Content uploaded by Todd Sanderson
Author content
All content in this area was uploaded by Todd Sanderson on Nov 27, 2017
Content may be subject to copyright.
Cultivating Trust: Towards
an Australian Agricultural
Data Market
Todd Sanderson, Andrew Reeson, Paul Box
Cultivating Trust: Towards an Australian Agricultural Data Market | 2
Citation
Sanderson T, Reeson A and Box P (2017) Cultivating trust: Towards an Australian agricultural data
market. CSIRO, Australia.
Licensed under Creative Commons Attribution 3.0 Australia
Copyright
© Commonwealth Scientific and Industrial Research Organisation 2017. To the extent permitted by
law, all rights are reserved and no part of this publication covered by copyright may be reproduced
or copied in any form or by any means except with the written permission of CSIRO.
Important disclaimer
CSIRO advises that the information contained in this publication comprises general statements
based on scientific research. The reader is advised and needs to be aware that such information may
be incomplete or unable to be used in any specific situation. No reliance or actions must therefore
be made on that information without seeking prior expert professional, scientific and technical
advice. To the extent permitted by law, CSIRO (including its employees and consultants) excludes all
liability to any person for any consequences, including but not limited to all losses, damages, costs,
expenses and any other compensation, arising directly or indirectly from using this publication (in
part or in whole) and any information or material contained in it.
CSIRO is committed to providing web accessible content wherever possible. If you are having
difficulties with accessing this document please contact enquiries@csiro.au.
Cultivating Trust: Towards an Australian Agricultural Data Market | 3
Acknowledgments
The research presented in this report was undertaken as part of the National Soil Data project, co-
funded by CSIRO and the Australian National Data Service (ANDS). The authors would like to
acknowledge this important strategic investment that contributes to an improved understanding
of economics of data communities and markets.
In addition, the authors acknowledge the contribution of the following individuals and organisations
to the Soil Data Community Workshop which led to the development of this report: Peter Wilson
and Ross Searle (CSIRO), Chris Sounness and other members of the Birchip Cropping Group, Mark
Pawsey (SST Software) Julia Martin and Melanie Barlow (Australian National Data Service) and Peter
Dahlhaus (Federation University). Their input and feedback has been of enormous value in the
development of this work. We also thank Melanie Ayre and Tim Capon for constructive comments
and review.
Cultivating Trust: Towards an Australian Agricultural Data Mar ket | 4
Contents
1 Introduction 5
2 The agricultural data ecosystem 6
2.1 Who are the players in the data ecosystem? ................................................................... 6
2.2 Platforms as data market environments .......................................................................... 7
3 Agricultural data as a public vs private good 10
3.1 Data clubs, cooperatives and collectives ....................................................................... 13
4 The cost and value of sharing agricultural data 14
4.1 The cost of sharing data ................................................................................................ 14
4.2 The values of shared data.............................................................................................. 15
5 Institutions and incentives 19
5.1 Understanding the needs of data providers .................................................................. 19
5.2 Incentives for data sharing ............................................................................................ 22
5.3 Approaching markets for sharing data........................................................................... 24
5.4 Balancing privacy .......................................................................................................... 25
6 Towards an Australian agricultural data market 27
References 28
Cultivating Trust: Towards an Australian Agricultural Data Market | 5
1 Introduction
Sensors and other data-gathering devices are ubiquitous in the digitally connected world. In the
agricultural context sensors are increasingly embedded in mobile equipment, such as harvesters,
and as stationary monitoring devices for weather, soil moisture and other production-relevant
variable phenomena. The proliferation has been driven by a combination of lower installation and
operation costs, and a growing understanding of the production value of data to support farm level
decision-making. The streams of sensor data add to the vast collections of research, government
and private agricultural systems data which have been collected and stored for decades. These are
the weather records, the soil physical and chemical properties lab tests from in-field samples, and
the crop and livestock trials data held by research institutions, to name only a few.
Such data streams offer considerable benefits to farmers and other industry participants, but there
is also growing awareness of the risks of big data enabling a small number of large companies to
increase their market power (Jakku et al. 2016). While the technology has advanced rapidly, the
institutions to support the efficient and equitable exchange of data between those who can
generate and use it remain under-developed. Institutions include the rules, organisations and
expectations which determine how people interact to use data. The right institutions will cultivate
the trust that is essential to realising the broader benefits of agricultural data. Farmers, and others,
will be loath to share their data if they fear it may be used against them.
