I think I basically agree with the bottom line here, but I think one point seems a bit off to me.
Another major issue is that a full or near-full land value tax would likely establish a troubling precedent by signaling that the government has the appetite to effectively confiscate an additional category of assets that people have already acquired long ago through their labor and purchases.
[...]
Beyond setting a harmful precedent that could influence people’s future behavior, a land value tax also creates a major disruption to people’s current financial plans, particularly for those who have spent decades developing strategies to preserve their wealth.
I agree this applies if the land value tax operates via the government effectively owning all land from now on (e.g. via charging annual rent proportional to unimproved value or in principle the government could do a one time tax equivalent to land value and then tax increases in value). Another way to put this is that this version of an LVT is effectively the same as doing a 100% one time capital levy on a particular type of asset and has the corresponding issues that capital levies have as well as increasing variance via just targeting one type of asset.
One proposal for (partially) getting around this issue is to say that the government owns increases in land value from the time the LVT is introduced, but does not tax existing land value. Precisely, suppose your land is currently worth $1 million. If your land would stay the same value, you pay nothing. If the land appreciates to being worth $1.1 million, you have to pay the government $0.1 million that year. (Or alternatively, the government will charge rent for that $0.1 million share it “owns” ongoingly.) With this approach, it’s important for the government to give money back if land depreciates to avoid disincentivizing variance (similar to the sort of proposal discussed here).
(I also think that a LVT should be more like 80% than 100%, but I’ve used 100% in the numbers above for simplicity.)
Of course, this means the LVT brings in even less revenue in the short term if the tax is implemented in a rental style. If land appreciates at 4% per year on average, then in 30 years[1], the government will own about 70% of land by value via this type of LVT, so it shouldn’t take that long before this is equivalent to the government owning unimproved land value. (I think you’d probably want to gradually increase the tax after testing it out for a bit to reduce disruption.)
If we imagine that the goverment directly charges for increases in land value (or sells off the rental rights to get the revenue immediately), then this means government revenue increases by total_land_value * appreciation * LVT_rate which is perhaps (for the US) $23 trillion * 4% * 80% = $0.7 trillion. Current tax revenues are about $4.5 trillion in the US, so this would be about 15% of revenue which isn’t that bad: you certainly still need other taxes, but you can displace a substantial amount.
Ok, but isn’t this still a bad precedent and disruptive, just to a lesser extent? After all, people were expecting that they would get the appreciation in land value and the government is taking that. I think it sets no more of a bad precendent and is no more disruptive than increasing income or capital gains tax (analogously, people were expecting to see returns on their investment in education or their capital investments). Quantitatively, there is something different about taxing a certain type of appreciation at a high rate (e.g. 80%) rather than increasing taxes more marginally, but it doesn’t seem that bad.
I think this type of land value tax where you tax just appreciation seems basically strictly better than income or capital gains taxation even for a well implemented capital gains tax. (That is, putting aside evaluation issues and assuming you can get kinda reasonable exemptions for value discovery.) It still seems bad to suddenly increase a tax to a very high amount due to disruptiveness, but certainly no worse than if you got $0.7 trillion via suddenly increasing some other tax.
I expect transformative AI prior to 30 years and generally think this makes discussion of mundane goverance much more confusing, but in this comment, I’m ignoring this.
(Inexpert speculation, please forgive my errors.)
I think I basically agree with the bottom line here, but I think one point seems a bit off to me.
I agree this applies if the land value tax operates via the government effectively owning all land from now on (e.g. via charging annual rent proportional to unimproved value or in principle the government could do a one time tax equivalent to land value and then tax increases in value). Another way to put this is that this version of an LVT is effectively the same as doing a 100% one time capital levy on a particular type of asset and has the corresponding issues that capital levies have as well as increasing variance via just targeting one type of asset.
One proposal for (partially) getting around this issue is to say that the government owns increases in land value from the time the LVT is introduced, but does not tax existing land value. Precisely, suppose your land is currently worth $1 million. If your land would stay the same value, you pay nothing. If the land appreciates to being worth $1.1 million, you have to pay the government $0.1 million that year. (Or alternatively, the government will charge rent for that $0.1 million share it “owns” ongoingly.) With this approach, it’s important for the government to give money back if land depreciates to avoid disincentivizing variance (similar to the sort of proposal discussed here).
(I also think that a LVT should be more like 80% than 100%, but I’ve used 100% in the numbers above for simplicity.)
Of course, this means the LVT brings in even less revenue in the short term if the tax is implemented in a rental style. If land appreciates at 4% per year on average, then in 30 years[1], the government will own about 70% of land by value via this type of LVT, so it shouldn’t take that long before this is equivalent to the government owning unimproved land value. (I think you’d probably want to gradually increase the tax after testing it out for a bit to reduce disruption.)
If we imagine that the goverment directly charges for increases in land value (or sells off the rental rights to get the revenue immediately), then this means government revenue increases by total_land_value * appreciation * LVT_rate which is perhaps (for the US) $23 trillion * 4% * 80% = $0.7 trillion. Current tax revenues are about $4.5 trillion in the US, so this would be about 15% of revenue which isn’t that bad: you certainly still need other taxes, but you can displace a substantial amount.
Ok, but isn’t this still a bad precedent and disruptive, just to a lesser extent? After all, people were expecting that they would get the appreciation in land value and the government is taking that. I think it sets no more of a bad precendent and is no more disruptive than increasing income or capital gains tax (analogously, people were expecting to see returns on their investment in education or their capital investments). Quantitatively, there is something different about taxing a certain type of appreciation at a high rate (e.g. 80%) rather than increasing taxes more marginally, but it doesn’t seem that bad.
I think this type of land value tax where you tax just appreciation seems basically strictly better than income or capital gains taxation even for a well implemented capital gains tax. (That is, putting aside evaluation issues and assuming you can get kinda reasonable exemptions for value discovery.) It still seems bad to suddenly increase a tax to a very high amount due to disruptiveness, but certainly no worse than if you got $0.7 trillion via suddenly increasing some other tax.
I expect transformative AI prior to 30 years and generally think this makes discussion of mundane goverance much more confusing, but in this comment, I’m ignoring this.