Disclosure “Scriptability”

• Author: JAMES R. MOON,KRISTIAN D. ALLEE,MATTHEW D. DEANGELIS
• Date: 01 May 2018
• DOI: http://doi.org/10.1111/1475-679X.12203
• Published date: 01 May 2018

DOI: 10.1111/1475-679X.12203

Journal of Accounting Research

Vol. 56 No. 2 May 2018

Printed in U.S.A.

Disclosure “Scriptability”

KRISTIAN D. ALLEE,

∗MATTHEW D. DEANGELIS,

†

AND JAMES R. MOON, JR.

†

Received 31 October 2015; accepted 20 December 2017

ABSTRACT

In response to the increasing use of computer programs to process firm dis-

closures, this registered report develops a new measure of “scriptability” that

reflects computerized, rather than human, information processing costs. We

validate our measure using SEC filing-derived data from prior research and

identify firm and disclosure characteristics related to it. In our planned hy-

pothesis tests, we find some evidence that the speed of the market response

to filings increases with scriptability, but find little evidence that scriptability

affects the incidence and speed of news dissemination by Dow Jones. In ad-

ditional analyses, we find that scriptability exhibits both positive and negative

associations with changes in information asymmetry between market partici-

pants, depending on the filing, trading window, and measure examined. We

also find little evidence that XBRL interacts with scriptability in a meaningful

way. Overall, our study broadens our understanding of information process-

ing costs and provides opportunities for new avenues of research.

∗University of Arkansas; †Georgia State University.

Accepted by Christian Leuz. This paper is the final Registered Report resulting from

the Registration-based Editorial Process (REP) implemented by JAR for its 2017 conference;

details of the process are available here: https://research.chicagobooth.edu/arc/journal-of-

accounting-research/2017-registered-reports. The accepted proposal and an online ap-

pendix for this report are available here: https://research.chicagobooth.edu/arc/journal-

of-accounting-research/online-supplements. We thank Patrick Caldon (Morningstar), Fran

Finnegan (SECInfo.com), J.W. Mike Starr (Workiva), Alex Zervakos (Securex) for excellent

industry perspectives on preparation and processing of SEC filings. We thank Semir Sarajlic,

Suranga Edirisinghe, and Kwai Wong for technical assistance. Wealso gratefully acknowledge

John Campbell, Scott Dyreng, Kurt Gee, Craig Holden, Stacey Jacobsen, Todd Kravet, Bill Mc-

Donald, Jeff McMullin, Volkan Muslu, Elizabeth Odders-White, Kyle Peterson, Mark Ready,

Bei Shi, Logan Steele, Brady Twedt, Ben Whipple, Ke Xu, and our anonymous reviewer for

help with this project, and conference participants at the 2017 Journal of Accounting Research

Conference for valuable insights and suggestions.

363

Copyright C, University of Chicago on behalf of the Accounting Research Center,2018

364 K.D.ALLEE,M.D.DEANGELIS,AND J.R.MOON,JR.

JEL codes: G10; G12; G14; G24; G30; G39; G41; M40; M41; M49

Keywords: disclosure; textual analysis; financial reporting; information pro-

cessing costs; high-frequency trading; XBRL

1. Introduction

Over the last several decades, the Internet has significantly increased both

the volume of information and the speed of information dissemination. In

response to these changes and increasing competition for returns, market

traders now utilize computing power to assist or replace human effort in

the acquisition and analysis of financial information and the execution of

trading strategies (Foxman [2016]). For instance, journalists estimate that

50% to 60% of all trades in the U.S. stock market are now made by al-

gorithms (Chilton [2014], Philips [2015]), often occurring within seconds

of information release (Lewis [2015]). As a result, the analysis of finan-

cial information in SEC filings has shifted away from manual processing

by investors and analysts toward programmers and computers in ways that

research has only begun to acknowledge.