This report takes an economic and market design perspective to consider institutional arrangements
which could support an inclusive agricultural data market. There are numerous analogues in digital
(eBay, Facebook, Google, etc.) and physical goods markets which suggest that a mechanism, such
as a data market, to connect data holders with the myriad of service providers and secondary data
users is feasible (for example, Poppe et al. 2015). In this context a ‘data market’ is the facilitated
interaction of those who have data (i.e. data owners and generators) with those who want data (i.e.
data demanders), for the purpose of mutually beneficial exchange. Data markets may exist at all
manner of scales and scopes, ranging from the one-to-one exchange of a single data collection, to
the one-to-many exchange of one or multiple collections.
This report first examines the nature of the players in the agricultural data ecosystem, and examines
the scope for digital platforms to provide structure to establish a data market. Second, we tackle
the notion that data is a natural public good, whereby value is maximised by sharing it openly. While
this may work well for some types such as scientific research data (Sanderson et al. 2017), it should
not be expected to apply universally to all data. Third, the costs and value potential of sharing
agricultural are considered, and we derive particular implications for the scale at which data is
aggregated. Finally, we review the institutional and incentive considerations which will be important
as we move from an agricultural data ecosystem towards an Australian agricultural data market.
Cultivating Trust: Towards an Australian Agricultural Data Market | 6
2 The agricultural data ecosystem
The agricultural data ecosystem encompasses the existing networks within which agricultural data
is generated and shared. This emergent ecosystem is mostly ad hoc and unstructured, resulting in
much of the potential values of agricultural data not being fully realised. A market design analysis
can show how the system might be made more efficient. In this instance, our intervention seeks to
improve the nature of connections among data holder and demanders by providing a more formal
institutional structure. In order to understand how this might be achieved, we first need to
understand the nature of the players in the data ecosystem, and second examine the potential of a
digital market platform to bring these players together.
2.1 Who are the players in the data ecosystem?
It is useful first to define some of the possible actors involved in soils and agricultural data supply
chains (Figure 1). There are three significant groups engaged directly or indirectly in the exchange
of data which establish the market: data providers, service providers and data/service users. Data
providers are the individuals or groups who generate data, such as the owners/operators of the in-
situ sensors which measure phenomena of interest, for example soils moisture, crop yield, rainfall
or temperature. They can also be those who have generated data ex-situ, for example through
analysis of soils samples in a lab.
Property rights are a pre-requisite to trade. However, ownership of agricultural data is not always,
with landholders, machinery manufacturers and service providers all potentially claiming stakes.
Data rights may be formally defined in terms of use or legal contracts between the land/sample
owner and the data generator (though individuals may often be unaware that they have ceded
control).
Figure 1: the data ecosystem
The underpinning economics of data provision varies depending upon factors such as the costs of
generating data, the economies of scale and scope in aggregation and exchange, and the benefits
of aggregating data across dimensions of space, time and type. For example, the transaction costs
Cultivating Trust: Towards an Australian Agricultural Data Market | 7
for re-appropriating data are typically high at the smaller scale individual level (i.e. due to the lack
of clarity around data rights) and so the barriers to data sharing tend to be substantial. Designing
institutions which help to develop the incentives to overcome the data lock-in among some types
of data providers will be essential to realising the potential benefits of the data market.
Service providers are a type of data demander, they demand data from data providers because data
is an input into their value-adding production of services. In the agricultural context, service
providers could include research agencies (both private and publicly funded), agronomic and
agribusiness information groups, as well as government agencies. Service providers can reward data
providers for their data by providing services (i.e. service-for-data) or by paying (dollars-for-data).
In many digital platforms data providers are often rewarded with services, for example, users of
Google Maps provide data on their location and movements in exchange for map and navigation
services. For many service providers, a key part of their business/operational model involves
drawing in data providers through some reward mechanism, developing services on that data, and
providing (possibly selling) the services on to third-party groups (i.e. data/service users).
The data/service users are also data demanders, seeking to access data directly or indirectly through
using services which draw on it. These data/service users may be firmly within the agricultural
sector, for example, a landholder purchasing crop or weather forecast services. Alternatively, users
may appear on the periphery of the sector, for example, a commodity futures market trader using
regional-scale soil moisture data to project future yields and possible prices. Collectively the
data/service users represent the final consumers in the data supply chain, their preferences and
capacity to pay for data/services will determine the scale of rewards which find their way back to
data providers. The kind of value signals data/service users provide through the data market is
important for providers of data and services because it directly informs further developments of
new data sources and service offerings.
2.2 Platforms as data market environments
Platforms are intermediaries which facilitate the formation and function of a market by bringing
together buyers and sellers. In pre-digital times market facilitating platforms were the physical
environments that supported the structured meeting of buyers and sellers, such as trading and
auction houses, organised marketplaces and shopping centres. In the digital age platforms have
proliferated as the environments that facilitate the algorithmic (i.e. refined search) matching of
those who have with those who want. An increasing range of goods and services are now exchanged
through digital platforms, which have greatly reduced the transaction costs of finding and making
trades and so opened up markets such as retail (e.g. eBay) and accommodation (e.g. Airbnb) to a
far greater range of providers. There is potential for digital platforms to transform the existing
agricultural data ecosystem into a more formalised marketplace, facilitating participation of many
more data providers and users.
