobs and can receive government support in the case of job loss, we can evaluate the costs and benefits of the benefit system by comparing the consumption smoothing benefits of the system against the moral hazard cost.
Consumption smoothing benefits just meant that people are able to make it through a period of job loss without dropping consumption. Moral hazard costs reflect the fiscal cost associated with people deciding to spend more time out of work due to the payment.
If you want to see this idea put into practice I’ll link to a narrative going up later this afternoon by Rose Khattar and Gianni La Cava to tie together a lot of the work.
But because I can’t help myself I’m going to waffle about some stuff below.
On labour supply responses/time out of work
A while back we showed no clear intensive margin labour supply response to benefit changes - so if you changed abatement a little bit it is unlikely to change hours worked too much. This is consistent with international evidence on local responses.
Now we are talking about the extensive margin - if the benefit goes up do people spend longer out of work, and are going to leave work a bit more quickly.
We find that a 10% benefit hike would likely lead to people spending a week more out of work. When comparing approaches we were also of the view that - even though higher benefits did lead to more people leaving jobs during COVID - for a 10% hike nothing would happen. These are both consistent with international evidence (linking David Card as believe we missed this in our lit review).
So these are the results, and to an audience outside of Australia these would be uncontroversial to the point of boring.
However, the bunching result was quite controversial at the time - so just in case I wake up to the new stuff sounding controversial it is important to note that these responses are not large or particularly surprising.
They are just a reflection that financial incentives can matter, and this matters for fiscal nerds who want to model things and taxpayers who want to understand what they are paying for.
Why not explain what you all did?
Again, fair - I will do more detail in future posts, but lets outline here.
So on the labour supply work we took two approaches. Both have flaws, but given that they are close to prior Australian and international work they do help us update our Bayesian Probabilities - or to talk straight, they add to overwhelming evidence of labour market responses of about this modest size.
The first one is by Erin Clarke, Ali Vergili, and myself. During COVID the JobSeeker Payment was suddenly increased by $550 per fortnight - that’s pretty big. So why not just look at when that happened to see if people spent more time out of work/where less likely to find a job.
Well, the payment was introduced due to lockdowns, concerns about catching COVID, and a general feeling this could be the end of the world. So of course people were less likely to find a job!!
As a result, to investigate this we need a control group - people who couldn’t receive the Supplement but with the same work rights, access to a lot of the same support systems, in the same labour markets, facing the same COVID stress, and with the same mildly uncomfortable sense of humor.
Amazingly Australia has this group living in our midst - New Zealanders. Comparing these two groups we see the job-finding rate drop 19% on the announcement of the subsidy, that’s our result. Well, that divided by 9 (to scale down the payment increase).
The second approach is by Jiaqi Luo and myself. For this we fit a standard search and matching model to the Australian economy before COVID, and then asked how things would change if the benefit rate changed. This does not use the COVID evidence, but is a useful thought experiment to sense check the results.
This suggests similar results in terms of quarterly job-finding rates - with a 10% increase lifting time out of work by a week, and a 25% increase lifting time out of work by 3 weeks.
My key concerns about these two approaches is that:
COVID is weird - the control group helps by differencing out weirdness that hits both groups, but COVID is still a weird time.
A SAM model like this - in part - reflects our assumptions about how job matches are formed (and thereby the evolution of the distribution of productivity). It does not try to match a known causal estimate of benefit responses. As a result, although the model is consistent with labour market flows and business cycle variability, the labour supply response to benefits to some degree reflects these assumptions rather than true causal evidence of an effect.
But these facts are as likely to have lead us to either overstate or understate the results. Ultimately, they reflect evidence that Australia acts like every other country on earth - with maybe slightly lower labour supply responses than most.
How do I think about this?
In the framework above we cared about this because of the “moral hazard” of the benefit payment - we aren’t arguing that people spending more time out of work is good or bad, it just has a fiscal cost that needs to be funded.
But before you start saying its good because of “wage scarring” or “match value” - lets just say the evidence doesn’t really back you up.
Instead the good part might just relate to reducing a bit of life stress and a bit of help managing finances.
The benefits of a higher payment
There are two forms of “insurance” benefits from higher benefits - and a separte fairness based “redistribution” benefit.
The redistributive effect is about effects on poverty and we won’t touch on that here - but if you want to reduce measured poverty, higher benefit payments are a direct and effective means to do so. [The divide here is often about the impact on persistent poverty - that is fine but one thing at a time please.]
So what is this insurance stuff?
Consumption declines due to lack of liquidity: If individuals were insured, consumption shouldn’t change on job loss. Clarke, Adams, La Cava and Nolan covered this a while back - with new work with the help of Pelin Akyol underway. Key stat, someone who receives the benefit reduces their spending by 15% in the year of job loss - and spending remains lower for at least five years as they try to rebuild.
Financial stress and mental health: If individuals were fully insured, a negative life shock like job loss would not lead to financial stress or declines in subjective wellbeing. However, it does. Upcoming work by Akyol and La Cava show that the COVID supplement lead to lower financial stress - so a higher benefit payment fill that gap.
So how do you balance these?
Good question. Sign up to email alerts from e61 and you’ll get a response by March ;)
To fill in the time let me say something. I’ve written a few related things on this blog on benefit payment. Looking across this led me to say:
But given the changes already taken in Budget 2023 the argument is not as strong as last year. Furthermore, we also shouldn’t just use that argument every Budget ad nauseam. The right level of support should be based on a clear understanding of the trade-offs, and a public discussion of what we deem to be the right trade-off to accept.
If you believe in a benefit increase, or that the benefit system should be more restrictive, you can make that argument - none of these facts invalidate that. Welfare policies are a game of tradeoffs.
Our values about how much government should be responsible for insuring risk, and about the minimum living standard that people deserve, are not empirical questions. But the trade-offs associated with making those choices are.
I personally think that both taxes and benefit payments should be higher. Always have. Nothing in this work changes that view in any direction.
But now I know that when I plug this labour supply response into the awesome PBO budget tool the net fiscal cost for a 25% increase in the JSP payment rate, the costing goes from $4.1bn to $5.6bn.
This costing excludes any increase in benefit takeup (which I think is reasonable in Australia given high takeup rates) and any increases in other benefits (i.e. Parenting Payments, which would have to rise).
If I want to make the case to others about this I should do it on the basis of these facts - because sustainable policy changes rely on trust. For example, if I told you nothing would happen if we introduced a different policy - and then people did find people worked less - you aren’t going to trust me for future change, and may even try to reverse payments due to the feeling of being misled.
On the flip slide, say you accept the poverty line is something biting and accept that current payment puts people below that line.
If you then hear that an unemployed person will spend a couple more weeks looking for work, and that is enough for you to not want to support them - then you do you, but that seems pretty wild from where I’m sitting. Especially if we accept the related fact that “welfare dependence” is actually falling.
Ultimately economics is about informing, not persuading - this is pretty much the only belief I have that I’m dogmatic about ;)