Tuesday 30 June 2020

Wealth and the life cycle

It's frustrating how hard it is to know some perfectly knowable things.

Over the weekend, the Greens announced that they support a wealth tax on individual net wealth in excess of $1m. 

In a country that had sensible priorities for its statistics department, I would have an easy time answering a simple question. Fixing the underlying problem here I understand to be a big job. But it would be a solution to problems that just keep coming up - not just this one. And it isn't like there aren't other things that the government has tasked Stats with doing that seem rather less important than fixing this problem. 

I would like to know the distribution of wealth by age bands. The 2018 wealth module of the Household Economic Survey has information on the distribution of households by net worth bands, and household net worth by household characteristics, and by a few other ways of slicing the data that would have occurred to a Stats analyst might be useful when they did the thing up in 2018, but nothing that can let me look at it by age. I don't blame anyone for not having this slice in - there are huge numbers of ways of slicing the data and it would be impossible to have all of them up in every release. 

In a system that had functional back-end systems, you'd be able to put in a query and it would just spit out the relevant cross-tabs - generating them on request from the underlying data. IPUMS has had this kind of functionality for ages - you can even run regressions on their data from inside your web browser.

Here, I can't even tell the age distribution of the 216,000 households with net worth in excess of $1.5m - the topcode in their income buckets. 

Why does this matter?

Imagine two states of the world. 

In state of the world A, 5% of individuals have net assets in excess of $1m. They are born with those assets endowed by bequest, and they die leaving those assets to their kids. Nobody else ever accumulates net assets in excess of $1m. There's a landed gentry, and everyone else. 

In state of the world B, 5% of individuals have net assets in excess of $1m. As people move through the life cycle, they move from being net debtors to net asset positions, and retire at age 65. Their wealth holdings peak around then, with draw-downs exceeding the return on their investments. 5% of people are aged 65-69, and every one of them has assets in excess of $1m when they retire, before they start drawing it down. They're all back below $1m at age 70.

We are absolutely not in either of those states of the world. But static numbers on wealth accumulation say nothing about the life cycle. The presumed equity considerations around a wealth tax would have to be different if small numbers of people are ever subject to it as compared to if large numbers of people became subject to it as they reached retirement. In state of the world B, a wealth tax transfers money from individuals' future richer selves to their current poorer selves in an overlapping generations sense, but where people tend to save money in anticipation of retirement it's an odd way of running things. The rest of the system (NZ Super) is designed around taxing people when they're working to help them fund their retirement. Not the other way round. 

Some of what Stats has is suggestive: couple-only households (retirees will be more likely to be couple-only) have median wealth of almost double that of couples with dependent kids. But that's highly imperfect - professional couples without kids will almost certainly find it easier to accumulate wealth if they don't have kids. And, even if you just lumped together everyone over 65, that would also catch lots of people who'd substantially drawn down their savings. The median across that bigger group wouldn't be the right answer either. 

While it would be nice if Stats would pull the number for me - I've requested it, and I understand that there's a two-week lag these days on custom data requests - it's still not a solution to the general problem. It's just frustrating that back-end systems have not been updated so that cross-tabs like this can be generated dynamically by users on the web interface, rather than by specialised request. Fixing this kind of thing once would save the custom data team a lot of work. It's impossible to anticipate all the cross-tabs that might wind up proving useful down the track; adding more tabs to downloaded excel sheets isn't the answer and can't be the answer. The systems need to be fixed so that things that are dead simple to know in principle can actually be known. 

In an election campaign, two-week lags for basic fact checking about the contours of policy can be an awfully long time. I'm not blaming the team at all for it - Stats is always crazy helpful. But the system really needs to be fixed so that things that should be knowable at a click on a browser don't require custom data pulls. 

No comments:

Post a Comment