The Firm Next Door: Using Satellite Images to Study Local Information Advantage

• Published date: 01 May 2021
• Author: JUNG KOO KANG,LORIEN STICE‐LAWRENCE,YU TING FORESTER WONG
• Date: 01 May 2021
• DOI: http://doi.org/10.1111/1475-679X.12360

DOI: 10.1111/1475-679X.12360

Journal of Accounting Research

Vol. 59 No. 2 May 2021

Printed in U.S.A.

The Firm Next Door: Using Satellite

Images to Study Local Information

Advantage

JUNG KOO KANG,∗LORIEN STICE-LAWRENCE,∗

AND YU TING FORESTER WONG∗

Received 2 December 2019; accepted 22 February 2021

ABSTRACT

We use novel satellite data that track the number of cars in the parking lots

of 92,668 stores for 71 publicly listed U.S. retailers to study the local informa-

tion advantage of institutional investors. We establish car counts as a timely

measure of store-level performance and find that institutional investors adjust

their holdings in response to the performance of local stores, and that these

trades are profitable on average. These results suggest that local investors

have an advantage when processing information about nearby operations.

However, some institutional investors do not adjust for the quality of their

∗University of Southern California

Accepted by Luzi Hail. We thank an anonymous referee for valuable comments that have

greatly improved this manuscript, as well as participants at the 2020 Journal of Accounting Re-

search Conference. We are also grateful to Eric Allen, Chad Ham, Jared Jennings, Xiumin

Martin, Dawn Matsumoto, Darren Roulstone (discussant), Eric So, Rencheng Wang, Regina

Wittenberg Moerman, Andrew Wu (discussant), and workshop participants at the 2019 Con-

ference on Emerging Technologies in Accounting and Financial Economics (CETAFE), the

University of Illinois Young Scholars Research Symposium, the University of Melbourne, the

University of Washington, and Washington University in St. Louis for their helpful com-

ments and suggestions. We also thank Orbital Insight for graciously providing the satellite

data used in this paper, Han Stice for sharing data on newspaper closures and layoffs used

in supplemental analyses, and Katherine Bruere for her research assistance. A previous ver-

sion of this paper was circulated under the title “Attention, Acquisition Costs, or Executive

Information? Using Satellite Data to Uncover the Sources of Local Information.” An on-

line appendix to this paper can be downloaded at http://research.chicagobooth.edu/arc/

journal-of-accounting-research/online-supplements.

713

© University of Chicago on behalf of the Chookaszian Accounting Research Center,2021

714 j. k. kang, l. stice-lawrence, and f. wong

local information and continue to rely on local signals even when they are

poor predictors of firm performance and returns. This overreliance on poor

local information is reduced for institutional investors with greater industry

expertise and those with greater incentives to maximize short-term trading

profits.

JEL codes: G23, G30, G40, G41, K00, L81, M40, M41, M48, O33

Keywords: satellite images; store-level performance; institutional investors;

local advantage; overweighting; processing costs; alternative data; big data;

emerging technologies

1. Introduction

A large stream of literature has concluded that local investors benefit from

an information advantage (e.g., Coval and Moskowitz [2001], Bernile, Ku-

mar, and Sulaeman [2015]). However, these studies measure local informa-

tion advantage indirectly, using the profitability of local trades, making it

difficult to identify the information to which local investors have access. We

build on these prior studies by using novel satellite image data to measure

intra-firm variation in the performance signals of stores located in different

geographic areas. As a result, we can pinpoint the information signals that

institutional investors incorporate into their trades, and our results sug-

gest that one source of local information advantage is information about

the performance of nearby stores. We also study the extent to which insti-

tutional investors appropriately integrate this local information into their

trades.

Prior research has been largely agnostic as to the type of information

that drives local information advantage. On one hand, local exposure to a

firm may provide institutional investors with information about firm-wide

performance and events. For example, local investors may gain private

company-wide information through greater access to firm executives. Addi-

tionally, exposure to local stores might increase investors’ knowledge about

the firm’s policies and business model, which have implications for firm

performance. On the other hand, geographic proximity may lead local in-

vestors to have an information advantage relating solely to local operations

and performance. Our data and empirical tests explore this latter type of

local information.

In the framework developed in Blankespoor, Dehaan, and Marinovic

[2020, 2019], an information signal effectively becomes private informa-

tion if the costs to become aware of, acquire, or integrate (analyze) that

signal become too high relative to the expected benefits of using it. In our

setting, geographic proximity, social ties, or institutional knowledge may

lower the costs that institutional investors face to obtain information about

local stores, leading them to have systematically more information about

local operations relative to nonlocal investors. For example, institutional

investors may observe store busyness when they shop. They may know local

using satellite images to study local information 715

employees or managers who mention store performance, or have ties with

reporters who can keep them apprised of local updates. Or they may be in-

vited to events, engage in policy making, or participate in social clubs and

organizations that provide information about factors that cause shifts in

supply and demand, for example, the closure or opening of a competitor,

or new state or city ordinances. On the other hand, nonlocal institutional

investors wishing to obtain the same information would need to purchase

business intelligence, pore over public reports, or travel to the region, and

even then, might not observe signals that local investors can.1

To investigate whether institutional investors have an information advan-

tage relating to local operations, we use satellite image data on the number

of cars in nearby store parking lots as a measure of store-level performance.

We validate the data by showing that changes in store-level car counts are

positively associated with quarterly changes in firm-level sales and net in-

come, and with abnormal returns. Although we do not expect local institu-

tional investors to directly observe car counts, we expect the local informa-

tion signals they do observe to be positively correlated with local car counts

through their joint relation with the performance of local stores. Further,

we ensure that none of the investors that we study respond directly to the

satellite data itself by restricting our sample period to end before the data

became commercially available. Our main sample includes 92,668 stores of

71 publicly listed U.S. retailers during the years 2009–2015.2

We define an institutional investor as local to a store if they are located

within 50 km.3If investors have greater information about local operations,

they should respond more to the performance of local stores than to that

of nonlocal stores. We provide evidence of this intuition by showing that

investors adjust their stock holdings in line with the car counts of local

stores, but not with the concurrent performance of the overall firm. On

average, a one-standard-deviation increase in local car counts is related to

investors increasing their holdings by 9–14%. Further, we provide evidence

that trades that incorporate local car count information yield annualized

abnormal returns of roughly 2%. These abnormal returns are comparable

in economic magnitude to those documented in the previous literature

(Coval and Moskowitz [2001], Baik, Kang, and Kim [2010]) and suggest

1In principle, these same factors should also apply to retail investors, even though they are

likely less sophisticated and have fewer resources at hand. We test our hypotheses using insti-

tutional investors because we have access to location and trading data for only these investors.

2When Orbital Insight was formed in 2013, it developed its algorithms by analyzing past

satellite images. Thus, when the Orbital Insight data first became available for purchase in

2015, the company already had historical data starting in 2009. We use only the 2009–2015

data so that, by construction, none of the investors in our main tests were responding directly

to the satellite data.

3In the online appendix, section A, we show that our results are robust to using alternative

distances to define local investors including: 100 km, 100 miles, within the same state, and

within the same “super county,” where a super county includes all counties with overlapping

zip codes.