honey bee pollination marketsandthe …...european honey bee (apis mellifera) is widely used to...

HONEY BEE POLLINATION MARKETS AND THE

INTERNALIZATION OF RECIPROCAL BENEFITS

RANDAL R. RUCKER, WALTER N. THURMAN, AND MICHAEL BURGETT

The world’s most extensive markets for pollination services are those for honey bee pollination in theUnited States. These markets play important roles in coordinating the behavior of migratory beekeep-ers, who both produce honey and provide substitutes for ecosystem pollination services. We analyzethe economic forces that drive migratory beekeeping and theoretically and empirically analyze thedeterminants of pollination fees in a larger and richer data set than has been studied before. Ourempirical results expand our understanding of pollination markets and market-supporting institutionsthat internalize external effects.

Key words: bees, ecosystem services, markets, pollination.

JEL codes: Q13, Q12, Q57.

Introduction

Pollination is essential for seed plant repro-duction. The transfer of pollen from stamento pistil is accomplished in different plantsby wind, water, insects, birds, and bats. Polli-nation by some insects can be economicallymanipulated by man, and the services of theseinsect pollinators have become vital agricul-tural inputs. In particular, the well- knownEuropean honey bee (Apis mellifera) is widelyused to enhance yields and promote uniformquality in nuts, fruits, and vegetables–notablyalmonds, kiwifruit, apples, cherries, pears, blue-berries, and cucumbers.1

While managed honey bees have providedpollination services in American agriculturesince colonial days, their importance has grownin recent years, and for at least two rea-sons. First, as modern agricultural produc-tion has come to employ large monocropped

Randal Rucker is a Professor in the Department of AgriculturalEconomics and Economics at Montana State University. WalterThurman is a William Neal Reynolds Distinguished Professor inthe Department of Agricultural and Resource Economics at NorthCarolina State University. Michael Burgett is an Emeritus Profes-sor of Entomology in the Department of Horticulture at OregonState University.

1 See Hoopingarner and Waller (1992) and Robinson et al.(1989). Although honey bees are the best known commercial polli-nators,other bees are used as well. Non-colonial bees are importantfor some commodities. For example, alfalfa leafcutting bees areused for alfalfa seed crops (Mayer and Johansen 2003), as well asfor hybrid canola seed crops in Canada, and bumblebees are usedfor greenhouse tomatoes (Thorp 2003).

farms, pollination by wild insects living on theperiphery of fields has become less reliable.Second, feral honey bees were decimated inthe mid- to late-1980s by Varroa and trachealmites– both of which are acarine parasites thatattack bee colonies. While both pests can becontrolled (at a cost) in managed colonies,feral honey bee populations were hit hard.Although these populations have slowly recov-ered, there are no good data available on theextent of the recovery. Experts agree, however,that across North America there is now muchless“natural”pollination to be relied upon thanin the past.2

The decline in wild pollinators has beentaken up as a cause celebre by those concernedwith wildlife preservation and biodiversity.3Concern over the loss of wild pollinators hasbeen accompanied in the past few years byconcern over the numbers of managed bees,as parasites and disease have been joined bya phenomenon dubbed Colony Collapse Dis-order. Virtually the entire bee populations oflarge numbers of overwintered colonies havedisappeared in various parts of the UnitedStates, and no single cause to explain the disap-pearances has been identified.To some,ColonyCollapse Disorder symbolizes the deleterious

2 See National Research Council (2007).3 See, for example, the web page sponsored by the Ecological

Society of America and the Union of Concerned Scientists:www.esa.org/ecoservices and the National Academy of Sciencesstudy (National Research Council 2007).

Amer. J. Agr. Econ. 94(4): 956–977; doi: 10.1093/ajae/aas031Received March 24, 2012

© The Author (2012). Published by Oxford University Press on behalf of the Agricultural and Applied EconomicsAssociation. All rights reserved. For permissions, please e-mail: [email protected]

by Murray B

elyk on September 5, 2012

http://ajae.oxfordjournals.org/D

ownloaded from

http://ajae.oxfordjournals.org/

Rucker et al. Honey Bee Pollination Markets 957

environmental consequences of industrializedagriculture, which takes widespread advantageof specialization, exchange, and economiesof scale, and that in the process may becompromising the health of managed honeybees. To other (commercial agricultural) inter-ests, Colony Collapse Disorder represents anew source of concern regarding their liveli-hoods, highlights the economic contributionof honey bees, and provides justification forgovernment-funded research.4

What is usually missing from discussions ofhoney bees and their plight is an apprecia-tion of the role of markets for pollinationservices. Pollination markets are also largelymissing from the large and growing body ofliterature on the economic value of ecosystemservices.5 Where they do exist, these marketsprovide readily available and relatively inex-pensive substitutes for the ecosystem servicesprovided by wild pollinators.

The world’s most extensive and active mar-ket for pollination services is for honey beesin the United States. A large-scale annualmigration of beekeepers moves hives fromfarm to farm, charging pollination fees as thecrops bloom. In recent years, fees paid to U.S.beekeepers were approximately $350 million.6Given the current and potential future impor-tance of pollination services, as well as thepublic’s interest in honey bees, there has beensurprisingly little economic analysis of pollina-tion markets.

Cheung (1973) was the first to take seriouslythe role of markets in allocating honey beeservices. Cheung challenged earlier theoretical

4 Arguments made in Congressional subcommittee hearings inthe past two decades assert the annual value of pollination servicesto be $9 billion (see Robinson, Nowogrodzki, and Morse 1989).Updates to that figure (see Morse and Calderone 2000) peg thevalue at $14.6 billion in 1999 dollars, a number that has been refer-enced by policy-makers. The accuracy of these estimates is an issueof active research and debate (see for example, Muth and Thur-man 1995, Gallai et al. 2009, and National Research Council 2007).Regarding the magnitude of the economic effects of CCD, littleresearch has been conducted. In 2007,U.S. Secretary ofAgricultureJohanns warned that, “if left unchecked, CCD has the potential tocause a $15 billion direct loss of crop production and $75 billionin indirect losses,” (USDA 2007). Rucker, Thurman, and Burgett(2012), however, estimate the economic impacts of CCD to be rel-atively small, a result attributed by Rucker and Thurman (2011)to “the resilience of honey bees and to the business acumen andperseverance of commercial beekeepers.”

5 For a review of recent literature on the economics of ecosystemservices, see Swinton et al. (2007).

6 The $350 million estimate is based on assigning an annual pol-lination rental income figure of $145 per colony to the number ofbee colonies reported by the U.S. Department of Agriculture in2009. For comparison, the USDA reports the 2009 value of honeyproduced in the United States (the other output of managed honeybees) as $208 million.

literature on externalities (Meade 1952, Bator1958, and precursors by Pigou 1912 and1920) that used honey bees and apples as anillustration of reciprocal externality. The ear-lier literature argued that apple trees provideuncompensated benefits to beekeepers (nec-tar for bees and honey production), and beesprovide uncompensated benefits to orchardowners (pollen transfer for apple production).From their assertion that apple farmers andbeekeepers do not transact,the writers inferredan underprovision of both apples and bees.Cheung responded with an examination of1971 data from a small number of beekeep-ers, arguing that (1) the existence of marketsfor pollination services rebutted the presump-tion of externality, and (2) variation in theobserved fees was consistent with differencesacross crops in the relative values of pollinationand honey.7

Pollination markets have developed consid-erably in the 40 years since Cheung wrote,and the policy issues surrounding pollina-tion agriculture have changed considerablyas well. In the current paper we theoreti-cally and empirically analyze the determinantsof pollination fees with a much larger andricher data set than has been studied before.Our data come from an annual survey ofPacific Northwest (PNW) beekeepers that hasbeen conducted by Michael Burgett at OregonState University since the late 1980s. In addi-tion to reinforcing Cheung’s refutation of theexternality myth, our empirical results expandour understanding of pollination markets andmarket-supporting institutions that internalizeexternal effects.

7 To our knowledge, other than Cheung’s 1973 paper, refereedpublications on the workings of pollination markets are the follow-ing: Johnson (1973), Olmstead and Wooten (1987), Muth, Rucker,Thurman, and Chuang (2003), Ward, Whyte, and James (2010). Ofthese, only Cheung, Muth et al. and Ward et al. statistically examinepollination market data. The differences between the current workand the three empirical studies–and the advances we offer beyondthem–are as follows. Cheung’s 1973 study analyzed detailed dataon one year of activity from eight beekeepers, all of whom livedin Washington. All of his statistical tests are relatively simple, andseveral of them are calculated using data obtained from only fourbeekeepers. An important finding was that pollination fees variedinversely with the honey value of the crop. The focus in Muth et al.is the honey support program. For pollination, they study an ear-lier data set (through 1995) in a simpler econometric setting thatdoes not take into account several potentially important covari-ates. Ward et al.’s 2010 study focuses only on almond and cherrypollination fees, and fails to take into account changes in the pur-chasing power of the U.S. dollar over a twenty-year sample period.Our analysis examines in a panel data setting a substantially moreextensive set of pollination market data than has heretofore beenstudied.

by Murray B



ownloaded from


958 July 2012 Amer. J. Agr. Econ.

Pollination and Markets

A typical large-scale North American commer-cial pollinator drives a tractor-trailer combina-tion that carries 400 to 500 bee hives, each ofwhich contains a single colony with a queen andbetween 15,000 and 30,000 workers.8 Trans-portation of the bees is facilitated by travelingat night with nets covering the hives–bees flyout of their hives only during the day. Oncethe truck arrives at a field or orchard for pol-lination, forklifts are used to move the hivesto strategic points to spread bees throughoutthe flowering area. Bees typically stay close tohome when placed in a pollen and nectar-richflowering field.They will,however,fly consider-able distances when pollen and nectar sourcesare more difficult to find.9 As bees forage acrossflowers, they pick up pollen (which contain themale gamete, or sperm) on their bodies andtransfer it to the pistils (the female reproduc-tive organs) of other flowers. In the case ofnuts and tree fruit, an important role playedby bees is cross-pollination: the transfer ofpollen between trees of one variety and thoseof another variety, strategically planted in adja-cent rows.10 The hybrid vigor that results frominter-variety pollen transfer promotes fruit setsand ultimately fruit quality and uniformity.

