Repeat After Me: Build Houses, But Use Good Economics

A Financial Times opinion piece addresses housing supply issues, but uses bad economics. Let’s correct that.

Thank you for reading our work! If you haven’t yet subscribed, please subscribe below:

If you would like to support us further, please consider sharing and liking this article!

As Nominal News grows larger, we will be able to make this a full-time project, and provide more content. If you would like to suggest a topic for us to cover, please leave it in the comment section.

---

An opinion piece in the Financial Times (FT) “Repeat after me: building any new homes reduces housing costs for all”, discussed the issues of housing, housing costs, and housing supply. The headline – that building any type of new home, whether low cost or luxury – reduces housing costs for all is correct, as economic research has shown. However, some other parts of the piece are questionable. Let’s discuss.

Overview

The main claims of the FT piece can be summarized as follows:

1. People misunderstand how increasing housing supply will affect prices.

2. Building any type of housing (luxury or affordable) will improve housing affordability.

3. Based on the author’s calculation, policies such as upzoning (allowing for higher density construction on a plot of land) can have a very large effect on housing prices – a 25% reduction in prices – which stands in stark contrast to research we described here, where prices only fell by 1.9%.

We agree with claim 2. Both empirical and theoretical research have often demonstrated this to be the case, and the FT piece does a good job of explaining it. Basically, building luxury housing allows high income individuals to upgrade their home – but now their previous home is available for others who have lower incomes.

However, claims 1 and 3 have certain issues. Today, we will look into Claim 3, while tomorrow we will focus on Claim 1.

Claim 3 - Impacts of Zoning

Summary of Argument 

Although economics is often defined as the study of allocation of resources, to me, economics is the study of causality. Often we are interested in whether and how a certain policy, such as expanding the child tax credit, impacts the economy. Regarding housing costs, we are often interested in what policies can we enact to make housing more affordable. Upzoning is often touted as one such policy.

In the FT piece, the author looked at the upzoning conducted in Auckland, New Zealand and summarized the analysis in the following chart: 

The author compared Auckland’s rents to the rental prices in another city in New Zealand, Wellington. The Auckland upzoning was enacted in 2016 (the dashed line). At the time, the nominal median monthly rent in Wellington was 25% lower than in Auckland. Seven years later, the rents were comparable. Thus, the author claims that upzoning resulted in a 25% reduction in rental prices.

Methodological Errors

In the above analysis, the approach used (implicitly) by the author is called difference-in-difference method – commonly referred to as diff-in-diff. Diff-in-diff is a great statistical method that aims to isolate the causal effect of changing one variable (in the case here, the variable is zoning policy). Diff-in-diff works in a situation where we have a specific 'treatment' occurring to something/someone (in this case, the treatment given to Auckland is 'changing the zoning laws in 2016') and the treatment not occuring to something/someone (Wellington did not have changes in zoning laws in 2016). The idea of diff-in-diff is to first look at the difference of the outcome variable of interest to us (in this case, rent prices between Auckland and Wellington) prior to the treatment (prior to the 2016 upzoning law). This the base level difference between the two observations (Auckland and Wellington) that we are studying. Next, we look at the difference of the outcome variable (rent prices between Auckland and Wellington) post-treatment (after enacting upzoning laws in 2016). The difference between these two differences (this is w

here the name comes from) is the causal impact of the variable we changed (i.e. upzoning). The key benefit of this method is that the level of the outcome variable (rental prices) between the treated (Auckland) and not treated (Wellington) does not have to be the same to begin with. Visually, diff-in-diff can be seen with this example:

The difference between P1 and S1 is the initial difference in variable Y. At Time 1, P gets treated. If P were not treated, at Time 2 it would be at Q. The difference between Q and S2 is the same as the difference between P1 and S1. Due to the treatment, P actually got to P2. So the effect of the treatment is P2 - Q, which is numerically the same as: (P2 - S2) - (P1 - S1).

This approach is exactly what the FT author did. He looked at the difference between rents in 2016 prior to the upzoning – Auckland rents were 25% higher than Wellington rents. He then looked at the difference in 2022 after the upzoning – there was approximately a 0% difference. Therefore, the effect of upzoning is 0% - 25% = -25% reduction in rental price. 

The problem with using this method here is that you need to satisfy one key assumption to be allowed to use diff-in-diff in such a way. The assumption is called the "parallel trends" assumption.

Parallel Trends Assumption

This assumption states that absent the treatment, the original difference would remain – that is without upzoning, the difference in rental price between Auckland and Wellington would remain 25% (i.e. in the chart above, P would be at point Q in Time 2). Intuitively, what this assumption means is that the two cities are very similar to each other and would grow over time in the same fashion. For example, immigration patterns (i.e. how attractive each city is for citizens) would need to have remained the same, and no major new local policies would have been enacted (for example, if Wellington or Auckland governments undertook large investments or enacted family support policies, that would be a concern regarding satisfying the parallel trends assumption). Specifically, per this assumption, their rental price growth rates would have remained the same if Auckland did not perform the upzoning. This is a big assumption and one of the reasons why diff-in-diff methods, although brilliant tools, are so difficult to use. There is no formal way to show that this assumption is satisfied, besides looking at a graph of the outcome variable over time and seeing if the lines, prior to treatment, are parallel (the difference would stay constant over time).