Three key elements underpin the economics of digital platforms which are important for
establishing data markets. First, the value return to platform participants is characterised by positive
network effects, and these effects typically grow at a faster rate than the growth in participants.
Second, platforms facilitate significant scale economies, which greatly reduce transaction costs for
Cultivating Trust: Towards an Australian Agricultural Data Market | 8
participants by centralising operational costs. Finally, platforms are institutions which cultivate trust
among participants, thereby encouraging strangers to engage in mutually beneficial exchange.
Networks are the distributed relationships through platform markets, which form around particular
goods such as data (Bansler and Havn 2004). These relationships are important, because the
stronger the interactions around particular types of data the greater the potential to generate value
for both providers and demanders. These are often called positive network externalities, or
demand-side economies of scale. The success of digital platforms, such as Facebook or eBay, has
been strongly determined by generating and harnessing positive network externalities. The more
users they have, the more useful they are, both to existing and new users.
Interactions among users also stimulate additional value, for example, through enhanced demand
for product innovations and as the drivers of their subsequent diffusion and adoption (Cabral 1990).
For data markets, harnessing possibilities for new combinations and applications of data to develop
new knowledge or new pathways for product innovation suggests that the connectedness of data
is strongly associated with value creation. Sharing can increase the value of data, while also reducing
its marginal costs (Gans 2013). The facilitation of feedback loops among platform participants may
also provide clear pricing signals to potential data providers to invest in new or expanded data
generation and aggregation.
There are some market design considerations which may be relevant to the cultivation and
harnessing of positive network externalities. Among these is the notion that achieving a critical mass
of market participants is important for ensuring the market is functional and can grow over time.
For example, the value of eBay lies in the substantial numbers of users, so that buyers and sellers of
both common and obscure products can be profitably matched. Critical mass can be achieved by
encouraging or inducing participants to join. Some platforms subsidise the engagement of particular
types of participants who are most likely to contribute to the positive network externalities. For
example, eBay charges sellers but not buyers, thereby facilitating a greater mass of potential buyers
for participants who are selling their goods. In data markets, the opposite might be the case, with
data providers more likely to contribute to the positive network externalities. In other words, the
data demanders might be better placed to finance the platform in order to encourage as many data
providers as possible to engage with the market.
Importantly for establishing data markets, platforms are particularly good at minimising transaction
costs for users, which encourages broad-based market participation and thereby maximises the
scope to match data providers with data users. While platforms have large upfront establishment
costs (i.e. infrastructure, software, building participant critical mass, etc.) the marginal costs of
exchange facilitation are often very low. Transaction costs can also be kept low with particular
design choices at the establishment of the platform, such as the creation of standard forms of
exchange agreements and fine-grained participant controls on the access to, and use of, their data.
An analogue can be found in the Airbnb platform which supports the exchange of diverse and non-
standard accommodation services. Exchange in markets of this kind can be transaction cost heavy,
with negotiation and contracting costs dominant. This has been overcome in the Airbnb platform
by employing standardised negotiation and contracting mechanisms. This has enabled participants
on Airbnb’s platform to buy/sell accommodation services at lower prices than comparable
formalised markets and facilitate a greater diversity in the provision of those services.
Cultivating Trust: Towards an Australian Agricultural Data Market | 9
The development of mechanisms which cultivate trust among participants has been a significant
innovation of digital platforms, where the vast majority of exchange occurs between strangers. Trust
among participants is an important feature of markets where the quality of goods or intentions of
participants is difficult to assess. In lieu of alternative mechanisms, traditional markets have relied
on the strength of personal relationships among participants to mitigate potential uncertainties in
quality or intensions. In digital markets, these kind of relationships have been difficult to establish
as buyers and sellers are widely distributed both geographically and culturally. The key insight of
eBay and Airbnb platforms has been to explicitly embed high visibility trust mechanisms, such as
reputation and quality metrics, so that participants can assess the trustworthiness of a potential
match. Importantly, participants have been shown to be very sensitive to adverse reputation and
quality rankings with the disreputable quickly driven from platforms (Standifrid 2001; Resnick et al.
2006; Mickey 2010). Given the diversity of both data providers, demanders and data sources,
platforms for data markets will likely require similar forms of platform-based institutional trust
facilitation to prosper.
