Can Markets Discipline Government Agencies? Evidence from the Weather Derivatives Market

• Author: AMIYATOSH PURNANANDAM,DANIEL WEAGLEY
• DOI: http://doi.org/10.1111/jofi.12366
• Published date: 01 February 2016
• Date: 01 February 2016

THE JOURNAL OF FINANCE •VOL. LXXI, NO. 1 •FEBRUARY 2016

Can Markets Discipline Government Agencies?

Evidence from the Weather Derivatives Market

AMIYATOSH PURNANANDAM and DANIEL WEAGLEY∗

ABSTRACT

We analyze the role of financial markets in shaping the incentives of government

agencies using a unique empirical setting: the weather derivatives market. We show

that the introduction of weather derivative contracts on the Chicago Mercantile Ex-

change (CME) improves the accuracy of temperature measurement by 13% to 20%

at the underlying weather stations. We argue that temperature-based financial mar-

kets generate additional scrutiny of the temperature data measured by the National

Weather Service, which motivates the agency to minimize measurement errors. Our

results have broader implications: the visibility and scrutiny generated by financial

markets can potentially improve the efficiency of government agencies.

HOW ARE THE INCENTIVES OF GOVERNMENT agencies shaped? This question is of

fundamental importance to both economists and policy makers. In this paper,

we examine a novel channel through which government agencies’ behavior may

be affected: financial market pressure. Specifically,we ask whether government

bureaucrats improve their performance when their actions are closely scruti-

nized because they affect financial markets? We address this broad question

by exploiting an interesting empirical setting: the launch of a new financial

market, namely, exchange-traded weather derivatives, that has payoffs linked

to the measurement of temperature by a government agency—the National

Weather Service (NWS). Financial markets depend on the actions of govern-

ment agencies in a number of other settings as well, such as commodity futures

and catastrophic insurance markets.1We focus on the weather derivatives

∗Purnanandam is at the Ross School of Business, University of Michigan. Weagley is at the

Scheller College of Business, Georgia Institute of Technology. The authors thank Gautam Ahuja;

Taylor Begley; Sugato Bhattacharyya; Ing-Haw Cheng; Charlie Hadlock; Zoran Ivkovic; Vojislav

Maksimovic; Paolo Pasquariello; Uday Rajan; Michael Roberts; TylerShumway; Ren ´

e Stulz; Maciej

Szefler; Vish Viswanathan; two anonymous referees; an anonymous associate editor; and seminar

participants at Dartmouth, Michigan State University, University of Michigan, and NBER meet-

ings on Economics of Commodities Markets for helpful comments. We thank CME and MDA Inc.

for providing data and several clarifications on the weather derivatives market. The authors are re-

sponsible for all remaining errors. We have received no external financial support for this research

and are awave of conflicts of interest.

1For example, a number of dairy, livestock, and commodity contracts on the CME and Chicago

Board of Trade (CBOT) are settled based on measures provided by the U.S. Department of Agri-

culture (USDA). Information provided by the U.S. Geological Survey (USGS) plays a crucial role in

the pricing of earthquake insurance. Similarly, market participants rely on the National Oceanic

DOI: 10.1111/jofi.12366

303

304 The Journal of Finance R

market because it provides us with a clean empirical setting to address the

central question.

The Chicago Mercantile Exchange introduced the first exchange-traded

weather derivative instruments in 1999. Since then, the CME has introduced

weather contracts on a number of U.S. cities in a staggered fashion. A vast

majority of these instruments are temperature-related, allowing end-users to

hedge their exposure to undesirable warm or cold weather conditions. These

contracts are city-specific and are settled based on the temperature readings of

a specific NWS weather station within or near the contract city. The weather

stations are prone to measurement error due to factors such as improper cali-

bration of the sensors, poor maintenance, and lax monitoring of the equipment.

We examine the effect of CME weather derivative introduction on NWS tem-

perature measurement performance. The introduction of derivative contracts

directly ties the NWS-reported temperature at these stations to the large eco-

nomic interests of traders and hedgers in the market. The resulting increase in

market scrutiny can impose significant reputation losses on the NWS and its

managers if the measurement is inaccurate. Thus, the introduction of a weather

derivatives market provides a reasonably exogenous shock to market-based

pressure on the NWS employees responsible for measuring the temperature at

a designated weather station in the contract city.