In the case of the FT article, the author does not tackle this question. Why should we expect Auckland and Wellington to have similar growth patterns? Just being in the same country is not sufficient evidence. Moreover if we look at the graph again, prior to 2016, the Auckland and Wellington rental price growths are not parallel.

Interestingly, if we look at the wider New Zealand market rents, from a different analysis cited by the author of the FT piece, it appears that Wellington’s median rent price behaved very differently than prices in either Auckland or the rest of New Zealand, while Auckland and the rest of New Zealand had similar growth trends. Furthermore, it is worth noting that, the Auckland price trajectory did not change during the upzoning, and remained quite similar to the wider New Zealand market. 

The Effect of Upzoning

To look at what actually occurred in Auckland, Greenway-McGrevy and Phillips (2023) undertook a more rigorous analysis at the question (the FT piece indirectly linked to this study). This study approached the question quite differently. Greenway-McGrevy and Phillips looked at what happened in areas that were upzoned in Auckland and compared them to areas in Auckland that weren’t upzoned. Note, this is also a diff-in-diff, but a diff-in-diff that appears more likely to satisfy the parallel trends assumption, because they focused on areas within the same city.

Greenway-Mcgrevy and Phillips found that because of upzoning, the number of dwellings increased by 4.1%. Interestingly, this is quite a bit larger than what was found by Anagol, Ferreira and Rexer (2023) when they looked at the Sao Paulo market, who found only 1.9% increase in housing supply. 

Upon closer inspection, I noticed that there is a slight difference between what the two papers are measuring. Greenway-Mcgrevy and Phillips focused on the number of permits issued, while Anagol et al. studied the actual number of new housing units built. This can account for this discrepancy, as 1) it takes time to convert permits into houses, and 2) not all permits might turn into houses (Greenway-McGrevy and Phillips, however, argue that in Auckland, the conversion rate is almost 100%, although this may take time). 

The Greenway-McGrevy and Phillips estimate of upzoning effects in Auckland seems much more reasonable than the one stated in the FT article. Although Greenway-Mcgrevy and Phillips do not address how much prices changed, we can infer from the increase in supply (at most 4.1%) that the price differential will not be too dissimilar from what was observed in the Sao Paulo study (house prices were 0.5% lower). 

Conclusion

Recapping the key FT piece claim:

Claim 3: Based on the author’s calculation, policies such as upzoning (allowing for higher density construction on a plot of land) can have a very large effect on housing prices – a 25% reduction in prices 

This claim is, unfortunately, wrong. The method used to show that upzoning had a 25% reduction on house prices in Auckland is incorrect. Moreover, numerous studies – 

• Freemark, 2019 showed that upzoning in Chicago had little to no impact;

• Murray and Limb, 2021 found no effects of upzoning in Brisbane; 

• Anagol, Ferreira and Rexer (2023) show that upzoning reduced prices by 0.5% in Sao Paulo;

– have shown that upzoning is not a policy that will have large impacts on housing supply and affordability. Thus, although building more housing, as stated by the FT author would alleviate the housing affordability crisis, the FT author proposes solutions that unfortunately would have little impact on the problem.

The FT piece shows the dangers of using a comparative analysis. Often, we don’t realize that we are actually comparing apples to oranges, and that we are not establishing any causal link. Such comparisons are very common, as they ‘sound’ logical. However, economics shows us how to properly implement comparative analysis via the diff-in-diff method. When used well, the diff-in-diff method can be extremely insightful – one of the most famous uses of diff-in-diff showed that raising minimum wages do not necessarily increase unemployment. 

Tomorrow, we will discuss claim 1. 

---

Interesting Reads from the Week

1. Article: The winner of this year’s Nobel Prize in Economics was Claudia Goldin for her work on women in the labor market.

2. Article/Podcast:

   Economics Happy Hour
   discusses the work of Claudia Goldin and the Nobel Prize.

3. Substack Article:

   Alice Evans
   goes over the seminal works of Claudia Goldin.

---

Photo by Matteus Silva.

If you enjoyed this article, you may also enjoy the following ones from Nominal News:

Early Child Investment - Child Tax Credit (January 15, 2023) – the benefits of the expanded child tax credit to society.

To Compete or Non-Compete (April 30, 2023) – why non-compete clauses do not solve any issues, but only create costs.

Correcting Economic Understanding of US Senators (March 19, 2023) – during the Senate hearing with Federal Reserve Chairman Jerome Powell, several Senators made questionable economic statements.