Bees are typically moved into an orchard orfield for just the flowering period. This periodis roughly three weeks for almonds and mosttree crops, but can vary with the weather, withhigher temperatures condensing the bloomperiod and lower temperatures extending it.The placement period for some crops is longer–cranberries can require four to five weeks.After pollination of a particular orchard, field,or bog, many beekeepers move their coloniesby truck to the next pollinating site, usuallya later blooming crop or possibly the samecrop farther north or at a higher altitude. Atsome point during the summer most beekeep-ers move their colonies to a location where the

8 Historical and contemporary accounts of beekeeping can befound in Nordhaus (2011), Horn (2005), Crane (1999), Mairson(1993), and Pellett (1938).

9 Seeley (1995, pp. 46–50) discusses the results of studies of theforaging range of honey bee colonies. He states that,“. . . the mediandistance was 1.6 km, the mean distance was 2.2 km, and the maxi-mum distance was 10.9 km. Perhaps the most important propertyof this distribution is the location of the 95thpercentile, which fallsat 6.0 km. This indicates that a circle large enough to enclose 95percent of the colony’s forage sites would have a radius of 6 km,hence an area greater than 100 km2.”

10 See, for example, Degrandi-Hoffman et al. (1992) on bee for-aging behavior in almond orchards and its implications for optimalplanting of trees and varieties.

focus of the bees’ efforts is commercial honeyproduction.

That pollination has been used as an exam-ple of a positive externality is understand-able. The physical facts of crop pollinationimply that transactions are difficult to mon-itor and agreements costly to enforce. Cropsrequire pollination for only a brief period eachyear, and crops at different latitudes and alti-tudes require pollination at different times.Thus, there are large economies of scale avail-able to mobile beekeepers, who use the samebees to pollinate several crops in a season.But mobility also makes market transactionscostly. Delivering pollination services must becoordinated across multiple crops during theirblooming seasons, and in the face of substan-tial uncertainty regarding the precise timingof the blooms. Further, markets must coordi-nate the joint production of pollination andhoney against a backdrop of continually evolv-ing scientific views of the efficacy of honey beepollination.

Pollination markets today consist ofcontracts between farmers and migratorybeekeepers. There are several large-scalemigration routes traveled by these beekeepersand their bees, including the route traveled byWashington and Oregon beekeepers, who arethe focus of our empirical analysis. Their polli-nation season begins each year in February inthe almond groves of California.

Following almond pollination, California-based bee colonies are often put into nearbycitrus orchards. Although there are no polli-nation benefits to citrus farmers and no feesreceived by beekeepers, the nectar is plentifuland valuable honey is produced. The migra-tory PNW beekeepers, on the other hand,typically move their colonies north to theirhome bases in Oregon and Washington. Fromthere, they distribute their colonies amonglocal apple, pear, and cherry orchards. Follow-ing that, the majority of the beekeepers withhome bases west of the Cascade Mountainsrent their colonies out to pollinate additionalcrops–typically soft fruits (strawberries, rasp-berries, and blueberries) in May and June,followed by seed crops (especially vegeta-bles such as onions and carrots), then cucum-bers, pumpkins, squash and some legume seeds(e.g. clovers) and occasionally alfalfa seed.11

11 It is noteworthy that most alfalfa seed in the Pacific Northwestis pollinated by two species of“managed wild bees,” the alfalfa leaf-cutting bee and the alkali bee. Further, survey data from the PNWsurvey suggest that PNW beekeepers engage in substantially more

by Murray B



ownloaded from



Not only does the timing of colony placementvary across crops, but the pollination fees col-lected by the beekeepers also differ. Thesedifferences are a focus of our theoretical modeland empirical analysis below.

The east side of the Cascades is much richerin honey sources than the west side. As aresult, beekeepers whose home bases are onthe east side of the Cascades typically polli-nate California almonds and then tree fruitsin eastern Washington and Oregon. After that,they spend the rest of the season using theirbee colonies to produce honey.12 Many of thecolonies remain in eastern Oregon and Wash-ington, with a minority of them being movedto summer honey locations in Montana andother Northern Plains states. There, they joinlarge numbers of other U.S. beekeepers, whoalso find summer ranges for their colonies inthe region. For the rest of the summer, thehives remain at these sites, and the bees visitsunflowers, clover, basswood trees, and othernectar sources, producing honey for consump-tion by the hive and extraction for sale by thebeekeeper.

In the fall, many U.S. beekeepers move theirbees again, this time to winter in the south or inthe Central Valley of California. Some Wash-ington and Oregon beekeepers move their beesto California in December and find locationsto hold their bees until the following year’salmond bloom. The majority of PNW bee-keepers wait until January or early February,however, to move their bees to California,where they often are placed in temporary hold-ing yards (apiaries) until it is time to move intoalmond orchards for pollination.

Parallel migratory routes move up theAtlantic coast, from fruit and vegetable cropsin Florida to blueberry bushes in Maine.Although surveys of eastern beekeepers havebeen initiated recently, there is no surveywhose longevity comes close to approachingthat of the PNW surveys we analyze below.

The markets that connect beekeepers withcontracting farmers range across the UnitedStates, addressing complex problems of infor-mation gathering and processing. To gain a

pollination activities than their California counterparts. On aver-age, PNW respondents reported that each colony pollinates almost1.5 additional crops after almonds. This number masks to somedegree the extent of beekeeper (as distinct from bee) activities.For example, on average, individual beekeepers reported takingbees to pollinate 5.5 different crops in 6.8 counties each year. Seetable 1 in Burgett et al. (2010).

12 West-side beekeepers produce honey, but the quantity is smallrelative to east-side production. The primary source of west-sideproduction is blackberries.

quantitative understanding of these markets,we develop a formal model of price relation-ships in pollination markets,and derive and testits predictions.

A Competitive Model of Pollination Fees

Honey and fruit (representing here also nutsand seeds) are the joint outputs of a pro-duction process that employs both land andbees. The typical organization of productioninvolves farmers who hire beekeepers to pro-vide bee colonies to pollinate their crops. Farm-ers receive the output of fruit while beekeepersreceive the honey (and the nutritional value ofthe nectar for their bees). In equilibrium, thereis a side payment from farmers to beekeepers.13

To determine the equilibrium pollinationpayment, we first consider an equilibrium thatwould result in a fictional market where multi-output firms hire the services of land and beesin order to produce fruit and honey, whichthe firms sell in competitive markets. We thenderive the implications for an equilibrium pay-ment from crop growers to beekeepers in thereal world, where farmers retain ownership ofthe fruit and pay pollination fees to beekeepers,who retain ownership of the honey.

Optimal Stocking of Bees on Land

The two-output production function isdescribed by a pair of constant-returns-to-scaleproduction functions:

(1) H = GH(A, B) and F = GF(A, B)

where H and F are the quantities of honey andfruit produced, and A and B are the numbersof acres and bee colonies employed. Becausethe honey and fruit production functions areconstant returns to scale, they can be written inper-acre terms:

GH(A, B) = A · gH(b) and

GF(A, B) = A · gF(b),

where b ≡ B/A is the stocking density, which isthe sole determinant of the per-acre outputs ofboth honey and fruit.

13 In the model we develop below, situations can arise whereside payments flow from beekeepers to farmers. From our study ofactual market arrangements, it is clear that such situations are rarewhen bees provide pollination services.

by Murray B



ownloaded from



An owner of a fixed quantity of acres solvesthe per-acre profit maximization problem:

max{b}

π(b) = PH · gH(b)(2)

+ PF · gF(b) − w · b,

where PH and PF are the market prices ofhoney and fruit, and w is the market wageof bees. The bee wage is the market price ofpollination services, a parameter from the per-spective of the individual beekeeper. Determi-nation of the equilibrium bee wage is discussedbelow.

The first and second-order conditions for theprofit-maximization problem are:

FOC :dπ

db= PH · g′

H(b)(3)

+ PF · g′F(b) − w = 0

SOC :d2π

db2= PH · g′′

H(b)(4)

+ PF · g′′F(b) < 0.

Note that (4) is not a global restriction on theshapes of the individual production functions,gH and gF. At the optimum, a linear combina-tion of the two production functions must beconcave.

The first-order condition in (3) can be inter-preted as setting the total value marginal prod-uct of bees (TVMPB) equal to the bee wage,where TVMPB is the sum of the VMPs forhoney and fruit. Solving (3) for b as a functionof w, PF, and PH, and substituting that function[b∗ = b∗(w, PF, PH)] back into (3) yields

TVMPB ≡ PH · g′H(b∗) + PF · g′

F(b∗)(5)

≡VMPHB (b∗) +VMPF

B(b∗) = w.

The left-hand panel of figure 1 displays theoptimal choice.

Log-differentiating (5) results in

[ϕηH + (1 − ϕ)ηF] d ln b∗ + ϕ d ln PH(6)

+ (1 − ϕ) d ln PF = d ln w,

where ηH ≡ d ln g′H

d ln b , ηF ≡ d ln g′F

d ln b , and

ϕ ≡ PHg′H

PHg′H+PFg′

F. Note that the parameter ϕ

is the honey share of the total VMP of beesand is represented as the ratio cd to ce infigure 1. Solving for the comparative staticresponses of b∗ to changes in w, PH and PF

gives

d ln b∗ = 1ϕηH + (1 − ϕ)ηF

(7)

× [d ln w − ϕd ln PH

− (1 − ϕ) d ln PF].Note that the own-price elasticity of the opti-mal stocking density is

ηb ≡ ∂ ln b∗

∂ ln w(8)

= 1ϕηH + (1 − ϕ)ηF

≤ 0 by the SOC.