Cultivating Trust: Towards an Australian Agricultural Data Market | 10
3 Agricultural data as a public vs private good
In economic terms a public good is something which is non-rival (i.e. one person’s use of it does not
diminish another’s use) and non-excludable (i.e. it is infeasible to prevent all who want it from
having access). Private goods are the converse of public goods, as they are rival (e.g. cannot be
consumed twice) and excludable (it is feasible to prevent access). These properties are not absolute
– different goods and services vary in their degree of rivalry and excludability. These distinctions
matter, because the defining properties of data provide clues as to how we ought to ideally manage
the sharing (or non-sharing) of data for the benefit of society.
When shared digitally, data has public good properties, as it is not depleted by repeated use, and it
can be challenging to prevent people from accessing it. This has resulted in much of the discourse
around the economics of data being predicated on the assumption that it is, or should be, a public
good. However, this is not always helpful, because while data can be repeatedly used at no
additional cost, the underlying value of data may be rival (Gans 2013). For example, the information
used by insider traders is only valuable is only valuable because others do not have it.
While some types of data such as weather observations or scientific data have clear public good
properties, many others do not. For example, data generated by farmers in the course of production
could conceivably be used against them, for example by suppliers or purchasers seeking to negotiate
a better price. Privacy concerns can also be interpreted as an expression of rivalry. Privacy is a
challenging concept in economics (and beyond) as it combines tangible (e.g. information
asymmetries can provide an advantage in negotiations) and intangible (people feel discomfort
about personal information being revealed); it can be both a means to an end and an end in itself
(Acquisti et al. 2016).
Similarly, while excludability can be difficult (though far from impossible) to maintain for published
data, it can be maintained completely by not sharing data in the first place. It may at first glance
appear optimal to compel sharing as much as possible in order to maximise the public value of data.
However, public goods tend to be undersupplied in the absence of incentives to create and share
them. Such communal approaches are seldom effective for other goods and services, so we should
not expect data to be any different.
Excludability is imperfect – digital data can be readily copied and shared with unauthorised users,
as the music industry knows all too well – but there are a range of technical and legal defences. A
more meaningful designation is to characterise data in terms of practical exclusivity, for instance,
how many users can reasonably claim an access-right (i.e. a direct or indirect ownership stake) for
some data collection (Figure 2). Where many individuals can claim a stake in the data there is less
to be gained from attempting to exclude others. This property alone is sufficient to define regular
public goods, which is given by the cluster at the top of Figure 2, of which pure public goods are a
subset.
Cultivating Trust: Towards an Australian Agricultural Data Market | 11
Figure 2: Data exclusivity, sharing disbenefit and possible access models
We can identify examples of public goods in government sponsored or generated data (i.e.
government transactions and statistics, publicly funded research, etc.), which has a significant
number of owners. Under the conventional definition above, it can be described as a pure public
good if there also exists the property of non-rivalry. In our data context, this is the same as saying
the sharing of data is associated with very low or zero disbenefit. This is likely to be the case with
non-sensitive government data (i.e. a pure public good), but unlikely to be the case with sensitive
data such as records of individuals or security details. The extent to which this kind of sensitive data
must be withheld suggests that the more others access the data the greater the resulting cost (aka
disbenefit). In other words, it is displaying some degree of rivalry. We can still call this kind of data
a public good, but it is no longer a pure public good which has implications for its optimal
management.
The status of being a public good (i.e. non-sensitive data, Point A in Figure 2) means that its value
to society is maximised where data is treated as open-by-default (World Bank 2014; Houghton and
Gruen 2014) and disseminated under open-access to anyone. For example, weather observations
data from the Bureau of Metrology is associated with very low sharing disbenefits from the
perspective of the Bureau, and is provided in an open-access format. This is not the case for all
publicly held data (i.e. sensitive government data, Point B in Figure 2), for which optimal
management will involve some degree of withholding, even if for a defined period. For example,
public health records at the aggregate level are a type of data in the public domain, but for which
there exist potential disbenefits of sharing. The ability to re-identify individuals means there are
reasons to control access, for instance, through licence limited public-access or authenticated
group-based access.
Cultivating Trust: Towards an Australian Agricultural Data Market | 12
When data is characterised by a high degree of exclusivity (i.e. one or very few owners with access
rights), we are dealing with private goods. Like public goods, these can span the range from high to
low or zero sharing disbenefits (i.e. from rival to non-rival), with a subset defined as pure private
goods which are both highly exclusive and associated with high sharing disbenefits (i.e. Point C in
Figure 2). Most data generated by individuals is a private good, for example, soil moisture sensor
data generated by a private landholder. If the data provider identifies high sharing disbenefits
related to underlying concerns over privacy, then the sensor data essentially a pure private good.
For example, the landholder may be concerned about financial institutions accessing data from the
in-field soil moisture sensor and using that data to determine access to credit. Conversely, if privacy
concerns are non-existent then we can probably say that the data has low or zero sharing disbenefits
(i.e. Point D in Figure 2). For example, data on a farmer’s crop choice and planting area for this
coming season may well have low sharing disbenefits.
