Mini Post #16: A Moral Multiplier?
Mini Post #16: A Moral Multiplier?
Can government spending projects inspire people to pay their taxes?
Lately, I’ve been writing a shorter, narrower post focusing on one specific paper instead of usual deep dives. Last time, I wrote about whether election monitors have an impact on election results.
This is the tab with all the previous posts.
Sometimes, you hear discussion of government spending framed around the “multiplier effect” - basically, in macro 101 terms, because government has to pay either people or companies if it does anything, then it can raise corporate and household incomes, “stimulating” the economy. And, since people and companies pay taxes on their own expenditures, this means that (at least in principle) some of the money can be recouped that way - and if the economy is genuinely stimulated (i.e. is bigger), then that effect can be large enough to offset the cost. Now, this is an interesting empirical and theoretical debate (my stance is that government spending cannot stimulate consumption permanently, but can stimulate investment, and that neither type of spending can pay for itself), but beyond the scope here - besides introducing the notion that, perhaps, government spending can lead to more revenue.
Latin American countries, owing to a variety of factors (mostly cultural issues and low state capacity to prosecute evaders) also have very low rates of tax compliance - that is, people just don’t pay taxes they statutorily are obligated to pay. A common saying, at least in Latin America, is that government taxes are too high for the services provided (usually framed as an international comparison) - that is, the taxes people pay are very high, but the quality of services is very low. This ties with something called the “fiscal contract”, where the government gained the power to tax the population and spend the money by investing in projects (mostly wars) that the taxpayers initially, and society as a whole later, desired. This means that an interesting question can emerge: can there be a “moral multiplier” of government spending, where government investment inspires people to pay their taxes?
Let’s look at Mexico for an example. Mexico City (from now on also abbreviated as CDMX, short for Ciudad de Mexico) has a tax evasion rate of roughly 40%, for reasons obviously beyond the scope of this post. The paper in question is titled “The Fiscal Contract up Close: Experimental Evidence from Mexico City” (2024) by Anne Brockmeyer, Francisco Garfias, and Juan Carlos Suárez Serrato, and studies the effect of an infrastructure spending program on tax evasion among exposed households.
In CDMX taxes, particularly property tax, are directly linked to the funding of the state, and especially infrastructure projects (because of legal frameworks for infrastructure, mainly). One such project is the Hábitat (from now on: just Habitat) program, where, starting in 2003 (but chiefly in 2009 to 2011) , the Mexico City government invested in local amenities in lower income neighborhoods - for example street paving, water works, and urban renovation projects like sidewalks, community facilities, and parks. The costs were split between the federal government and the local/municipal governments in question.
According to theories of the fiscal contract, what would happen next is simple: if the government provides goods or services that people want, then those who benefit at the very least will reciprocate and pay their taxes. The study tests this directly, by measuring the impact of the Habitat projects on nearby tax compliance, as well as by performing an experiment where some taxpayers are reminded that their tax payments contribute to the Habitat program.
One quirk of the Habitat program that the study exploits is that Habitat projects were evaluated with a nationwide randomized controlled trial, where a group of municipalities meeting certain characteristics (high poverty, deficient infrastructure, no land ownership related conflicts) was divided in “polygons” of at least 15,000 people, and various polygons were assigned a Habitat project (or not) at random. In CDMX, the sample consisted of 20 polygons across four municipalities totalling 7,947 lots. Twelve of these polygons (with 3,210 lots) were assigned to be control polygons. On average, Habitat projects invested $141 US dollars per property in the treated polygons, whichis 3.7 times higher than the average property tax payment, and 1.8 times the average monthly per capita income in the sample. Due to successful evaluations finding that Habitat projects were successful at improving quality of life and boosting income and investment, such that they were expanded starting in 2017.
The analysis is based on administrative CDMX tax records that cover each property in the city register, and the authors measure both whether any tax payments were made, as well as what share of the tax liability owed was paid, as well as the average tax payment by all relevant taxpayers, and the property’s value. The authors also utilize the Habitat database to find the relevant polygons for each property as well as when each project was started and finished. They also survey houses which are given questionnaires of two kinds (long and short).
The average property tax compliance rates in the control group is 49%, lower than the 65% average for Mexico City, consistent with the higher poverty shares in the treated neighborhoods. Three quarters of Habitat treated properties and over two thirds of control properties were included in the final database, with no significant observable differences in match rates. The treatment and control lots do not have any statistically significant differences in 23 of 25 characteristics (treated polygons are less likely to have open stores or be adjacent to garbage dumps).