Substituting the expression for ηb into (7)results in

d ln b∗ = ηb d ln w − ϕηb d ln PH(9)

− (1 − ϕ)ηb d ln PF.

From (9) one can see that the comparativestatic effects of PH and PF on b∗ are of thesame signs as ϕ and (1 − ϕ). Both ϕ and(1 − ϕ) are positive if, at the optimum, themarginal products of bees in fruit and honeyare positive. If such is the case, then beestocking density increases from increases ineither output price.14

Equilibrium in the Market for PollinationServices

Equilibrium is characterized by an aggregatedemand, which is the sum of the optimalstocking density function (b∗) across A(assumed identical) acres on which bees areemployed as described above,and an aggregatesupply of bee services for pollination.

Supply depends upon the bee wage (w), var-ious factors that affect the costs of beekeeping(k), and the price of honey (PH). The aggregateequilibrium appears as:

(10) A∗ · b∗(w, PF, PH) = QS(w, PH, k),

where A∗ is the aggregate equilibrium num-ber of acres pollinated. While we analyze the

14 Note that one of ϕ and (1 − ϕ) (but not both) can be nega-tive at the optimum, implying that the marginal product of beesin, say, honey could be negative. Expression (9) then implies thatan increase in the price of honey would increase the opportunitycost of a negative marginal product of bees in honey, and resultin a decrease in stocking density (and an increase in the marginalproduct of bees in honey).

by Murray B



ownloaded from



w

B

B

B

$/b $/b

S

* eB

d TVMPB

VMPH

VMPF

c

A* • TVMP = D

*

b* b (stocking density or colonies/acre)

B

= b* • A*

B (1000s of colonies)

Individual Beekeeper (A=1) Market for Beekeeping Services

Figure 1. The optimal bee stocking rate

effects of changes in A∗, we take those changesto be exogenous with respect to changes inmarket prices in the length of run considered,which is most reasonable for long-lived peren-nials such as almonds and tree fruit. On theright-hand side of (10), QS(.) represents theaggregate pollination supply function, which isassumed to be increasing in w and decreasing ink. The equilibrium in equation (10) is depictedin the right-hand panel of figure 1.

Log differentiate (10) to determine the equi-librium response of w to exogenous changes inA, PF, PH, and k:

d ln A + d ln b∗ = ε d ln w + β d ln PH(11)

+ γ d ln k,

where ε ≡ ∂ ln Qs

∂ ln w> 0, β ≡ ∂ ln Qs

∂ ln PH

>

<0 and

γ ≡ ∂ ln Qs

∂ ln k< 0.

Substitute into (11) from (9) and collectterms to obtain

d ln w = 1ε − ηb

d ln A(12)

− ϕηb + β

ε − ηbd ln PH

− (1 − ϕ)ηb

ε − ηbd ln PF

− γ

ε − ηbd ln k.

Equation (12) gives the equilibriumresponse in the bee wage to exogenouschanges in A, PH, PF, and k. Substitute from(12) back into (9) to derive the equilibriumchange in b∗ induced directly by exogenouschanges in PH and PF, and indirectly byinduced changes in equilibrium w:

d ln b∗ = ηb

ε − ηbd ln A(13)

− ηb(ϕε + β)

ε − ηbd ln PH

− ηb(1 − ϕ)ε

ε − ηbd ln PF

− ηbγ

ε − ηbd ln k.

Equations (12) and (13) comprise a reducedform for equilibrium changes in w and b∗ thatresult from exogenous changes in A, PH, PF,and k.The coefficients on d ln A,d ln PH,d ln PF,and d ln k in (12) and (13) are the comparativestatic effects of changes in those variables onthe (unobservable) bee wage (equation 12) andstocking density (equation 13). Their signs canbe summarized as follows.

Each of the terms in (13) involves a multi-plicative coefficient of ηb, the field-level own-price elasticity of bee stocking density. By thesecond-order conditionηb ≤ 0. Ifηb < 0,we cansay the following. First, referring to figure 1, anincrease in A shifts aggregate demand for beeservices to the right, increasing w, and inducingan upward movement along the TVMPB curve,

by Murray B



ownloaded from



which decreases b∗. Second, an increase in PFshifts to the right the TVMPB curve, inducingan increase in equilibrium b∗, a rightward shiftin the aggregate demand for bee services, andan increase in w assuming that (1 − ϕ) > 0.This latter condition is equivalent to the VMPof bees in fruit production being positive atthe equilibrium b∗. We consider 1 − ϕ > 0to be the normal case: at the margin, higherstocking densities of bees produce more fruit,but may result in less honey due to bees’ ownconsumption and competition among bees fornectar.

Third, an increase in PH shifts the TVMPBcurve to the right, thereby shifting the aggre-gate demand for bee services to the right. Anincrease in PH also shifts the aggregate sup-ply of bee services, either leftward or rightwarddepending on whether or not pollination andhoney production are net substitute or comple-mentary outputs. Fourth, an increase in k–thecosts of beekeeping–shifts the supply curve ofbee services upwards,resulting in an increase inthe bee wage (w),and a decrease in the stockingdensity (b∗).

Comparative Statics of Pollination Fees

Now consider the real-world situation wherefarmers own or rent the orchards and rent theservices of beekeepers. Beekeepers are paidpartly in kind by the honey, which they keep,and partly by a pollination fee paid by farmers.Farmers retain their claim to the fruit producedby the joint efforts of bees and land.

The fact that the standard contractual agree-ment for pollination services assigns propertyrights to the beekeeper to all honey produced isa result of measurement and monitoring costs,and an optimal allocation of price and out-put risk. A contract in which the farmer wasdue the honey would be more difficult to mon-itor than one in which the farmer was dueonly the pollination services of a fixed num-ber of easily observed bee colonies.15 Such acontract would require the extraction, or atleast the accurate estimation,of honey reservesafter each crop was pollinated. Given that

15 This is not to suggest that monitoring pollination services iswithout cost.The issue of hive quality is of great concern to farmers.The strength of hives is assessed partly by the inspection of hives byfarmers, partly by third party inspectors and, in some cases, by theassurances of bee brokers who guarantee colony strength in rentedcolonies. Reputation effects provide an incentive for beekeepersto provide full strength hives, especially in instances where hivesare placed with a particular farmer year after year (see Klein andLeffler 1981). Some almond growers specify a sliding pollination feebased on hive strength, which is assessed by third party inspectors.

extraction equipment is typically located at abeekeeper’s home base, extraction after eachpollination set would be costly, and it is dif-ficult to imagine how a farmer might validateestimates of honey reserves accumulated whilecolonies were located on their land. More-over, as argued by Barzel (1997, chapter 1),production and price risk will tend to beborne by the transacting party who has themost influence over those sources of incomevariability.

Thus, the allocation of honey price and yieldrisk to beekeepers,and fruit price and yield riskto orchard owners, is in accord with their rel-ative abilities to maximize economic value inresponse to those sources of income variability.See Knoeber and Thurman (1994) regardinga similar contractual allocation of price andproduction risk to enhance value in contractbroiler production.

The bee wage, w, is comprised (in differentproportions across crops) of a per colony polli-nation fee (PP) and an in-kind honey payment,or w = PP + gHPH/b.16

Equivalently, the equilibrium pollinationfee is

(14) PP = w − gH · PH

b.

Log differentiate (14) to obtain

d ln PP = 11 − α

d ln w(15)

− α

1 − α[d ln gH + d ln PH

− d ln b],

where α = PH·gHw·b , the honey share of compen-

sation. Denoting the elasticity of honey out-

put with respect to b as ηHb = d ln gH(b)

d ln b, (15)

16 Figure 2 shows the equilibrium payment to beekeepers peracre of land, PP b, in two situations. The first panel shows the situa-tion where the per-acre pollination payment (which will equal thedifference between the product of the bee wage and the numberof colonies, and the total value of the honey produced at the opti-mal stocking density) is positive and equal to area C. The secondpanel shows the uncommon case in which the equilibrium paymentis negative (and equal to area D-C), that is, beekeepers pay farmersfor the privilege of placing their bees on the cultivated land. Thepossibility of zero or negative pollination fees can most easily beimagined for a crop that yields substantial marketable honey out-put,or little or no marginal fruit product at the equilibrium stockingrate, such as oranges in California and, in Oregon’s Willamette Val-ley, crimson clover and hairy vetch (both of which are legumesgrown for seed).

by Murray B



ownloaded from



Figure 2. Pollination fees

reduces to

d ln PP = 11 − α

d ln w(16)

− α

1 − α

[d ln PH

+ (ηHb − 1)d ln b

].

To determine the effect on PP from exogenouschanges in A, PH, PF, and k, substitute for d lnw from (12), and for d ln b from (13), into (16).This results in

d ln PP(17)

= 1(1 − α)

[1 − αηb(ηH

b − 1)]

(ε − ηb)d ln A

− γ

(1 − α)

[1 − αηb(ηH

b − 1)]

(ε − ηb)d ln k

+ (1 − ϕ)ηb

(1 − α)

[αε(ηH

b − 1) − 1]

(ε − ηb)

× d ln PF

+ 1(1 − α)

×

⎧⎪⎪⎨⎪⎪⎩

ηbα(ηHb − 1)(ϕε + β)

−(ϕηb + β)

(ε − ηb)− α

⎫⎪⎪⎬⎪⎪⎭

× d ln PH.

The comparative static effects from (17) can-not be signed in general. However, if stockingdensities are constant with respect to changes

in w (ηb = 0) then all but one of the effectsin (17) can be signed. In this important and–as argued below–empirically relevant case, wehave the following:

If ηb = 0,∂ ln PP

∂ ln A= ε

(1 − α)> 0.(18)


∂ ln k= γ

(1 − α)ε> 0.(19)


∂ ln PF= 0.(20)

The sign of the comparative static effect of achange in PH remains ambiguous, but reducesto:

(21) If ηb = 0,∂ ln PP

∂ ln PH= 1

(1 − α)

(−β

ε− α

).