In many cases sharing disbenefits will have an objective basis, such as with data which is
commercial-in-confidence, the release of which has tangible costs to an organisation or individual.
In practice, there will be a degree of subjectivity in a data owner’s assessment of sharing disbenefit,
particularly where privacy is concerned. In some cases, sharing disbenefits could be a behaviourally
variable, rather than a fixed, property of the data. For example, some individuals are willing to share
seemingly private data (i.e. holiday photos) over social media, whereas others will seek to vigorously
prevent their release. This suggests that there may be an important role for designing institutional
arrangements which can positively influence the subjective determinants of sharing disbenefits.
The optimal sharing model is highly unlikely to correspond to open or public access, as private data
providers will require some form of incentive (i.e. payment, recognition, etc.) for sharing their data.
The need for incentives suggests sharing models using contract establishment for internal or named-
access (Figure 2). The failure to recognise the need for rewards to exchange data of this type usually
result in its under-provision. That is, data providers will withhold their data from those who could
otherwise have a value and use for it. In cases where soil moisture data is assessed to have low or
zero sharing disbenefits (i.e. it doesn’t matter if everyone has access to the data) the required
rewards may be small, perhaps only enough to cover the costs of sharing. When sharing disbenefits
exist, the size of the reward required to induce sharing are likely to be more substantial and
contracting conditions more stringent.
An interesting feature of a model of good-types presented in Figure 2, is the associated proposition
that efficient provision models ought to occur approximately along the diagonal between pure
private and pure public goods (Randall 1983). The social welfare maximising approach indicates that
largely non-rival (low or zero sharing disbenefit) goods are most efficiently provided as public goods.
In our present context, this suggests that society’s wellbeing is maximised where private owners of
low sharing disbenefit data (Point D) are incentivised (e.g. subsidised by the government) to provide
their data as if it were a public good (Point A), i.e. under an open-licence to anyone. Following the
same logic, strongly rival goods (i.e. data with high sharing disbenefit) are most efficiently provided
as if they are private goods. For this to occur, there is a role for the government to establish and
support institutions which help to protect private data holder’s rights, for example, through legal
mechanism like the Privacy Act. In the absence of these legal protections, data with high sharing
disbenefits could be obtained and widely disseminated without the explicit consent of the data
owner. This has the effect of creating a strong disincentive to generate the data in the first instance,
meaning that the data will be under-supplied and society’s wellbeing could be diminished.
Cultivating Trust: Towards an Australian Agricultural Data Market | 13
Therefore, society’s wellbeing is maximised where government supported institutions enable data
which may otherwise be at Point E (Figure 2) to be generated and managed safely as if it is a private
good (i.e. Point C).
3.1 Data clubs, cooperatives and collectives
Somewhere in between public and private goods are club goods, which are characterised by an
intermediate level of access exclusivity (i.e. somewhere between several and several thousand
access rights holders) (Figure 2). Within this category, clubs and cooperatives emerge as
mechanisms to share the burden of some costly feature which generates a stream of benefits to its
members (Buchanan 1965). In the context of data cooperatives, these costly features could include
the investment cost of data storage, as well as the transactions costs of engaging with data markets
or service providers. The presence of scale economies can enhance the stream of member benefits
which may not otherwise be obtainable acting as individuals. Digital technology both reduces the
transaction costs of data sharing (e.g. it can now simply be emailed or posted online) and increases
the scale economies of shared Information Communication and Technology (ICT) infrastructure
(servers, cybersecurity, etc), dramatically enhancing the potential for efficient sharing models.
In cases where data sharing disbenefits are low, cooperative institutional mechanisms are well
suited to organising and distributing the benefits from pooling members’ data. Sharing of data
within the cooperative can enhance the value of each individual member’s data, as well as possibly
allowing non-data owning members to benefit from the data of others. As sharing disbenefits
become high (i.e. Point F in Figure 2), there is a greater requirement for cooperatives to manage the
complexities of distributing benefits to members and protecting member’s data rights. For example,
a farmer’s cooperative sharing data on successful management practices/experiences for some new
commercial-in-confidence crop variety. Sharing such data beyond the cooperative may diminish the
commercial edge that the farmers could enjoy with the new crop. In practice clubs or cooperatives
should be able to function with these conditions, although there will be substantially higher costs
of forming and maintaining the appropriate institutional arrangements. Occupying the middle
ground between private and public goods also means that a wide range of access arrangements are
possible, albeit that contracted internal or named-access types would tend towards the more data
collections where sharing disbenefits were higher.
