• @JamesFire
    link
    322 hours ago

    I mean, if they removed the means-testing from benefits, you could significantly reduce the staff required to administer it.

    https://www.theguardian.com/social-care-network/2013/jan/14/means-testing-benefits-not-efficient-fair

    Means-testing almost always costs more to implement than just giving out the benefit. Or is otherwise such an insignificant part of the total that it just doesn’t matter.

    For a “simple” example, let’s say the US implements a UBI. A flat 2k a month to every resident, or 24k a year. We will not adjust this by COL anywhere, to keep it simple.

    The current (Estimated) US population is 340,110,998. This brings the total cost of our program to ~$680 billion dollars a month, or ~$8.1 Trillion a year. (We’re going to ignore how the US actually gets this money, or the impacts it would have, and the fact that the federal budget for 2024 was only $6.7 Trillion. But also fun fact, the combined federal and state budgets is $8.4 Trillion, which would be just enough to implement this plan. And basically nothing else, but oh well.)

    So we have our number. Let’s apply means-testing to it. What would that look like?

    If we assume they hire people at the US Median Wage of ~$37k a year, plus benefits, which adds 15-30% of the salary (We’ll use 30%, since government benefits are usually very good), that’s a total of ~$49k per year. So each worker has to deny at least 3 people a year to pay for themselves. But they can’t just deny people, they have to look at every application, because the whole point is to make sure nobody gets benefits they “don’t need”

    Now what criteria would be used for this means-testing (And thus how many people would actually be denied, and how long would it actually take to handle the application), and how often will it have to be handled (Having to handle an application every month is a lot more work, but doing it every month would allow the program to be flexible in responding to changes in people’s lives)?

    Lets say we want the flexibility of monthly to catch people quickly. A lot can change in a year, and we wouldn’t want anyone to go for too long without income they’re entitled to. So our workers have to deny at least 3 people a month, not a year. So let’s look at what handling applications will actually take. Surely it won’t take a week, or more, for each one?

    Simple criteria like making too much money is pretty easy to implement, right? Then you can just ask the IRS for income records. Well, no. For basically everyone, their income is submitted yearly as part of the various documentation you need to file your taxes. So what do we do? We could change it so the IRS gets this data every month. But compiling it into usable form is a lot of work, so you’d definitely need to hire more people at the IRS. How many? Would probably need to increase their workforce by a pretty large multiple.

    As-is, they do everyone’s taxes in ~6 months. (You can file anytime after January 1st, until April 15th, and turnaround time is given as 6-8 weeks at most.) Compiling income data is certainly less work than doing a full tax filing, but it’s still a lot more work than they’re currently doing. Let’s say it’s 50% of the work, times 12 for doing it every month. So they need 6x their current employee count, in addition to what they already have (This also isn’t counting how much more work accounting departments would need to do to submit this data, but they’re not part of the government, so they don’t matter). Their current employee count is 93,654. They would need to increase that by 561,924. If we keep things simple and just assume each one of those new employees gets that median wage+benefits we calculated earlier, the IRS would now cost an additional $27.5 billion a year. In order for that to result in total decreased costs for our UBI program, we would now need to deny at least 1.15 million people from the UBI program every month. And we still haven’t even gotten to how many people it would require to deny that many people, and then how many more we need to deny to cover the cost of those deniers, and so on. (This is also assuming this change to monthly stuff for the IRS doesn’t reduce the workload at all related to filing taxes yearly. It might, but it’s way easier to just ignore that)

    So let’s say we do that to implement this means-testing. Let’s say the criteria is just a simple “you make above $X”, and that using the expanded IRS, they have reasonable access to records proving income. They’ll still need to manually view every application. So how long will it take to actually handle each application?