The effect of an increase in PH can be seento be positive if −β > αε: the substitutioneffect away from pollination supply due toa honey price increase must be larger thanthe share-weighted own-price elasticity ofpollination services supply.

Testing the Model of Pollination Fees

In this section we test the implications ofthe economic model of equilibrium pollinationfees developed above.We relate time series andcross-sectional variation in pollination fees tovariations in the price of honey, the price of thepollinated crop, factors influencing the costs ofbeekeeping, and observable characteristics ofthe crop, such as its honey yield.

by Murray B



ownloaded from



For data on pollination fees, we exploit adetailed annual survey of Oregon and Wash-ington beekeepers that has been conductedannually by one of the authors over a 23-yearperiod (see Burgett,various years,also Burgett,Rucker, and Thurman 2004.) The data set wehave constructed includes information aggre-gated from the survey respondents on averageannual pollination fees by crop from 1987–2009. We augment the survey data with annualdata from other sources on Oregon crop andhoney prices,and on factors that affect the costsof beekeeping.

Empirical Predictions

Equation (14) relates the equilibrium pollina-tion fee to the unobserved equilibrium beewage and the value of the in-kind payment ofhoney. The equation’s most direct implicationconcerns variation in pollination fees acrosscrops.

Prediction 1: Pollination fees will be lower forcrops that yield more honey.

That this is true in theory can be seen fromthe fact that the bee wage, w (the sum ofin-kind and pecuniary payments), is constantacross crops. Therefore, from equation (14), anincrease in the per colony honey yield,gH(b)/b,reduces the pollination fee, Pp.

Our next three empirical predictions followfrom the comparative static effects in (17) cou-pled with the auxiliary empirical assertion thatstocking densities for a given crop are fixed as amatter of agronomic practice, and do not varyyear-by-year with crop prices, beekeeper costs,or the price of honey.We support this empiricalclaim below.

Prediction 2: If stocking densities are fixed,pollination fees will vary directly overtime with beekeeper costs. (Follows fromequation 19.)Prediction 3: If stocking densities are fixed,pollination fees will rise with increases in theacreage of pollinated crops. (Follows fromequation 18.)Prediction 4: If stocking densities arefixed, pollination fees will not change inresponse to crop price changes. (Follows fromequation 20.)

Another potentially important determinantof pollination fees is the price of honey, theeffect of which is theoretically ambiguous.Even if stocking densities are fixed, an increasein PH will increase PP if the elasticity of the

bee wage with respect to the price of honeyexceeds the in-kind share of beekeeper com-pensation. This follows from equation 21.Intuitively, one reason why the effect of PHon PP might be large is that an increase in PHcould induce a large shift in bees away frompollinated crops and toward nectar-and-honeyemployments, such as uncultivated areas andthe sunflower fields of North Dakota. Thisincrease in the opportunity cost of placing beeson crops for pollination could raise w to suchan extent that an increase in the price of honeyresults in an overall increase in pollinationfees, despite the increased value to beekeepersof the in-kind payment.

The Importance of Almonds

Any analysis of contemporary U.S. pollinationmarkets is incomplete without considering therole played by California almond pollinationduring February and March. Almonds as cur-rently grown are highly dependent on honeybee pollination; figure 3 provides insights intohow almond pollination has expanded overtime. The two lines show almond pollinationdemand relative to the number of honey beecolonies available for pollination in (1) Califor-nia, and (2) California plus the PNW (Oregonand Washington).

The quantity of almond pollinationdemanded in a given year is estimated asthe number of bearing acres of almonds inCalifornia multiplied by the number of hivesper acre used for pollinating almonds–forthe purposes of figure 3, we use the averagedensity of 2.1 hives per acre reported by PNWbeekeepers.

The first series, titled “90% of Californiacolonies,” provides an indication of when thequantity demanded for almond pollination ser-vices might have exceeded the quantity sup-plied by California beekeepers. If one esti-mates that 90% of the colonies in Califor-nia are owned by beekeepers willing to pro-vide commercial almond pollination services,then the approximate point at which out-of-state pollination services would have beenimported was 1973.17 Two obvious sources of

17 Data from the 2002 Agricultural Census indicate that about88% of the colonies in the United States are on commercialoperations with 300 or more colonies, and another 10% are onsemi-commercial operations with 25 to 299 colonies; see table 1in Daberkow et al. (2009). Assuming that virtually all of thebeekeepers with 300 or more colonies pollinate some crops and thata fraction of the smaller operations are also involved in pollination,our use of the 90% estimate in the text seems reasonable.

by Murray B



ownloaded from



Tho

usan

ds o

f col

onie

s

1500

Bees required for almonds

1000

90% of CA, OR, & WA colonies

50090% of CA colonies

1973

1977

01940 1950 1960 1970 1980 1990 2000 2010

Figure 3. Almond bee requirements rise beyond regional colony numbers

outside pollination services are Washingtonand Oregon beekeepers. If one again assumesthat about 90% of the colonies in Califor-nia and the PNW are available for commer-cial pollination services, the second series infigure 3 suggests that the quantity demandedfor almond pollination services exceeded theregional quantity supplied in the mid-to-late1970s. Not shown in the figure is the substan-tial rise in the demand for almond pollina-tion services relative to total U.S. honey beecolony numbers. Whereas almond producersemployed less than 5% of U.S. colonies priorto the mid-1960s, they employed about 15% bythe late 1970s, and currently employ 60%.

Today, many beekeepers are paid substantialpremia to bring their colonies to California andplace their bees in almond orchards during thebloom. In recent years,almond pollination feesreported by PNW beekeepers have increaseddramatically–from an average of about $66 percolony in 2004 to almost $157 in 2006, in 2009dollars.18

One response to increased almond pollina-tion fees has been the attraction of beekeepers

18 For California beekeepers, reported almond pollination feesincreased from about $73 in 2004 to $166 in 2006. See Burgett et al.(2010), tables 3 and 4.

from greater distances, some from as far awayas the East Coast. The fee increases havebeen attributed to increases in almond acresand expectations of further future increasesin almond acres. Some sources have suggestedthat the onset of Colony Collapse Disorder hasresulted in increased pollination fees.19 In ourempirical analysis, we examine the impact ofthese and other factors.

In light of the importance of almonds, it isuseful to explicitly link our theoretical modelwith actual pollination markets. Three spe-cial cases in the model describe three distinctphases of the pollination calendar: almonds,post-almond pollination activity, and post-pollination honey production. While pollinat-ing almonds, bee colonies run an energy deficit,and beekeepers typically feed their coloniessupplemental calories–usually a sucrose solu-tion. Bees produce no surplus honey while pol-linating almonds. Further, honey from almondnectar has a bitter taste and so is undesirableif produced. In this situation, the theoreticalmodel of the previous section applies, but thevalue, or effective price, of honey is zero. While

19 See Sumner and Borriss (2006), who cite almond acres asimportant drivers. Ward et al. (2010) and Carman (2011) add CCDto the list of factors they believe to have increased almond fees.

by Murray B



ownloaded from



Table 1. Bee Stocking Density Summary Statistics by Crop

(4) (5) (6)(2) (3) Mean Minimum Maximum

Years of Average Annual Annual Annual(1) Density Observations Average Average Average

Crop Data per Year Density Density Density

Almonds 15 8.53 2.08 1.51 2.61Apples 17 8.94 1.34 0.97 1.87Blueberries 17 6.82 2.07 1.16 3.00Cherries 17 12.12 1.54 0.89 2.31Red and White

Clover Seed17 4.76 1.30 0.90 2.31

Cranberries 7 1.57 1.59 1.24 2.20Crimson

Clover Seed16 3.81 1.07 0.74 1.87

Cucumbers 16 2.25 0.98 0.20 1.60Pears 17 8.76 1.40 0.89 2.29Radish Seed 17 3.65 1.60 1.00 2.88Squash and

Pumpkins17 5.23 0.86 0.49 1.33

Vetch Seed 7 2.29 0.97 0.60 1.30

pollinating almonds, beekeepers are paid byfee the value of their marginal product inproducing almonds, which is unadjusted forhoney value.

After the almonds bloom, beekeepers movetheir colonies to a variety of locations andcrops. During this post-almond spring period,the full model is descriptive, with pollinationfees taking into account both pollination valueand the value of any honey produced. Finally,in late spring, commercial crops are throughblooming and the value of bees’ marginal prod-uct in the production of crops is zero; all com-mercial bee activity is directed toward honeyproduction (and the buildup and maintenanceof colony health).

Our survey data on pollination fees referto transactions during the first two phases ofthe seasonal cycle: almonds and post-almondcrops. During the third phase, pollination feesare zero or negative–they are considered ascosts by beekeepers, not as sources of rev-enue. The PNW surveys include no observa-tions from this period. In terms of our model,this portion of the beekeeping season cor-responds to the situation where the valueof the marginal product of bees in the pro-duction of fruit is zero (imagine any of thegraphs in figures 1, 2, or 4 with the entireVMP of bees coming from the production ofhoney). While the bee wage in our modelis constant across crops within any of thethree sub-seasons, it may differ across thesub-seasons.

The Relative Constancy of Bee StockingDensities

One potential influence in the theoreticalmodel developed above is the responsivenessof stocking density (b) to changes in exogenousfactors. Here we examine the issue empirically.

Beekeeper respondents to the annual PNWsurveys were asked to report, for each crop,the number of colonies rented and the num-ber of acres pollinated. Although responses tothese questions are missing on many surveys,we have sufficient colony density data to esti-mate an empirical model that corresponds toequation (13): the comparative static result forhow stocking density responds to changes incrop acreage, the (expected) prices of honeyand crops, and economically significant costfactors–including the appearance of the Var-roa mite in the PNW and the onset of ColonyCollapse Disorder.