Within the agricultural data domain there are a number of cooperatives (e.g. Grower Information
Services Cooperative, Agricultural Data Coalition, The Farmers Business Network) which have been
developed to provide data management and marketing services to landholders, albeit that all of
these examples are based in the US. Member-contributed data cooperative environments like these
can substantially reduce the transactions costs (search, negotiation, etc.) involved in connecting
data providers to possible service providers (i.e. research agencies, digital agronomists, government
agencies, financial service providers, etc.), which can represent a substantial impediment to
realising the value of data. An alternative cooperative model exists whereby the data cooperatives
can generate data providing the data and associated services to members. Given the range of
possible data sources and differences in methods of collection, it seems most likely that a
cooperative could be composed of a combination of member contributed data and cooperative
generated data.
Cultivating Trust: Towards an Australian Agricultural Data Market | 14
4 The cost and value of sharing agricultural data
The sharing of data is fundamental to realising the value of data, however the scale of values realised
will be moderated by the associated costs of sharing. Understanding how value from data sharing is
realised in different contexts and how costs arise, helps us to identify opportunities to maximise the
potential for the data market to create value.
4.1 The cost of sharing data
In a simple model of costs, we can identify that among our three groups of market participants (data
providers, service providers and data/service users) some costs are specific to one group or other,
while some costs may be shifted from one group to another (Figure 3). For example, only
data/service users can experience data utilisation costs, only service providers can bear the cost for
developing services based on data inputs, and only data providers can bear the costs of generating
data. However, the incidence of transaction costs which arise from the interaction of market
participants may be shifted from one group to another. Who pays the transactions costs will depend
upon factors like market power, but also upon the design choices we make when establishing the
market.
Figure 3: Costs of data sharing
In general, transaction costs are the market engagement costs associated with any interaction with
other groups in the broader data supply chain. Specifically, we include costs arising from (1) market
search and information acquisition, (2) negotiation and contracting, and (3) policing and
enforcement of agreements. Transaction costs are the frictions in the data/service supply chain,
which may overwhelm the potential benefits of market exchange. We know for example, that when
uncertainties exist in the scale of potential benefits (i.e. in experimental and innovative applications)
that transaction costs will exert even greater friction and discourage use (Baca 2006; Quiggin 2010).
Importantly, some groups may be better placed to efficiently bear those costs by taking advantage
of economies of scale and scope. This suggests the possibility to explore options for designing
markets where the frictions exerted by transaction costs are minimised and thereby maximise the
potential for data to find its way efficiently from providers to demanders.
Economies of scale describe an inverse relationship between quantity of a good/service produced
and its marginal cost of provision, that is, average costs decline as we provide more. Scale economies
Cultivating Trust: Towards an Australian Agricultural Data Market | 15
are particularly evident where there are large upfront costs relative to ongoing production costs.
For example, scale economies will be present in large investments for establishing sensors, database
infrastructure and service layers (i.e. algorithms, apps, etc) with low ongoing costs of operation. The
more data that are stored in and pass through the infrastructure, the lower the average cost
becomes.
Economies of scope occur where average costs are reduced by providing a greater diversity of
products through a single infrastructure, rather than providing each separately (Rosen 1983). For
example, when we can apply the same reproducible services to a large range of data collections,
then the cost of developing those services can be shared and so the average cost per data collection
will decrease. Economies of scope are key drivers in reducing the average costs of data generation
and analysis, which partly explain the large data-service conglomerates (i.e. Facebook, Google, etc.)
that have emerged in recent years (Varian 2014).
Some economic provision models exist to identify and take advantage of service provision scale and
scope economies, such as cooperatives and clubs. Within cooperatives a typical offering may involve
the provision of a centralized data repository, some basic analytics and data market services, each
of which entail the possibility of scale economies. Data cooperatives are well placed to enjoy these
economies of scale and scope because of the repetitious nature of market engagements with data
demanders. Scale economies are certainly likely to exist for transaction costs, for example,
contracting and negotiation will be broadly similar for broadly similar types of data, as will
monitoring and enforcement processes. Search and information costs are strongly characterised by
scale economies, because when one member of the cooperative has acquired market relevant
information, all members of the cooperative will now have access to the same information. These
are all scale economies that a small-scale data provider engaging in the market as an individual is
unlikely to be able to enjoy.
At a broader level, scale and scope economies can be enjoyed by tackling industry-level frictions,
like data reporting and management standards. For example, Box et al. (2015) have identified that
with multiple geospatial data providers investments in data standards substantially reduce
transaction costs. Where providers apply agreed data standards and bear the associated costs the
benefits are experienced throughout the data supply chain. In the absence of standards, data
demanders bear the (possibly significant) costs of having to wrangle alternative data reporting
formats before use. Frictions through the supply chain along these lines have the effect of reducing
the use of, and potential to realise value from, data.
