What could B.C.’s land wealth make possible?

Why this project?

There is no shortage of good arguments for society to collect its land rents.

If you have made it this far, you probably know most of them already. Land cannot be produced or moved somewhere else in response to a levy. Using its rents to fund government does not penalize someone for building a better house, hiring another worker, or making a productive investment. And the value of a good location is largely created by the community around it.

The harder question is: if the case is so good, why do so few places actually do it?

Politics is obviously part of the answer. But there is another problem that gets less attention: the details matter enormously.

What rate should you use? What taxes should it replace? How quickly should it be introduced? How much does the price of land fall? What happens to someone who bought just before the reform and is carrying a large mortgage? What about an older homeowner with valuable land and very little income? How much new development should we really expect? What happens to renters? And if the policy raises a lot of money, what should we do with it?

It is possible to believe very strongly in collecting land rent and still give quite different answers to all of those questions.

These are the origins for this project. We are trying to model those questions in enough detail that we can move beyond the general case for land value taxation and start comparing actual policy designs, and advocating for them.

A quick note on terminology. “Tax” is not always the most useful or accurate description of what is happening here. The underlying idea is that some portion of the value generated by land and location is being returned to the community that helps create it. For that reason, we increasingly use Land Value Return (LVR) as the broader term for this family of policies. A land value tax is one way of accomplishing that return, but the design choices around rates, offsets, dividends and other uses of the revenue can produce quite different policies. We will use LVR in that broader sense below.

Why British Columbia?

B.C. happens to be an unusually good place to get into the weeds.

For one thing, the problem is not abstract here. Land is extraordinarily valuable, housing costs are a major political issue, and a very large share of the value of residential property sits in the land underneath the building.

But B.C. also has something many jurisdictions do not: much of the machinery needed to seriously study a land value return already exists.

BC Assessment conducts regular, province-wide assessments and separately estimates the value of land and improvements. That distinction is fundamental. Before a government can collect land value separately from buildings, it has to be able to say what each is worth.

B.C. also has unusually rich complementary data on households, incomes, wealth, mortgages, taxes and property. The pieces do not all sit neatly in the same dataset, which is one of the problems discussed below. But enough of them exist that we can begin asking much more specific questions than “Would LVR be good?”

We can ask what different versions of it would actually do.

That is what the five options below are for.

They are not five recommendations, and they are certainly not five forecasts of what B.C. is about to adopt. Think of them as five different things we can ask the same tax base to do.

What if the priority were cutting taxes on income? What if it were reducing taxes on buildings? What if some of the revenue went directly back to households? What if we deliberately designed around the households most exposed to the transition?

Each choice produces a significantly different policy, and all of them can reasonably sit under the heading of Land Value Return.

That is really the point of this exercise. There is no single LVR policy. There is an unusually efficient and, we would argue, just source of public revenue, and then there are a great many choices about how to use it.

The results below are preliminary. They are being produced with a calibrated version of the model before the final work with the full research dataset in all its detail. So the exact numbers should not be read as final estimates for British Columbia.

What they already allow us to do is more interesting than that: start exploring the design space.

How far can we shift the tax system? What can we accomplish with the revenue? Where do the trade-offs appear? Who becomes vulnerable as we push a reform further? And can we design around those problems rather than discovering them after the fact?

That, ultimately, is the ambition of the project: to move the conversation one step beyond “this is a good idea” and toward “what would a good version of this policy actually look like?”

Getting into the weeds

The frustrating thing about trying to model a land value return is that most of the information you want exists. It just does not exist in the same place.

BC Assessment can tell us an enormous amount about property and land values. Statistics Canada can tell us an enormous amount about household incomes, wealth, mortgages and taxes. Administrative tax data can tell us how incomes and taxes change over time.

What we cannot do is open one spreadsheet and find every B.C. household sitting beside the value of the land they own, their mortgage, their income, their other assets, the taxes they pay and their location.

Public datasets are especially limiting here. They simply do not contain the resolution required to connect all of those things at the level needed for serious distributional work.

And even if they did, that would only tell us who writes the cheque on day one.

It would not tell us what happens next.

This is where modelling becomes useful.

Suppose we introduce an LVR and use the revenue to cut income tax.