To empirically estimate (13) we constructeda data set of annual average stocking densitiesby crop for each year over 1989–2009. Table 1presents summary information. For ten of thetwelve crops,we have 15 to 17 observations. Forcranberries and vetch seed, we are only ableto determine average densities for seven years.With regard to the number of useable ques-tionnaire responses per year, almonds, apples,cherries and pears all have averages of morethan eight annual observations. Five of theother crops have averages of at least 3.5. Densi-ties for cranberries, cucumbers, and vetch seed

by Murray B



ownloaded from



are the least frequently reported,with the aver-age number of responses for all three of thesebeing less than 2.5.

Column (4) of table 1 shows notable vari-ability in these densities across crops, withthe means for almonds and blueberries beinggreater than two colonies per acre and themean densities for vetch, cucumbers, andsquash and pumpkins being less than one.

Further, columns (5) and (6) imply year-to-year variation in stocking densities for each ofthe crops.

The data summarized in table 1 comprise anunbalanced panel of stocking densities by crop(180 observations in total), which we use toestimate the following model:

Densityit = b0i + b1Crop Priceit(22)

+ b2Honey Pricet

+ b3Almond Acrest

+ b4Varroat + b5CCDt

+ eit,

where for crop i in year t, Densityit is the aver-age annual hive stocking density; Crop Priceitand Honey Pricet are measures of expectedcrop and honey prices constructed from AR(2)time series models; Almond Acrest measuresbearing almond acres in California;20 Varroat isa dichotomous variable equal to one for obser-vations after 1990 (when Varroa first appearedin the PNW), and zero otherwise; and CCDt isa dichotomous variable equal to one for obser-vations after 2006 (when CCD first appeared),and zero otherwise. In addition to the linearspecification shown, we also examine log-log,semi-log, and inverse semi-log versions of (21).All specifications include crop fixed effects,denoted as b0i.

Ordinary least squares estimates of (22) arereported as specification (1) in table 2.The onlystatistically significant coefficients are the cropfixed effects, which are highly significant. Weconclude that crops vary significantly in theirtypical stocking densities. But despite the evi-dent variation over time seen in table 1, thevariation is not explained by time series varia-tion in almond acres, crop price, honey price,and cost-of-beekeeping factors. This conclu-sion is not affected by including a linear timetrend (results not shown here).

20 We include almond acres as an explanatory variable for stock-ing densities for all crops due to the aforementioned importance ofalmonds in pollination markets.

A natural interpretation of this non-significance is that ηb—the factor commonto all the comparative static effects—is zero.Unfortunately, the parameter ηb is not identi-fied in the specification of equation (22); butone can come close. The ratio ηb/ε is iden-tified from the elasticity of stocking densitywith respect to acres–see equation (13)–usingvariation in almond acres as the empiricallyimportant variation in acres.The parameter ε isthe elasticity of the pollination services supplycurve. Using equation (13), and the estimate ofthe elasticity of stocking density with respectto acres from specification (2), one can use thedelta method to construct the following 99%confidence interval for ηb/ε: (−0.079, 0.535).If ηb is non-zero, theory tells us it is neg-ative. Further, the most negative value forηb/ε supported by the data is −0.079. Thus,if ηb is negative, the confidence intervaltells us that it is smaller in magnitude thanone-tenth the supply elasticity of bee ser-vices. By this comparison, ηb is economicallyinsignificant.

To further examine the determinants ofhive density, we supplement the regressions intable 2 with separate OLS time series regres-sions on each of the twelve crops in our dataset. We estimate these regressions for each ofthe four specifications reported in table 2, andthe results of these by-crop regressions are con-sistent with those for the full stocking data set.For each of the specifications, estimation of thetwelve by-crop regressions yields 58 estimatedcoefficients.21 In none of the four specificationsare more than four of the 58 (6.9%) estimatedcoefficients statistically significant at the 5%level. Moreover, for the four specifications intable 2, the estimated coefficients are jointlysignificant in,at most, two of the twelve by-cropregressions.22

The results of this empirical exercise are easyto summarize by making two points. First, theanalysis of the full data set indicates that thereare significant differences in stocking densities

21 This number excludes the estimated intercepts, which are notof interest. Ten of the twelve regressions include five explana-tory variables. The almond regression excludes the Varroa variablebecause our almond data start in 1991, so all values of this vari-able are one for that crop. The vetch data set ends before CCDappeared, so that regression excludes the CCD variable.

22 A more detailed presentation of the by-crop regression resultsis available upon request. A number of other specifications werealso estimated. In all cases, the number of significant coefficientsis small, and the estimated coefficients are not jointly significant inthe vast majority of the individual by-crop regressions.

by Murray B



ownloaded from



Table 2. Determinants of Bee Stocking Densities

(4)(1) (2) (3) Inverse

Linear Log-lineara Semi-log Semi-loga

Dependent Variable Density ln(Density) ln(Density) Density

Independent VariableCrop Price 0.024 0.142 0.051 0.07

(0.203) (0.118) (0.155) (0.155)Honey Price −0.002 −0.267 0.012 −0.037

(0.151) (0.144) (0.116) (0.189)Almond Acres 1.14 0.295 0.562 0.596

(0.515) (0.200) (0.393) (0.262)Varroa −0.019 −0.011 −0.026 −0.013

(0.121) (0.095) (0.092) (0.124)CCD −0.091 −0.054 −0.058 −0.074

(0.112) (0.079) (0.086) (0.104)Crop Fixed Effects Included Yes∗∗∗ Yes∗∗∗ Yes∗∗∗ Yes∗∗∗Adjusted R2 0.548 0.522 0.519 0.549

Number of observations = 180. Standard errors appear in parentheses. Asterisks ∗ , ∗∗ , and ∗∗∗ indicate one-tailed significance at the 0.10, 0.05, and 0.01 levels,respectively.aDenotes that the natural logarithms of Crop Price, Honey Price, and Almond Acres (i.e., all non-binary variables) are used as regressors.

across crops. Second, there is virtually no evi-dence that colony density is related to annualchanges in crop or honey prices, almond acres,the advent of Varroa, or the onset of CCD.We attribute this lack of relationship to twofactors.

The first is that the cost share of pollina-tion is small in the production of most crops.For example, apple yields in recent years havebeen roughly 10 tons per acre and prices havebeen roughly 17 cents/lb, for per acre revenuesof about $3,600. Recent (nominal) apple pol-lination fees are roughly $45 per hive and thesurveys suggest that 1.34 hives per acre is typ-ical. The fraction of total revenue accountedfor by pollination fees for apples is, then,$60.30/$3, 600 = .0168, or 1.7 percent. One ofMarshall’s laws posits that when the cost shareof an input is small, the derived demand elas-ticity is small as well. (See Marshall 1920, andMuth 1964.) While there may be non-zeroeffects of temporal variation in crop prices,almond acres, honey prices, and cost factors onhive density, they are small enough to not beidentifiable empirically.

The second reason that the derived demandfor pollination services is likely to be inelasticis due to the state of knowledge regarding thebenefits of bee pollination to crops. Althoughcrop science advisors make recommendationsto farmers regarding the proper number ofcolonies to put on an acre of, say, almonds,we have never seen such advice conditionedon economic factors. This observation, com-bined with our estimation results discussed

above, suggests that advisors and farmers actas if they perceive their production processesto be fixed regarding proportions of landand bees.

Figure 4 shows the situation described abovein which the value of marginal product ofbees in fruit production has a vertical section,as it does for a fixed-proportion productionfunction. This figure is heuristically useful forunderstanding the predictions of our model,given our empirical finding that hive stockingdensity is generally insensitive to the factors inour empirical analysis of the determinants ofpollination fees.

Figure 4. Pollination fees with fixed stockingdensity

by Murray B



ownloaded from



Table 3. Pacific Northwest Pollination Fees byCrop, 1987–2009(2009 dollars per colony rental)

N Mean Min Max

Honey CropsBlueberries 23 33.18 21.34 43.44Crimson

Clover21 8.95 0.00 36.84

Radishes 23 31.42 15.92 49.23Red Clover 22 27.07 9.93 46.47Vetch 15 3.11 0.00 11.86Group 104 22.27 0.00 49.23

statisticsNon-honeycrops

Apples 23 38.80 22.67 49.68Cherries 23 38.70 27.08 53.71Cranberries 20 45.48 29.05 60.00Cucumbers 21 39.46 23.47 71.04Pears 23 38.23 24.01 51.41Squash 21 40.64 26.20 60.78Group 131 40.10 22.67 71.04

statisticsAlmonds 17 75.61 46.94 150.27All crops 252 35.14 0.00 150.27

An Econometric Model of Pollination Fees

We now turn to panel regression analysis of252 observations on crop-average per-colonypollination fees from the PNW beekeeper sur-veys. The data span the years 1987–2009, andinclude information on 12 crops. Table 3 dis-plays sample-period averages of fees, dividedinto three categories: almonds, non-almondcrops from which bee pollination producesno marketable honey (non-honey crops), andnon-almond crops from which bee pollinationproduces marketable honey (honey crops).

Standard economic theory suggests thatboth cost and demand factors explain polli-nation fees. The theory developed in the pre-vious section also suggests that the value ofthe in-kind payment of honey to beekeepersinfluences the equilibrium monetary payment,which in turn implies the importance of thehoney output of the pollinated crop and theprice of honey.

Strong evidence for the importance of thehoney output of the pollinated crop is foundin figure 5, which shows the annual real(2009 dollar) averages of pollination fees percolony for almonds, non-honey crops otherthan almonds, and honey crops. Almondscould be included in the non-honey crop

category, but due to the previously dis-cussed importance of almonds and the factthat almond fee behavior departed fromprevious norms after 2004, we consider itseparately.