    Well, it depends on how the IRS provides the data, and what the worker actually needs to do to make the approval/denial. Sure you could develop dedicated software to make this easy and take a likely 30s-1m per application, or even entirely automatic, but that’s additional cost and IT support. Let’s say the IRS provides the data as grouped income totals. So employment income is separate from capital gains, which is separate from contracts work, and so on. The worker needs to sum these up, and enter this total as proof, and approve or deny the application. Let’s say this will take 3m per application on average. A single worker, in an 8 hour day, assuming perfect performance, will be able to handle 160 applications a day. Or about 4800 a month. With 340,110,998 applications a month, you would need about 70,857 workers. Now, this isn’t nearly as many as the IRS needed, but it’s still a lot of people that would cost ~$3.5 billion a year. Surely peanuts compared to the total program cost of $8.1 trillion. To cover this, the program would need to deny an additional 145k people.

    But wait, people aren’t perfectly productive for 8 hours a day every day. People usually only work 5 days a week, and office workers are actually only productive for a little under 3 hours. https://www.inc.com/melanie-curtin/in-an-8-hour-day-the-average-worker-is-productive-for-this-many-hours.html

    So with this in mind, a work would actually only be able to put out 1153 applications a month, requiring a much larger 294,980 workers, costing ~$14.5 billion in wages+benefits alone. To cover this, the program would need to deny an additional 603k applications.

    But lets keep going. What income level would we need to set to deny enough people to make this an actual cost savings? 1.7 million people is only ~0.5% of the US population, so it should be fairly high, right?

    97.82% of working people earned less than $250k in 2022. There are, by the US Census, 271,500,000 people in the US over the age of 15 who make an income.

    The remaining 2.18% of “working” people, totaling 5,238,000, is much higher than our 1.7 million people. So we can set the income limit to at least $250k (And maybe even lower, but this is good enough for now), and save ~$125 billion. ~1.5% of the total program cost, at the cost of $42 billion in just wages+benefits for additional IRS workers, and workers for these applications. This does not include support workers (Like HR, accounting, and managers) for them, or office space. It also does not include the extra work that the private sector would have to do to provide this data monthly to the IRS, nor does it take into account how these monthly applications would actually be done. Would each person have to make it? Would it just come up automatically every month?

    This also doesn’t take into account that this much money would have other knock-on effects in the economy, that the jobs we’re making are frankly tedious as fuck, nor the fact that this income counts as income. Even on payouts from the government, the income is taxable because it is administratively easier.

    The current US tax brackets have $231,251-$578,125 taxed at 35%. (a significant amount of people “making” this much money make it as part of capital gains, not employment income, so in that event the actual tax rate will be lower. But we wanna keep it simple.)

    That’s ~$44 billion in lost tax revenue (And only income tax. The money the people actually get could still come back as various forms of taxes, further hitting revenue). So now we’ve saved ~$125 billion, at the total cost of $87 billion (Which is still just a low-end guess that only includes direct government costs), so we’ve only actually saved $38 billion. A measely ~0.5% of the program’s budget.

    But what if we lowered the threshold? How much could we lower it?

    Let’s say we lower it to $100k a year. This does increase the people we can cut from 2.18% of people with income, to 15.05%, or from 5,238,000 to 36,013,000 people. Who probably don’t need it (Which is the point, right?). This would increase our initial savings to ~$814 billion, a much better number, at 10.6% of the initial budget.

    But what’s the tax decrease look like for this? (There’s an “even distribution” assumption here, to keep it simple)

    Well, 19,674,000 people are taxed at 24%.

    64% of 7,765,000 are also taxed at 24%. The remaining 36% are taxed at 32%.

    62% of 3,336,000 are taxed at 32%, while the other 38% are taxed at 35%.

    5,238,000 are still taxed at 35%

    So in total, we have ~$235 billion in lost revenue.

    24,643,600 people, taxed at 24%, that would have paid ~$142 billion.

    4,863,720 people, taxed at 32%, that would have paid ~$38 billion

    6,505,680 people, taxed at 35%, that would have paid ~$55 billion.

    Including our low-end guess of direct costs of $42 billion, and the lost tax revenue, we’ve actually only saved ~$537 billion, down to ~6.6% of the initial budget.

    And we’re still not including any knock-on effects of this on the rest of the economy! Or an overly-complex set of criteria that would make reviewing cases take much longer, as is usually implemented when means-testing is done. For instance, the means-testing implemented on AISH, an Alberta Government Program for disabled people, has a whole slew of requirements beyond “Be disabled”