Data Collection & Creation
A wide variety of public agency, private, and open-sourced data was utilized to create a robust dataset including information from 2000-2021 across hundreds of rules and thousands of comment letters/meetings.
2.1 SEC Rule Publication
The text of all Proposed and Final Rules in addition to the associated comment letters and meeting memorandums between 2000 and 2021 was collected. The text was then parsed to create a novel dataset of textual information about each textual element (e.g. word count).
The core of our analysis is based on published rules from the Federal Register and their associated comment letters and meetings. In total 330 Proposed Rules and 303 Final Rules were identified. There were a number of cases in which there was a Proposed Rule issued, and no Final Rule was adopted by the Commissioners of the SEC. In total 300 proposed-final rule pairs were identified or 80.08% of all rule extracted from the Federal Register. The released of Proposed and Final Rules varied across time significantly between 2000-2021 (see Figure 2.1 and 2.2).
.png)
.png)
2.2 Classifications & Coalitions
[ ]
In sum 22,630 comment letters and 4,003 meetings were identified originating from 25,607 non-unique entities. After matching commenting entities together that represented the same organization the number of entities that commented or meet at least once between 2000-2021 was 9,792 and 1,350, respectively (see Appendix 1.2 for description of this process). Across both comment letters and meetings 9,945 unique entities submitted a comment to or met with the SEC a total of 26,663 times between 2000-2021. There was significant variation overtime in the number of comments and meetings received, moving within relation to the number of proposed rules released during the year.
.png)
.png)
The 9,945 unique entities that submitted comment to or met with the SEC at least once were classified into one of six categories (see Table 1.1 for category descriptions). As discussed in the Data Collection & Creation section, these classifications were manually assigned and then programmatically verified so that a single entity retained the same classification throughout the entire dataset. Table 2.1 shows the classifications of every entity in the dataset without consideration for the number of comment letters or meetings participated in. Table 2.2 shows the number of classification occurrences including the rate at which an entity participated in the rulemaking process.
.png)
.png)
As seen in Table 2.1 and Table 2.2 corporations and interest groups make up the bulk of comment letters and meetings. In addition to individual entity participation entities also joined coalitions in their attempt to influence the SEC rulemaking process. Within the 2000-2021 comment letter dataset 516 comments were from a coalition or 2.28% of all comment letters in this period. Coalition meetings occurred 999 times or 25.07% the whole meeting dataset. It is clear that coalitions are far more common, both absolutely and relatively, within meetings rather than comment letters.
.png)
.png)
Of the cumulative 1,515 coalitions from 2000-2021 735 (49.13%) were ad hoc in nature with that coalition only occurring one time. The mean size of a comment coalitions and meeting coalitions was 4.76 and 3.10 with a standard deviation of 4.24 and 1.97, respectively. Across time there was not significant variance in the size coalitions (with the exception of 2014) across time (see Appendix 2.7).
.png)
Coalitions are typically formed by entities within the same classification from 2000-2021 927 coalitions or 61.97% of all coalitions were made up of entities with the same classification. The number of uniformly classified coalitions varied overtime, but it seems to have generally increased following 2005 (see Figure 2.8).
.png)
2.3 Financial Data
[ ]
The data accessed through WRDS and Compustat contained 526,143 observations from 2000-2021. This quarterly data was over 27,363 different publicly traded companies. The average market value was $5,434.37 million with a standard deviation of $28,966.46 million. Through the manually matching process 498 entities were identified within the WRDS dataset or 1.79% of reported companies and 4.92% of unique entities. These matches and their susbquent observations were not uniformly distributed throughout time with some years (2009 and 2020) having more data available (see Figure 2.10).
.png)
Companies which participated in the rulemaking process were notably larger than those that did not with an average market value of $47,331.45 and standard deviation of $71,407.09 [Financial vales reported following are adjusted for inflation/deflation using the Federal Reserve’s GDP Deflator]. Compared to the average market value of $5,170.15 and standard deviation of $26,296.07 for companies which did not participate in the rulemaking process. See Figure 2.11 for the distribution of log-adjusted market values segmented by engagement with the rulemaking process.
.png)
The data accessed through IRS-990 contained 2.8 million observations and was reported annually for filers. On average the total assets at period end for an entity was 18.33 million with a standard deviation of 348.03 million. Within the IRS-990 data 731 associations with commenting or meeting entities were made or [#]% of the total unique filers within the IRS-990 data. Filers which participated in the rulemaking process at least once had a mean total assets of 549.53 million with a standard deviation of 5,132.98 million. There are also multiple classifications present within the IRS-990 data unlike in the WRDS or Federal Reserve data. Comparing the reported financial data for each classification shows clear differences.