The first thing to note from figure 5 isthat the pollination fees for non-honey cropsare consistently higher than those for honeycrops. The difference between the sample aver-age of honey-crop fees and non-honey-cropfees is $17.84 in 2009 dollars (we considera more formal test of this difference in theregressions to come). If almonds were includedin the non-honey crop group, the differencebetween honey and non-honey crop fees wouldbe larger. This corroborates the predictionthat the higher the value of the honey pay-ment to beekeepers, the lower is the monetarypayment.

The second issue to note from figure 5 isthe unprecedented increase in almond feesthat occurred after 2004–behavior not seenfor other surveyed crops. Almond pollinationfees rose from $59 to $89 between 2004 and2005, and increased again to close to $140in inflation-adjusted terms from 2006–2009. Itis tempting to attribute these fees to ColonyCollapse Disorder, and CCD may be partlyto blame, but the timing is not right. Thefirst reported instance of CCD was during thefall of 2006, which could only have affectedfees beginning in spring 2007. But as figure 5shows, almond fees rose earlier, in 2005 and2006.23

Because recent changes in almond fees are aprominent part of the data we analyze, we con-sider the time patterns of other possible expla-nations of recent high almond fees. Figure 6displays two of them:the real price of diesel fuel(an important input into migratory beekeep-ing) and the numbers of nut-bearing almondacres. It is clear from figure 6 that there is ahigh degree of collinearity between these twoseries (the correlation coefficient between thetwo is 0.86). Coupling that fact with the limitedduration of the recent period of high almondfees should cause one from the outset to notbe optimistic about the ability of a few annualobservations to definitively disentangle theseeffects.

Our empirical analysis is based on thefollowing crop-wise heteroskedastic linear

23 A comprehensive study of the economic consequences ofCCD can be found in Rucker, Thurman, and Burgett (2012).

by Murray B



ownloaded from



160

140

120

100

80 almondsnon-honey crop averagehoney crop average

60

40

20

01990 1995 2000 2005 2010

Figure 5. Real pollination fees per colony (2009 dollars)

350

300

250

200

almond fees

real diesel price

almond acres

150

100

501985 1990 1995 2000 2005 2010

Figure 6. Index numbers (1993 = 100)

by Murray B



ownloaded from



panel model:

yit =12∑

j=1

μjdjit + ϕVit + x′

itdAit βA(23)

+ x′it(1 − dA

it )βN + εit,

for i = 1, . . . , 12

and t = 1987, . . . , 2009

The variables are defined as follows:yit = real pollination fee for crop i in

year t, measured by the average ofresponses across beekeepers,

djit = crop dummy variables equaling 1

when i = j,Vit = a dummy variable equaling 1 when

t ≥ 1991, the year in which Varroabegan to cause large-scale losses inthe Pacific Northwest,

xit = a vector of explanatory variables,including subsets of the following:the log of one-step-ahead forecasthoney price, the log of one-step-ahead forecast crop price, the log ofdiesel fuel price, the log of almond-bearing acres, a post-2004 dummyvariable, and a post-2006 dummyvariable,

dAit = a dummy variable equaling 1 for

observations on almond fees.The disturbance terms are assumed to be het-

eroskedastic but uncorrelated across crops andover time:

Var(εit) = σ2i , Cov(εit, εis) = 0 for t �= s,

and Cov(εit, εjt) = 0 for i �= j.

The semi-log specification in (22) allowsfor sensible aggregation across crops. For allcrops, the left-hand side variable is measuredin real dollars per colony of bees. On theright-hand side, prices of crops refer to differ-ent commodities and taking logarithms con-verts their changes into comparable percent-age change magnitudes. Notice that in (22),crops are distinguished from one another bycrop-specific intercepts and crop-specific vari-ances. Almonds are further distinguished fromnon-almond crops by the almond dummy inter-action terms.

The panel is nearly complete but containsholes due to survey non-response. Comparedto a total potential number of observations of

12 crops × 23 years = 276, our data set com-prises 252 usable observations. We have noreason to suspect that the pattern of non-response is correlated with any of the covari-ates in our model. In table 4 we reportWeighted Least Squares (WLS) estimates ofequation 22, where first-stage Ordinary LeastSquares residuals are used to estimate thecrop-specific variances used in the secondstage.

A summary of the expected effects inequation (22) is as follows: (1) We predict thatgrowers of crops that provide valuable nec-tar to the bees (honey crops) will pay lessin pollination fees than do growers of non-honey crops. (2) We predict that cost factorsinfluence pollination fees. Specifically, we pre-dict that the arrival of Varroa shifted upwardthe supply of pollination and, hence, increasedpollination fees. Similarly, the price of dieselfuel represents a more continuously-varyingcost factor, which we predict has a positiveinfluence on prices. (3) We predict that to theextent that the supply of pollination servicesis upward sloping, almond pollination fees willvary directly with almond acres (although pos-sibly only for almond pollination fees and notfor later- pollinated crops). (4) We predict thathigher expected prices of crops will have noeffect on pollination fees to the extent that beestocking rates are unaffected by crop prices(which we argued in the previous section tobe the case). (5) A change in the honey pricehas a theoretically ambiguous effect and is tobe determined empirically.

Empirical results are reported in table 4. Thesix specifications in table 4 all include effects forVarroa, honey price, and crop price. The spec-ifications differ in their inclusion of effects fordiesel fuel price, almond acreage, and CCD.Further, all regressions in table 4 include apost-2004 dummy variable, motivated by thesubstantial increase in almond fees after 2004,displayed in figure 5.While not so visually obvi-ous, the post-2004 dummy variable is allowedto have an effect on non-almond fees as well.The other recent time effect considered in halfof the specifications is a post-2006 dummy vari-able that we interpret as a CCD indicator;CCD was first widely reported after the 2006pollinating season, but prior to the 2007 polli-nating season. The first three specifications intable 4 do not include a CCD effect; the lastthree do.

All specifications in table 4 include crop-specific intercepts, and tests of the equalityof the intercepts (not shown) are strongly

by Murray B



ownloaded from


972July

2012A

mer.

J.Agr.

Econ.

Table 4. Pollination Fee Panel Regressions - Cropwise Heteroskedastic GLS

Dependent variable = Real (2009 dollars) pollination feePanel of 12 crops over 23 years: 1987–2009, n = 252Crop-specific intercepts fit in all specifications:

almonds, pears, cherries, apples, red clover, crimson clover, vetch, cucumbers, blueberries, cranberries, radishes, squash

Specifications with no CCD effect Specifications with CCD effect

(1) (2) (3) (4) (5) (6)

Variable coefficient t-stat coefficient t-stat coefficient t-stat coefficient t-stat coefficient t-stat coefficient t-stat

Honey crop effect: differencebetween average non-honeycrop intercept (excludingalmonds) and average honey-crop intercept 16.74 5.83∗∗∗ 17.07 5.86∗∗∗ 17.10 5.87∗∗∗ 16.54 5.65∗∗∗ 17.12 5.76∗∗∗ 16.90 5.69∗∗∗

Varroa 6.13 5.25∗∗∗ 6.31 5.69∗∗∗ 7.03 5.87∗∗∗ 6.17 5.12∗∗∗ 6.28 5.58∗∗∗ 7.07 5.73∗∗∗Log honey price almonds 6.09 0.52 −21.44 −1.05 7.40 0.38 7.00 0.65 −18.94 −1.15 −1.24 −0.07

others 8.43 3.74∗∗∗ 5.02 1.76∗ 2.30 0.69 8.53 3.73∗∗∗ 4.85 1.61 2.43 0.72Log crop price almonds −5.54 −0.35 3.74 0.15 −6.73 −0.32 2.44 0.16 24.64 1.13 10.97 0.51

others −1.62 −0.78 −1.35 −0.64 −1.32 −0.62 −1.77 −0.83 −1.30 −0.60 −1.47 −0.68Log diesel price almonds 77.90 3.81∗∗∗ 79.51 2.85∗∗∗ 63.10 3.05∗∗∗ 51.10 1.71∗

others 1.17 0.31 −7.74 −1.51 0.15 0.03 −8.82 −1.53Log almond acres almonds 88.04 2.05∗∗ −4.05 −0.08 83.32 2.40∗∗ 25.70 0.55

others 8.78 1.92∗ 15.66 2.45∗∗ 9.30 1.86∗ 15.61 2.41∗∗Post 2004 = 1 almonds 37.56 3.32∗∗∗ 52.26 3.93∗∗∗ 37.89 3.17∗∗∗ 35.74 3.42∗∗∗ 37.93 3.21∗∗∗ 33.45 3.00∗∗∗

others 4.71 2.24∗∗ 2.82 1.77∗∗ 4.79 2.31∗∗ 4.82 2.26∗∗ 3.00 1.77∗ 4.90 2.33∗∗CCD (post 2006 = 1) almonds 14.89 1.71∗ 25.00 2.93∗∗∗ 16.70 1.81∗

others 0.83 0.42 −0.38 −0.22 0.89 0.46R2 0.889 0.886 0.891 0.890 0.891 0.892

Note: ∗ , ∗∗ , and ∗∗∗denote significance at the 10%, 5%, and 1% levels, respectively, with two-sided tests.

by Murray Belyk on September 5, 2012 http://ajae.oxfordjournals.org/ Downloaded from



rejected.24 The first row of table 4 reports ahoney crop effect,constructed as the differencebetween the average intercept for non-honeycrops (apples, cherries, cranberries, cucumbers,pears, and squash, excluding almonds) and theaverage intercept for honey crops (blueber-ries, crimson clover, radishes, red clover, andvetch). The standard error for the differencein averages is calculated using the variancesand covariances of the estimated intercepts.The estimated honey-crop effect is roughly $17in all specifications (average non-honey cropfee higher than average honey-crop fee),whichis quite close to the difference between theunconditional averages of the honey and non-honey crop fees displayed in figure 5 ($17.84).The estimate is highly significant statistically,with a 95% confidence interval of the col-umn (1) estimate spanning the interval ($11.08,$22.40).