As of June 30, 2012, there are 24 U.S. cities with temperature-related deriva-

tive contracts traded on the CME. These contracts were issued in four waves:

1999 to 2000, 2003, 2005, and 2008. Our empirical setting allows us to compare

the improvement in temperature accuracy of the weather stations with deriva-

tives (the treatment group) around the derivative launch dates with a set of

nonderivative stations (the control group) during the same period. The stag-

gered nature of the derivative launch allows us to separate improvements due

to derivative introduction from the effect of any time trend in error rate or any

general improvement in the NWS’s technology over time. Furthermore, unlike

stocks or bonds, the variable underlying the weather derivatives contract is

not a traded commodity. Thus, the introduction of the derivative contract is not

going to affect the underlying weather itself, which helps us establish a causal

relationship from financial market pressure to measurement accuracy.

Our main performance measure captures temperature measurement errors

made by the NWS. We define measurement error as any discrepancy between

the initial temperature recording of the NWS weather station and the subse-

quent corrected values issued by the National Climatic Data Center (NCDC,

a sister agency of the NWS) or a private third party. Using a sample period of

1999 to 2012 for 49 treatment and control stations, we find that the median

weather station in our sample has an error rate of 12 days per year. Using a

difference-in-differences model, we show that the treated station’s error rate

and Atmospheric Administration (NOAA) for assessing, pricing, and settling a number of contracts

such as hurricane and flood insurance. At a much broader level, across the world’s markets, partic-

ipants focus closely on government-released macroeconomic data such as GDP growth, inflation,

and unemployment for their investment strategies.

Can Markets Discipline Government Agencies? 305

declines significantly after the introduction of weather derivative contracts.

Depending on the model specification, the point estimate ranges from 1.6 to

2.4 days of improvement in measurement accuracy at the treated stations. The

decline in error rate represents about 13% to 20% of the median error rate in

our sample. Our results remain similar if we omit the nonshocked stations from

the sample, and thus achieve identification solely from the group of derivative

stations. On a station-by-station basis, we find that 18 of 20 stations experience

an improvement in error rate after weather derivative introduction. Further-

more, we show that our results are not driven by differential trends across

the treatment and control groups. Taken together, these results paint a clear

picture: weather stations with derivative contracts have a lower incidence of

inaccurate data after their recorded temperature numbers become reference

points for billions of dollars of financial contracts in an open market.

In our next set of tests, we show that the improvement is larger for stations

with higher economic interests. Specifically, we find the improvement to be

higher for stations that received derivative contracts in earlier waves. These

stations are likely to have relatively higher economic interests based on CME’s

revealed preferences. Moreover, we find that the effects are stronger for cities

that are likely to have higher energy demand and hence higher interest in

weather derivative products.

Our results are consistent with the idea that bureaucracies are concerned

about a loss in reputation and improve performance when the probability of

a loss in reputation increases. The additional visibility and scrutiny of an

agency’s actions by market forces can create significant reputational risks for

the agency, which in turn can have adverse consequences for the agency’s

budget and its managers’ career paths (Wilson (1989)).2Consequently, these

concerns are likely to motivate the bureaucrats to control managerial slack in

response to increased pressure from financial markets.3

An alternative channel that can explain our main result is based on tech-

nological improvements at the treatment stations. If the NWS deploys more

technical resources at the treatment stations precisely at the time of deriva-

tive launch, then we are likely to observe lower error rates even without any

improvement in managerial effort. While we cannot conclusively establish the

effect of the managerial effort channel, we provide several tests to show that

the technological channel is unlikely to explain all of our results. First, we show

2See Dewatripont, Jewitt, and Tirole (1999) for a discussion on the importance of career concerns

in bureaucracies.

3This mechanism is different from the traditional models of corporate finance that study the

disciplining role of financial markets in the context of corporate managers. Numerous important

contributions in this area have been nicely summarized in survey articles such as Shleifer and

Vishny (1997), Gillan and Starks (1998), Black (1998), Karpoff (1998), Romano (2001), Hermalin

and Weisbach (2003), and Becht, Bolton, and R¨

oell (2003), among others. This line of research

argues that market participants such as blockholders and pension funds can discipline corporate

managers through explicit or implicit performance-based incentive contracts. See Shleifer and

Vishny (1986), Holmstr¨

om and Tirole (1993), Burkart, Gromb, and Panunzi (1997), Bolton

and Von Thadden (1998), Kahn and Winton(1998), Gopalan (2009), Admati and Pfleiderer (2009),

and Edmans (2009), among others.