The baseline analysis is fairly straightforward: utilizing Intention to Treat, they measure the effect of the Habitat project being assigned to a given polygon. This is a slight difference with most RCTs, which measure the Average Treatment Effect - the difference between receiving the treatment and not receiving it after it’s assigned. But anyways, the ITT is calculated as the difference in outcomes for lots in treated and untreated polygons with controls for property, polygon, and municipality level fixed effects, as well as time fixed effects, and a series of pre-Habitat factors that may affect urban quality in general.
To begin, the authors find that Habitat (not very surprisingly) improved infrastructure quality for all types of public works, but significant effects for water access and sewage coverage (by 16% and 10% respectively). Constructing an index that aggregates all infrastructure quality measures, the Habitat projects raised conditions by 0.22 units, to a statistically significant level.
Now onto the variable of interest: the impact (ITT, remember) of Habitat projects on the tax compliance at the polygon level, at the block level within each polygon, and at the street per block level. Across all three approaches, and with and without controls, the author find no effect of benefitting from public works on making any tax payments whatsoever, or on the ratio of tax payments to total liability. The effects are so small that, given the methodology, the authors can rule out even small effects. If we also look at the impact of Habitat projects on fiscal property value (which would result in higher revenue for existing lots due to upgraded amenities), the authors find no change either - meaning the project just didn’t result in any increases for the bulk of property owners, neither at the extensive margin (i.e. "making versus not making payments”) or the intensive margins (how many payments were made”).
To adjust for heterogenous qualities per lot, particularly nominal tax liability, tax rates on properties, and individiual and pre-Habitat neighborhood level tax compliance, the authors try to test whether people who believed their neighbors were already paying taxes (i.e. low liability, low rate, and high compliance) were affected more strongly. The authors include a control for heterogeneities for each variable and an interaction between the variables and Habitat assignment (in case they correlate in some way); there are also no statistically significant effects adjuting for these factors. Other tests of heterogeneity also fail to detect any big differences within groups.
This means that, generally speaking, CDMX residents benefitted from Habitat programs simply did not care about the benefits enough to pay their taxes at higher rates than previously (in fact, tax compliance was reduced across both the treatment and control groups). But could it be that the residents didn’t realize that Habitat was funded from their taxes in particular? Utilizing a similar methodology but for household survey data, the authors find that the program was 7.5 p.p. better known among beneficiaries than the general public (versus a mean of 20%), and for CDMX, the effect on “fame” was even larger - an increase of 9.4%, i.e. nearly 50%. Using these surveys, the authors find that Habitat led to statistically significant increaes of 3.7% in trust in public officials, and 3.4% of trust in neighbor leaders (5.6% and 7.8% respectively for Mexico City). Given the effects of Habitat on neighborhood level infrastructure quality, and on economic and quality of life outcomes, then this means that the lack of a “moral multiplier” wasn’t driven by ineffectiveness of the Habitat program, but rather, by the fact that taxpayers simply do not reciprocate government spending of that nature, at least not for a program of these qualities.
As one final test, the authors utilize results from a 2014 government campaign that sent letters to 80,000 debtors (with a control group of 10,000) urging them to pay their tax debts. Half the letters emphasized that the tax revenue was used for infrastructure projects such as Habitat, while the other half emphasized potential legal sanctions. Comparing the households that received a letter with the households in the Habitat program, and looking at the tax compliance of both, the authors find that households that received letters increased their tax compliance rates by 5.1% for property owners receiving the one emphasizing public services… and 7.3% for the ones who were threatened by sanctions. Both effects are statistically significance. And while receiving a letter raised compliance among Habitat beneficiaries, this did not raise it with any statistically significant difference versus letter non-recipients: there are no statistically significant interactions between either type of letter or Habitat.
This all means, put together, that households who benefitted from public works program saw their standards of living rise and quality of infrastructure improve, but did not increase their tax payments in response across any relevant metric, even when specifically reminded that their taxes paid for those very projects. This was not driven by lack of information or by inefficiencies in the projects, and beneficiaries saw higher trust in politicians and other civic figures. All in all, this just means that “moral suasion” (whatever that is) is not very good at convincing people to pay their taxes.
To finish up the post, some links
The paper in question
A paper by McIntosh et al studying the impacts of Habitat on quality of living
A paper by Iacovone et al studying the impacts of Habitat on economic conditions
A similar paper by Carrillo, Castro, & Scartascini finding that a lottery for property tax payers in Argentina increased tax compliance
A paper by Antinyan & Asatryan about nudges (letters and such) and tax compliance outcomes
A paper by De La O and a lot of other people carrying out similar evaluations fo r a number of developing countries
A paper by Bergeron, Tourek, & Weigel on the impact of tax reductions on tax compliance for a given rate of government enforcement capacity
A paper by Okunogbe & Tourek studying technology and tax compliance
An OECD study about tax evasion across 38 OECD members