The second row of table 4 reports estimatesof a Varroa effect, the coefficient on a dummyvariable equal to one in 1991 and beyond.The point estimates are similar and stronglysignificant in all specifications. In column 1,the point estimate of $6.13 is significant atthe 1% level, and centers a 95% confidenceinterval of between $3.83 and $8.44. Indus-try observers noted after the 1990 pollinationseason that the long-awaited Varroa infesta-tion had arrived in the PNW, causing bee-keeper costs to rise. The same observers notedthat it was the good fortune of beekeep-ers that pollination fees rose by just aboutthe right amount to offset the Varroa-inducedincrease in costs.25 We attribute this coinci-dence not to the good fortune of the indus-try but rather to the forces of competitiveequilibrium.

Beekeepers and farmers agree on pollina-tion fees at the time that colonies are placedin orchards and fields, typically in the springor early summer months. Fees, then, are deter-mined prior to the time that actual crop pricesfor the year are known. We model the expec-tation for crop prices as one-step-ahead fore-casts from second-order autoregressive models

24 Champetier,Sumner,andWilen (2012) argue that biodynamicfactors over the pollinating season should induce seasonality inpollination fees, which would be reflected in the values of the crop-specific intercepts.

25 The following calculation justifies this statement.Annual costsof Varroa control (miticide and application costs) are approxi-mately $18 per colony (Barnett 2002.) A typical colony in oursample was rented out 2.5 times in a year,yielding a per-pollination-placement treatment cost of $18/2.5 = $7.20, which is close to ourpoint estimate of the Varroa effect and comfortably within itsconfidence interval.

in real prices. The parameters are estimatedfrom 1976–2009. The same issues arise forOregon honey prices, which are determinedafter pollination fees are agreed upon. Inthe presence of the honey price support pro-gram,however,each year’s honey price supportlevel was known at the time that pollinationfees were specified. (See Muth et al. 2003)Accordingly, to account for the presence of thehoney program during a portion of our sampleperiod, we use the honey price for year t whenthe program was in effect (until 1993), andthereafter use a one-step-ahead forecast froma second-order autoregressive model for theprice of honey expected at the time that polli-nation fees are set (estimated using data duringthe period that the honey price support wasnot effective).26

The effect of honey price on pollination feesis theoretically ambiguous. As per our theory,there are two channels of influence from honeyprice to pollination fees. One is that an increasein honey price makes more valuable the in-kindpayment to beekeepers and therefore reducesthe required pollination fee. The second isthat an increase in expected honey price candecrease the aggregate level of crop pollinationactivity,which can thereby increase fees.Whichof these effects dominates is an empirical ques-tion addressed in the table 4 specifications. Inthe first column of table 4, the estimated loghoney price coefficient is reported as 8.43 fornon-almond crops and 6.09 for almonds. Thenon-almond honey coefficient is significant atthe 1% level, while the almond honey coeffi-cient is insignificant. This pattern is repeated inspecifications (2) and (4).

In the semi-log specification, the non-almond coefficient of 8.43 implies that a 10%increase in the price of honey induces an$0.843 increase in pollination fee per crop. Theaverage annual absolute percentage change inhoney prices between 2000 and 2009 is 15.3%,which according to the table 4, column (1)estimate would induce a $1.29 change in thetypical non-almond pollination fee. Notice thatthe almond honey price effect is statisticallyinsignificant, but similar in size to the non-almond honey coefficient.

26 Note that we use the actual average honey price rather thanthe support price. Given that there were different support prices fordifferent grades of honey, and given that support prices were bind-ing during the period of our analysis, we assume that the observedaverage honey price was an appropriately weighted average of thevarious support prices.

by Murray B



ownloaded from



A logical interpretation of the honey priceresults is as follows. Higher honey prices do notbid beekeepers away from almonds, becausealmonds bloom in late winter when other nec-tar sources are scarce; higher honey prices dotrigger a substitute supply response away fromnon-almond crops, which bloom later in theseason when honey production is a more viablealternative to pollination work. We also notethat the insignificance of the almond coefficientis partly due to the fact that there are only 17observations on almond fees in the panel andthat modest sub-sample size is likely to revealonly a strong effect.

The fourth effect reported in table 4 is thecrop price effect. Across all specifications itis robustly statistically insignificant. This sup-ports the joint hypothesis of the equilibriummodel of equation (17) and the constancyof stocking densities with respect to cropprices.

Next, turn to the effects of diesel price andalmond acres on pollination fees. Due to thepreviously discussed collinearity between thetwo variables, we consider specifications withonly diesel price (specification 1), only almondacres (specification 2), and both diesel priceand almond acres (column 3). The same threemodels are repeated in specifications 4, 5, and6, which also include a CCD dummy variable.As with honey and crop prices, the effectsof almond acres are estimated separately foralmonds and non-almond crops.The diesel fuelprice variable is lagged one year to reflect thefact that pollination fees are received in theearly part of the year, and the most relevantfuel price is likely to be that involved in truck-ing bees in preparation for the winter/springpollination season.

In specification 1, the coefficient on dieselprice is large and significant for almond fees.The estimated coefficient of 77.90 implies thata 10% increase in the price of diesel fuel willlead to a $7.79 increase in almond fees. Inter-estingly, for non-almond fees the estimatedeffect of diesel fuel is much smaller and notstatistically significant. The contrasting resultscan plausibly be attributed to the differencebetween the two types of crops in the impor-tance of pollination transport costs. In recentyears, bees have been trucked to the Califor-nia almond orchards from as far away as NorthCarolina, and the PNW beekeepers in our dataset truck their bees substantial distances southto the examined almond orchards. After pol-linating almonds, the PNW beekeepers returntheir bees to their home bases and, from there,

proceed to pollinate the other 11 crops. Trans-port represents a lower cost share for thepost-almond, home-based crops, which couldexplain why variations in diesel fuel pricehave smaller and statistically insignificant coef-ficients for those crops.

Specification 2 replaces diesel fuel priceswith the collinear almond acres. The esti-mated effect on almond pollination fees fromalmond acreage is large and statistically sig-nificant. The semi-elasticity with respect toalmond acres is similar in size to the semielas-ticity with respect to diesel price estimatedin specification 1. Almond acres in specifi-cation 2 is also a statistically significant fac-tor in explaining non-almond fees, but thealmond acre effect on non-almond fees is muchsmaller than the almond acre effect on almondfees.27

Specification (3) pits diesel fuel prices andalmond acres against one another as possi-ble explanations of pollination fees. In thisspecification,diesel price remains a statisticallysignificant factor in the presence of almondacres, but the almond acre effect does not sur-vive the inclusion of diesel prices. The oppositeis true for non-almond fees: when both dieselprices and almond acres are allowed into theregression (specification 3), almond acres isthe significant explanatory variable and dieselprice is not.28

Lastly, consider the significance and size ofrecent changes in pollination fees that are notexplained by factors considered so far. All ofthe specifications in table 4 include post-2004dummy variables, motivated by the notablerecent increases in almond fees, which arevisually apparent in figures 5 and 6. The post-2004 dummy variable is statistically significantfor both almond and non-almond crops andquite robust across specifications. The post-2004 effect ranges between $33 and $52 acrossspecifications for almond fees, and from $3 to$5 for non-almond fees. These are statisticallysignificant effects that are puzzling when con-sidered in connection with Colony CollapseDisorder, first reported in late 2006. Therewere, however, earlier indications of concerns

27 After pollinating almonds, bees are in a more weakened statethan they would be if they skipped almond pollination. This couldbe one reason that almond acres are found to have a positive effecton subsequent non-almond pollination fees.

28 As the t-tests suggest, diesel prices and almond acres havejointly significant effects on almond pollination fees in column 3. AWald test with two degrees of freedom has a p-value of 0.007. Thep-value for the joint significance of the effect of diesel prices andalmond acres on non-almond pollination fees is 0.047.

by Murray B



ownloaded from



over honey bee health in the spring of 2005. SeeMussen (2005) for a discussion of abnormallyhigh overwinter bee mortality among Cali-fornia beekeepers in the winter of 2004/2005,which could imply a leftward shift in the sup-ply of bees, and an increase in equilibrium feesfor all crops.29

To represent CCD itself, specifications 4–6replicate specifications 1–3, but add a post-2006 dummy variable, again interacted withan almond indicator to allow different CCDresponses in almond and non-almond fees. Theindicator coefficient shows a statistically sig-nificant $15–25 rise in almond pollination fees,depending on specification, beginning in 2007–in addition to the post-2004 effect. This couldrepresent a CCD supply shift, but if so, whydoes it not also show up in non-almond fees?Non-almond fee increases in the post-CCDperiod are statistically insignificant in specifi-cations 4–6 and their point estimates are small(less than $1). One possible explanation isthat if some beekeepers get hit particularlyhard by CCD over the winter, then they willnot be able to recover for almond pollination,and almond fees will increase. There is a lagbetween almonds and subsequent crops, allow-ing colonies to recover or new colonies to build,implying little or no impact on pollination feesfor these other crops.

The main conclusions of the empirical anal-ysis are as follows. The factors that we findto be significantly related to pollination feesare those that influence beekeeper costs andreturns: honey yield, Varroa mites, the price ofdiesel fuel (for California almonds) and theprice of honey. The factor that we find not tobe significantly related to pollination fees–cropprices–is one that influences farmer costs andreturns. This makes sense, as the cost shares ofpollination fees are typically small to farmers,and we infer from our estimates that there islittle or no year-to-year response to them byfarmers, just as we find little or no response instocking densities. On the other hand, honeyprice, the costs of Varroa control, and the valueof honey obtained from pollinating differentcrops loom large in the beekeepers’ balancesheets. As predicted by theory, we see evidenceof substantial response to these influences. The

29 Mussen suggested (p. 2) that the high overwinter mortality was“easiest to blame (on) Varroa,” but that some of the bee mortalitymay have been due to “lack of brood rearing during the extremelycritical time from August through September, when the bees arerearing their ‘winter bees.”’ Lack of brood results in “colony pop-ulations that just dwindle in numbers continually from October todeath before March.”

fact that farmers regularly employ bees to pol-linate their crops demonstrates the existenceof robust derived demands for pollination, butour results suggest the demands are relativelyinelastic. At the same time, the supply-sideinfluences from the costs of beekeeping andthe value of in-kind returns to beekeepers aremeasurably important factors in determiningpollination fees.