Land prices should respond. Household balance sheets change. A mortgaged owner may suddenly have a higher loan-to-value ratio. Holding idle land becomes more expensive. Development decisions may change. Lower income taxes change the return to working. New construction changes the supply of housing. Those changes affect rents, wages, consumption and, eventually, the amount of revenue the government collects.

Those things do not happen one at a time.

So we have been building a model that lets them happen together.

For the results shown here, the household side of the model summarizes the population into 30 representative groups: ten income deciles, and within each of them renters, mortgaged homeowners and homeowners without a mortgage. Behind that sits a much richer research program using household, tax and property data at a much finer level. As that work progresses, more of the grouped relationships in the model can be replaced and tested against the full microdata.

For now, the grouped results make it possible to present what is happening without asking readers to inspect thousands of household-level observations.

And they already let the model do something much more useful than calculate an LVR bill.

Change the policy and the rest of the model has to respond.

The fun part is deciding what to solve for.

A lot of LVR modelling starts with a tax rate.

What happens if we impose a 1% LVR?

That is a perfectly sensible question, and the model can answer it. But increasingly, the more interesting way to use this work has been to turn the question around.

What are we trying to accomplish, and what policy gets us there?

If we want to replace a given amount of income tax, what LVR is required?

If we want to provide a household dividend, how large can it be?

If a reform creates a group of clear losers, what would it cost to compensate them?

If the objective were simply to maximize the economic gains from shifting the tax base, how far would the model want to go?

And if that answer creates unacceptable transition risks, what changes when we constrain the problem?

That is where this starts becoming a policy-design exercise rather than a tax calculator.

There will never be a single mathematically “correct” LVR, because governments are trying to achieve several things at once. They care about revenue, growth, housing, distribution, administrative simplicity and political durability. Sometimes those objectives pull in the same direction. Sometimes they do not.

The useful thing a model can do is make those trade-offs visible. 

And it gives us quite a lot to look at.

The obvious output is revenue. But that may actually be one of the less interesting ones.

We can follow what happens to residential and commercial land values. We can look at rents, housing supply, development density and idle land. We can see changes in output and consumption. We can watch mortgage leverage and financial stress during the transition. We can compare renters, mortgaged owners and outright owners across the income distribution.

More importantly, when a household gains or loses, we can start asking why.

Was it the levy itself? A change in wages? Lower housing costs? A transfer from government? The loss of land wealth? A change in collateral? A behavioural response?

Those channels matter because two households with the same income can experience the same reform very differently.

That is also why some of the results can look odd at first glance.

Someone can pay a substantial land levy and still be better off because the reform removes a larger tax somewhere else. Someone else can have more disposable income but lose wealth because the market value of their land falls. Renters never receive an LVR bill, but that does not mean they are untouched by the reform. And phasing a levy in slowly can change household cash flow without necessarily preventing an anticipated future policy from showing up in today's land price.

This is precisely the sort of detail that gets lost when the LVC debate stays at the level of first principles.

A note on the numbers

The figures on this page are not the final empirical results of the research program.

For now, we are running these policy experiments through a calibrated version of the model. It uses real B.C. assessment information and empirical household data to discipline the model, but the results presented here still summarize households into representative groups rather than exploiting the full confidential microdata in all its detail.

That makes these results useful for understanding the model, testing policy designs and finding the questions we should be asking. It does not make every percentage point on the page a forecast.

The next stage is to bring the richer data to bear on the same questions.

That distinction is important, but so is the reason we are publishing some of this work now.

One of the advantages of building the model this way is that we do not have to wait until the very end to start learning from it. These early runs can expose weak assumptions, identify awkward edge cases, show where a policy is doing something unexpected and, perhaps most usefully, let other people working on LVC tell us what questions we have forgotten to ask. 

So treat what follows as a working discussion about what can be designed for.

The numbers will get better. The interesting part is already beginning to emerge in the choices.

What's next

These five variants are a starting point rather than the end of the exercise.

We will be publishing fuller results for these policies, along with additional variants, technical documentation and selected outputs from the full model as the research progresses. The aim is not simply to settle on one preferred version of LVR, but to make it easier to see what different designs accomplish, where they create problems, and what those trade-offs look like in practice.