4.2 The values of shared data
Data in agricultural settings is typically generated in discrete units, for example, individual
observations or streams of observations from a single sensor. A further complexity of valuing data
is that in some circumstances the marginal value of each unit will depend on how much data is
already available. This property is not unique to data. Many goods and services are characterised by
declining marginal values – your first morning coffee may be great, a second one probably less so,
and by the fourth or fifth they are probably worthless to you. In other (rarer) cases, marginal values
may increase – a collector may prize each additional item more highly as their collection grows. For
some other things, the values are simply additive and independent of the quantity already held.
Cultivating Trust: Towards an Australian Agricultural Data Market | 16
These relationships may be important to data sharing. Consider a setting in which a landholder
installs sensors which gather environmental data. Those sensors which generate definitive and
accurate data are most likely to display declining marginal returns from additional data sources. For
example, given maximum daily temperature readings from one sensor, there is little or no additional
value of combining this with a neighbouring sensor’s reading. If the initial sensor is accurate,
additional sensor data confirms what is already known. Given the nature of the phenomena,
temperature is relatively constant across a locality and there is little to be gained from multiple
readings.
The consequence is that for data of this type, a single provider model would minimise the costs and
secure the maximum benefits (Figure 4). A private data provider would need to be incentivised to
share their data, and in the presence of low or zero sharing disbenefits there is a case for public
subsidy to induce open-access sharing. A public agency, such as the Bureau of Meteorology BoM,
may be better placed to act as the single data provider where the costs of sharing data are high.
Figure 4: Data with low spatial heterogeneity and low measurement error
Where there is greater spatial heterogeneity, interdependencies and/or measurement error
additional data will be more valuable (Figure 5). Some phenomena, such as rainfall, will vary at a
decision-relevant local scale and more data sources will provide more detailed information.
Likewise, where data sources are associated with high measurement error or require ongoing
calibration, neighbouring data sources help to identify erroneous data. In this context, sharing data
between generation sources can provide more accurate and detailed information for each
participant, so the initial marginal value of additional data is likely to be high. However, this
additional value will likely diminish with further data sources once there are sufficient readings to
provide an optimal assessment at the relevant scale.
Cultivating Trust: Towards an Australian Agricultural Data Market | 17
Figure 5: Data is heterogeneous and measurement is uncertain.
As it is most useful for one data provider to combine with the data of others, this may well
correspond to properties of club/cooperative goods. For example, this could be achieved in a data
cooperative structure where one landholder’s benefit from their soil moisture data is enhanced by
combining it with other landholders’ data to achieve a mutually beneficial outcome. The benefits of
aggregation may be realised additively, where data of the same type are combined, or
synergistically, where data differing along temporal, spatial or phenomenological dimensions are
combined. The scope to realise the synergistic benefits of data aggregation is a key economic
argument for the benefits of data cooperative formation.
There is a risk that some may wish to free-ride, relying on their neighbours’ data and not actively
contributing to the collective data pool. The benefits of this would depend on how locally specific
the data are, and how costly it is to collect. The cooperative may need to rely on member
subscriptions for access to cover its costs. Other data demanders might also benefit from access to
the data, for example, downstream water managers and fire authorities with an interest in
understanding of catchment moisture levels. Data demanders like these may be able to provide
monetary or other incentives to support the activities of the cooperative, or the activities of the
data providers. Importantly, the individual benefits of cooperatively sharing data should be
sufficient in the first instance to sustain the cooperative, through incentivising participation, while
more general sharing with data demanders represents the realisation of collateral benefits.
Finally, there are data which are valuable at both the individual and aggregate level, but for which
value differs along lines of spatial scale (Figure 5). This could include data which are valued by
individuals, such as yield forecast data, but enjoy little additional value from sharing and combining
with the forecast data of others. In this case, the optimal individual quantity of data is likely to be
low compared to the socially optimal level. At the much larger scale of society, aggregated yield
forecasts can help predict prices, and guide better investment decisions about costly inputs to
production. The socially optimal quantity of data is likely to exceed that of an individual, and possibly
by many orders of magnitude. If the individual costs of data collection are low (as may be the case
if it is collected as a by-product of ongoing management) then sharing may still be optimal. However,
the presence of transaction costs will easily overshadow remaining sharing benefits, which suggests
an important role for both clear incentives and transaction cost minimising institutional structures.
Cultivating Trust: Towards an Australian Agricultural Data Market | 18
The extent to which there may be public good properties of such data in aggregate suggests the
possibility of a role for a government aggregator, although large scale cooperatives may also be able
to achieve socially optimal scale.