Beyond measuring and confirming effectsthat flow directly from the theoretical model,the results offer evidence on the effects ofrecent concerns over honey bee health. Signif-icant and sizeable increases in almond pollina-tion fees are seen to begin after 2004–and theyappear not to be explained by the cost factors inour analysis, or by increases in almond acres. Aportion of this increase ($15–25) can plausiblybe attributed to CCD. More modest increasesare measured for non-almond crops, but theydo not coincide with the arrival of CCD.

Summary and Conclusions

Despite the importance of honey bee polli-nation for many agricultural crops, there hasbeen little economic analysis of pollinationmarkets. In this paper, we develop a modelto represent modern pollination markets andempirically test its predictions. Our empiricalanalysis, based on annual surveys of Oregonand Washington beekeepers, suggests that pol-lination markets respond to changes in eco-nomic factors that are predictable in light ofthe model.

Our results have important implicationsfor the growing body of literature on thestatus of wild pollinators and the valuationof related ecosystem services–they suggestthat, at least in the United States, thereare relatively inexpensive substitutes for wildpollinators that are coordinated by marketprices. Economically defensible estimates ofthe value of related ecosystem services shouldaccount for the availability and costs ofsuch substitutes. Beyond the U.S. border lierelated research questions–which we leavefor future work–involving the evolution andcurrent status of pollination markets else-where: where have such markets developedand why? What are the resulting implica-tions for agricultural development, the orga-nization of production in agriculture, andthe ability of human-managed pollination toadapt to changing environmental and eco-nomic circumstances?

by Murray B



ownloaded from



Acknowledgments

The authors are grateful for helpful com-ments and encouragement from Ronald Coase,Chuck Knoeber, Mitch Renkow, Lester Telser,editor Erik Lichtenberg, two anonymous ref-erees, and participants at seminars at BrighamYoung, Clemson, Montana State, North Car-olina State, Oregon State, Utah State andWashington State Universities, and at the Uni-versities of California (Davis), and Maryland.

Funding Statement

Rucker acknowledges the support of the Mon-tana Agricultural Experiment Station andThurman the support of the North CarolinaAgricultural Research Service.

References

Barnett, R. 2002. Belgrade, Montana bee-keeper, personal communication.

Bator, F.M. 1958. The Anatomy of MarketFailure. Quarterly Journal of Economics72: 351–364.

Barzel,Y. 1997. Economic Analysis of PropertyRights. Cambridge: Cambridge UniversityPress.

Burgett,M. 1992–2010.“Pacific Northwest Pol-lination Survey.” Various issues of USDAHoney Market News.

Burgett, M., S. Daberkow, R.R., and W.N.Thurman. 2010. U.S. Pollination Markets:Recent Changes and Historical Perspec-tive. American Bee Journal 150(1): 35–40.

Burgett, M., R.R. Rucker, and W.N. Thurman.2004. Economics and Honey Bee Pol-lination Markets. American Bee Journal144(4): 269–271.

Carman, H. 2011. “The Estimated Impact ofBee Colony Collapse Disorder onAlmondPollination Fees.” ARE Update 14(5):9–11. Giannini Foundation of AgriculturalEconomics.

Champetier, A., D.A. Sumner, and J.E. Wilen.2012. “The Bioeconomics of Honey Beesand Pollination.” Working paper, Dept. ofAg. Econ., University of California, Davis.

Cheung, S.N.S. 1973. The Fable of the Bees:AnEconomic Investigation. Journal of Lawand Economics 16:11–33.

Crane, E. 1999. The World History of Bee-keeping and Honey Hunting. New York:Routledge Press.

Daberkow, S., R.R. Rucker, W.N. Thurman,and M. Burgett. 2009. U.S. Honey Mar-kets: Recent Changes and Historical Per-spective. American Bee Journal 149(12):1125–1129.

Degrandi-Hoffman,G.,R.Thorp,G. Loper,andD. Eisikowitch. 1992. Identification andDistribution of Cross-Pollinating Honey-Bees on Almonds. Journal of AppliedEcology 29: 238–246.

Gallai, N., J-M Salles, J. Settele, and B.E. Vais-sière. 2009. Economic Valuation of theVulnerability of World Agriculture Con-fronted with Pollinator Decline. Ecologi-cal Economics 68: 593–924.

Hoopingarner, R.A. and G.D. Waller. 1992.“Crop Pollination.” In J.M. Graham, ed.The Hive and the Honey Bee, J.M. Graham,ed. Hamilton, Illinois: Dadant & Sons,chapter 24.

Horn,T. 2005. Bees in America: How the HoneyBee Shaped a Nation. Lexington, Ken-tucky: The University Press of Kentucky.

Johnson, D.B. 1973. Meade, Bees, and Exter-nalities. Journal of Law and Economics 16:35–52.

Klein, B., and K.B. Leffler. 1981. The Roleof Market Forces in Assuring Contrac-tual Performance. Journal of PoliticalEconomy 89: 615–641.

Knoeber, C.R. and W.N. Thurman. 1994. Test-ing the Theory of Tournaments:An Empir-ical Analysis of Broiler Production. J. ofLabor Economics 12: 155–179. In E.P.Lazear and R. McNabb, eds. Person-nel Economics. Edward Elgar Publishing,Ltd.

Marshall, A. 1920. Principles of Economics,8thed. London: Macmillan and Co., Ltd.

Mairson, A. 1993. “America’s Beekeepers:Hives for Hire.” National Geographic183(5): 75–93.

Mayer, D.F and C.A. Johansen. 2003. “TheRise and Decline of Nomia melanderi(Hymenoptera: Halictidae) as a Com-mercial Pollinator for Alfalfa Seed.” InK. Strickler and J.H. Cane, eds. For Non-native Crops, Whence Pollinators of theFuture? (chapter 9). Lanham, Maryland:Entomological Society of America.

Meade, J.E. 1952. External Economies andDiseconomies in a Competitive Situation.Economic Journal 62: 54.

Morse, R.A. and N.W. Calderone. 2000. “TheValue of Honey Bees as Pollinators ofU.S. Crops in 2000.” Bee Culture 128(3):1–15.

by Murray B



ownloaded from



Mussen, E. 2005. “Almonds and Honey Bees.”Apiculture News. University of California,Davis Department of Entomology.March 2005. Available at: http://entomology.ucdavis.edu/faculty/mussen/news.cfm.

Muth, M.K. and W.N. Thurman. 1995. “WhySupport the Price of Honey?” Choices10(2): 19–21.

Muth, M.K., R.R. Rucker, W.N. Thurman, andC-T Chuang. 2003. The Fable of the BeesRevisited: Causes and Consequences ofthe U.S. Honey Program. Journal of Lawand Economics 46: 479–516.

Muth, R. 1964. The Derived Demand CurveFor a Productive Factor and the IndustrySupply Curve. Oxford Economic Papers16: 221–234.

National Research Council. 2007. Status ofPollinators in North America. Washington,D.C.: The National Academies Press.

Nordhaus, H. 2011. The Beekeeper’s Lament.Harper Perennial: New York.

Olmstead,A.L. and D.Wooten. 1987. Bee Polli-nation and Productivity Growth:The Caseof Alfalfa. American Journal of Agricul-tural Economics 69(1): 56–63.

Pellett, F.C. 1938. History of American Bee-keeping. Ames Iowa: Collegiate Press.

Pigou,A.C. 1912. Wealth and Welfare. London:Macmillan.

Pigou, A.C. 1920. The Economics of Welfare.London: Macmillan.

Robinson, W.S., R. Nowogrodzki, and R.A.Morse. 1989. The Value of Honey Bees asPollinators of U.S. Crops. American BeeJournal 139(6): 411–23 (Part I) and 139(7):477–87 (Part II).

Rucker, R.R. and W.N. Thurman. 2011.“Blessed Are the Beekeepers.” Wall StreetJournal, June 22.

Rucker, R.R., W.N. Thurman, and M. Burgett.2012. “The Economic Consequences ofColony Collapse Disorder.” Workingpaper, Dept. of Ag. and Res. Econ., NorthCarolina State University.

Seeley,T.D. 1995. The Wisdom of the Hive:TheSocial Physiology of Honey Bee Colonies.Cambridge: Harvard University Press.

Sumner, D.A. and H. Boriss. 2006. Bee-conomics and the Leap in PollinationFees. Agriculture and Resource EconomicsUpdate 9(3). Giannini Foundation of Agri-cultural Economics.

Swinton, S.M., F. Lupi, G.P. Robertson, andS.K. Hamilton. 2007. Ecosystem Servicesand Agriculture: Cultivating AgriculturalEcosystems for Diverse Benefits. Ecologi-cal Economics 64(2): 245–252.

Thorp, R.W. 2003. “Bumble Bees(Hymenoptera: Apidae).” In K. Stricklerand J.H. Cane, eds., For NonnativeCrops, Whence Pollinators of theFuture?, chapter 2. Lanham, Maryland:Entomological Society of America.

U.S. Department of Agriculture. 2007. Tran-script of Agriculture Secretary MikeJohanns Remarks to the Pollinating Part-ners Celebration Ceremony and Releaseof a New U.S. Stamp. Washington DC,June.

Ward, R., A. Whyte, and R.R. James. 2010.A Tale of Two Bees: Looking at Pol-lination Fees for Almonds and SweetCherries. American Entomologist 56(3):172–179.

by Murray B



ownloaded from

http://entomology.ucdavis.edu/faculty/mussen/news.cfm




honey bee pollination marketsandthe …...european honey bee (apis mellifera) is widely used to...

Documents