In this paper, we examine corporate disclosure from a programmer’s,

or “scripter’s,” perspective.1We start by identifying two basic tasks that a

scripter, whether an investor doing her own analysis or a computer scientist

working at an investment bank, is likely to perform when analyzing firm

disclosures. First, the scripter searches for data of interest, such as the Man-

agement’s Discussion and Analysis (MD&A) section of the 10-K or a table of

executive compensation figures. Second, the scripter processes identified

data into decision-relevant information. We then formulate two measures

based on these tasks and one composite measure of “scriptability,” or the

ease with which a scripter can perform these tasks on a given disclosure. In

creating and analyzing our measure, we provide comprehensive descriptive

statistics and empirical evidence regarding several previously unexamined

dimensions of disclosure quality, such as the amount of disclosure that is

machine-readable, the ease of separating text from tables, and the presence

of table metadata that facilitates the extraction of key financial metrics.2We

1This research was conducted as a two-stage editorial process based on what is known in

other fields as “Registered Reports.” In stage one, authors submit a proposal that describes the

hypotheses they will test, the data they will gather, and (in considerable detail) the research

design and analyses they will use to interpret their results. In stage two, authors submit the

“Registered Report,” which describes the original intent and actual execution of the study

approved in stage one, along with the results and interpretation of planned and unplanned

analyses. Consistent with the registered report process, the only changes to our study from the

accepted stage one proposal are reported in online appendix F.

2The proper use of tables facilitates the identification of numeric data and its separation

from textual data. We evaluate each table in a filing and apply a series of conditions to

determine whether the table contains properly formatted numerical data or textual data

that would be better presented in a bulleted list or paragraph. For those tables that contain

DISCLOSURE “SCRIPTABILITY”365

also offer insight into dimensions of disclosure quality that affect both pro-

grammers and readers, such as the usefulness of section headings and the

extent to which the disclosure references external content.3As such, our

study highlights both the opportunities and challenges of using computer

programs to analyze financial information and expands prior research on

information processing costs and frictions in capital markets.

We conduct three validation tests on our scriptability measures and find

evidence suggesting that our measures capture the costs of computerized

information processing in firm disclosures (see online appendix E). We

then examine this previously unexplored dimension of disclosure quality

by identifying a set of firm and disclosure characteristics that we expect

to relate to scriptability. Prior research suggests that disclosure quality in-

creases with investment in the financial reporting function, so we expect

improvements in scriptability when managers make these investments. We

also expect that both technologically sophisticated managers and managers

relying on external providers (“filing agents”) for filing preparation are

more likely to have more scriptable disclosures. While we find some evi-

dence from these tests that investment in the financial reporting function

is related to higher scriptability, our results do not suggest that technologi-

cal sophistication in general translates into more scriptable filings. Overall,

our most notable finding in this test is that the use of a top filing agent, a

previously unexplored information intermediary, is associated with signifi-

cantly better scriptability. This finding provides additional validation of our

measure, as filing agents appear to improve filings along the dimensions we

measure.

Controlling for the above factors, we proceed to test two hypotheses

on the possible benefits of scriptability. First, market efficiency requires

complete and rapid price formation (O’Hara [2003]). Investors utiliz-

ing programmatic processing techniques should be able to identify, pro-

cess, and act upon disclosed information more quickly when disclosures

are more scriptable. Using an intraperiod timeliness (IPT) measure from

prior research (e.g., Butler, Kraft, and Weiss [2007], Bushman, Smith, and

Wittenberg-Moerman [2010], McMullin, Miller, and Twedt [2015]) as well

as a newly designed IPT measure based on volume, we hypothesize that dis-

closure scriptability positively relates to the speed of the market’s response

to the disclosure. Our strongest evidence is consistent with scriptability in-

creasing the timeliness of trading (volume IPT), particularly in the 8-K fil-

ing group. These results suggest that at least some traders are impeded by

poor scriptability. With regard to the timeliness of price formation, most

of our evidence fails to detect a significant association between scriptability

numeric data, we further evaluate numeric tables for the presence of HTML markup, row and

column consistency, and descriptive row names, all of which aid the processing of numeric

data into information.

3Figure 1 in section 2 illustrates how our components relate to our two task-based measures

and a composite measure.