Congrats! You’ve seen my poster at EALE 2024
Contents in this page:
- Quick Recap
- Robustness of Financial Results
- Alternative Design: Regression Discontinuity
- Let’s stay in Touch!
- Did You Guess?
1. Quick Recap
Wondered off from the poster and need a quick reminder about the project?
- What’s this about? I am evaluating a 2017 Swedish sustainability reporting reform’s effects on firms’ financial outcomes, sustainability policies and sustainability outcomes.
- How have I estimated this? I’ve used a Difference-in-Differences approach on all Swedish firms with more than 10 employees with regards to financial outcomes, and specifically, looked among listed firms for sustainability policies and outcomes. Additionally, I’ve created a proxy for compliance with the reform that’s simple, intuitive and captures the essence of compliance.
- What have I found? High compliance with the reform, some financial benefit in terms of greater sales, no effect on productivity (within-firm value-added), little to no changes in sustainability targets and sustainability outcomes.
- Do I trust these results? Yes, with an asterisk. First, most robustness checks (as in Table 3 and 4 below) provide stable point estimates regardless of sample or exact definition of treatment. For the full sample, pre-trends are not completely parallel for value-added per employee, and we may thus violate the PTA. Therefore, as an additional analysis, I perform an RD design.
The results can be summarized as:
2. Robustness of Financial Results
Job advertisements show we shouldn't worry about anticipation
The EU directive is from 2014 and became Swedish law in 2016, to apply from January 1st 2017. On the largest job advertisements platform, we see that firms post advertisements for sustainability officers and managers in line with the Swedish reform and very rarely before that. The figure below shows the _number of job advertisements specifically mentioning sustainability disclosure (blue), sustainability reporting (orange), and environmental reporting (green)_. The last term (green) has been in the public discourse for longer than sustainability reporting or disclosures.
The financial outcomes are a great second-best
Essentially, I want to capture whether a firm is "rewarded" by the market for being transparent and if they become more productive by becoming transparent. How can we capture that in absence of sales data from firms? First, "net turnover per employee" is equivalent to net sales per employee. The "net" here refers to revenues from sales after discounts and taxes. Since this measure is "per employee", I use it in my Difference-in-Differences setting without risk of comparing apples to oranges in terms of firm sizes. I capture firms productivity by a measure of value-added per employee.
3. Alternative Design: Regression Discontinuity
The alternative RDD design shows no statistically significant effects, but is also hard to evaluate since no single running variable determines treatment on its own, while the combined running variable captures multiple margins of the ATE. The RD also doesn’t take into account the condition of having fulfilled two of three criteria for each of the past two years, but focuses on a single year instead.
RD Details
The Regression Discontinuity Design relies on the fact that the reform has sharp cutoffs that are difficult to manipulate precisely. Firstly, there are three paths into treatment given the three cutoffs ({1, 2}, {1,3}, {2, 3}). Hence, a firm with over 250 employees that would like to remain untreated would have to manipulate its books with regard to the balance sheet (i.e. total assets) and net turnover. Secondly, the cutoffs are based on fulfilling criteria over a period of two years. So, for example, a firm that would like to stay under a net turnover of 350 million SEK would have to make sure to do so for one of the past two years to avoid SR this year. Overall, it is unlikely that a firm would harm its growth potential simply to avoid reporting on their sustainability policies. Thus, we would expect there to be no manipulation of treatment status.
The main identifying assumption is continuity, i.e., in absence of treatment we would not see any discontinuities at the cutoff. However, crossing the cutoff makes a firm switch from being untreated to treated, and therefore any jump in outcomes is directly attributable to being treated.
What this implies is that firms around the cutoff are as-if randomly assigned to treatment. We do not have reason to believe that firms are able to precisely manipulate their size (in terms of number of employees, total assets and net turnover) around a combination of cutoffs for many reasons.
First, Sweden has relatively strict Employment Protection Laws (EPL) that regulate employer-employee relations. This implies that employees cannot be fired without reason. Employers could however choose not to expand by hiring employees. However, for the purposes of this paper, it is unlikely that firms would downsize the number of employees or curb their growth in order to avoid complying with such a low-cost law. Having said that, number of employees, as a running variable, is still the easiest to manipulate of the three variables. Second, to avoid treatment, a firm would have to ensure that they are below the cutoff in at least two of the three running variables for size. We have no reason to believe that firms could precisely manipulate their size in terms of at least two of three firm size variables, simply to avoid maintaining a sustainability report.
Testing the manipulation assumption with a McCrary test and by plotting histograms shows no bunching around any of our cutoffs.
An alternative to evaluating each running variable separately is to use a standardized distance to cutoff as a running variable. I construct a standardized distance to a "combined" cutoff. Imagine each firm's size represented by a point in 3D-space. Each dimension represents standardized distance to the respective cutoff. So, 1 on the x-axis is 1 standard deviation away from 250 employees normalized to 0. Having a positive distance on a dimension means being above that dimensions threshold. Being above two of three dimensions (i.e., {x,y,0}, {x,0,z}, or {0,y,z}) means that the firm is treated.
Overall, there is no statistically significant effect of the reform on the main financial outcomes. The results vary with increasing the bandwidth from the MSE-optimal bandwidth. Closest to the cutoff (and with the MSE-optimal bandwidth), there is no significant difference between treated and untreated firms, though sales have overall more positive coefficients than value-added. Estimating the RDD results each year separately shows no clear pattern in the RDD estimates per year.
There's a lot of exciting work in progress in this part of the paper!
4. Let’s Stay in Touch!
I’m so glad that you’ve shown this much interest in my work and in me. I’d also like to get to know your work, your city, and your interests!
E-mail me via marcos.demetry@lnu.se and say “Hello!”
I’m also very interested in segregation studies (how come most taxi drivers in Sweden have a foreign background? Why are most pharmaceutical students in Sweden females? Which factors would need to be tweaked in order for individuals’ self-selection into neighborhoods not create a segregated city?) as well as crime economics (specifically, school shootings in the U.S., prosecution duration and recidivism in Sweden, and the economic impact of crime on neighborhoods more broadly).
5. Did You Guess?
Press here to reveal the answers
- These two industries had the greatest number of big firms in 2017: Manufacturing and Transportation - These three industries had the greatest share of big or treated firms in 2017: Mining, Electricity, and Real Estate - Between 2008 and 2020, CO2 emissions among listed firms in Sweden decreased by around 17,000 kilotonnes, or 30% - Would you a priori have expected positive, null or negative effects on financial outcomes from having a sustainability report? Let me know what you think!