Cultivating Trust: Towards an Australian Agricultural Data Market | 19
5 Institutions and incentives
The extent to which data providers, service providers and data/service users will interact to
exchange data is largely determined by the institutional environment, provided by the
organisations, rules and norms within which they operate. In a market where we have private data
providers, the institutional arrangements can include definitions of the nature of property rights
over their data. Importantly, they also define the kinds of incentives which are available to induce
the subsequent exchange of data with service providers and secondary users. Taking an institutional
perspective on data sharing recognises that alternative sharing arrangements will present different
patterns of costs, benefits and risks and so will directly affect sharing behaviours.
5.1 Understanding the needs of data providers
As data generation has a value to its generator, the existence of data doesn’t initially depend upon
the presence or absence of a market. In this sense, data isn’t a commodity good like wheat or oil
which are specifically produced to satisfy the demands of users, but rather a good for which a large
portion of its supply to the market is incidental to its production (though the existence of a market
may incentivise enhanced data generation). For example, a landholder generates data on soil
moisture to satisfy their own information needs for decision-making, but its potential to be
exchanged subsequently through a market may simply be a bonus. This suggests that those who
have generated the data and may act as data providers are the key drivers of data markets. In this
view, understanding the needs of data providers and making market design considerations which
take these into account may provide the best chance of establishing a functional data market.
To help understand the market-design requirements of data providers it is useful to consider some
archetypal models of the organisations which could engage with the market. We have identified
four significant models of engagement for the exchange of data, all of which may co-exist in the
same market, and which have particular strengths and weaknesses depending on context (Table 1).
These are: (1) the individual data provider, (2) the cooperative aggregating data provider, (3) the
corporate aggregating data provider, and (4) the public data provider.
The individual data provider model involves the individual managing all of the stages of exchange to
downstream data demanders (i.e. service providers or data/service users). For example, this could
describe a landholder who owns a sensor, or several thousand sensors, directly negotiating the sale
of data to demanders. The volume (quantity) of the data will play a significant role in determining
whether provision scale economies are likely to be present, particularly in relation to transaction
costs. For the majority of farm-scale data providers scale economies will remain elusive, but for
some large multi-property agricultural companies they are likely to be present. The key benefit to
this model of market engagement is in maintaining a high-degree of control (i.e. requirement to
control use on a case-by-case basis due to privacy legislation or research ethics requirements),
through specific contracting, as to how and who is able to access and use the generated data. In
cases where there are very high disbenefits of sharing the model might be ideal, albeit that
substantial incentives may well be required.
Among the cooperative aggregating data provider models there are two distinct groups, the first
based on member-contributed data and the second based on cooperative generated data. In the
Cultivating Trust: Towards an Australian Agricultural Data Market | 20
first case, members are data generators who use the cooperative as a mechanism to share data with
each other, and possibly exchange data within the broader data market. The benefits of cooperative
data sharing may be amplified in cases where data is heterogeneous and there exists some degree
of individual measurement uncertainty (for example, Figure 5). In the second, the cooperative itself
generates the data with members enjoying the benefits of accessing the resulting data. This might
be a common model where the costs of establishing a sensor or collecting data are very high – the
costs can be spread over the cooperative’s membership. In both cases, the cooperative takes on the
role of engaging with upstream data demanders and data aggregating intermediaries. The larger
volume of data moving through cooperatives will likely reveal provision scale economies,
particularly because average transactions costs could diminish with higher volumes of exchange.
Cooperative data club models of market engagement are likely well suited to sorts of data generated
at the farm-scale, particularly because relatively small data contributions can be managed at lower
cost by the cooperative than by an individual. A potential drawback however, may be in the agreed
set of principles which guides the activities of the cooperative. While many members will have
strongly aligned attitudes over how their contributed data might be managed, shared and possibly
exchanged with data demanders, other members may not. In this sense, a cooperative may be
better suited to managing data which doesn’t have high individual disbenefits of sharing, or is able
to diminish individual disbenefits through aggregation.
The public data provider models also fit into two categories, one based on the acquisition and
aggregation of privately held data for distribution to the public and another acting as a data
generator on behalf of the public. In the first case, a government agency might act as an aggregating
purchaser from individual or cooperative data providers, and subsequently provide the data in an
open-access form. This approach may be well suited where there exist public good benefits of large-
scale aggregation (for example Figure 5), but individual data providers experience little private
benefit in achieving that aggregation themselves.
Public data generation may be most relevant in situations where there are significant economies of
scale, for example where significant initial investment is required (e.g. developing a climate model,
building a telescope) and the resulting data are non-rival. Government data generators can cover
their costs through general taxation, allowing data to be shared openly and equitably, and avoiding
the need for further transaction costs. However, this is no guarantee of efficiency, and there is no
market feedback to guide optimal investment or innovation.
Collectively, understanding where different types of data may be more or less suited to particular
market engagement models is useful, because it allows us to identify where such models might be
developed or adjusted. Some types of market engagement mechanism can be a challenge to
develop, such as cooperatives, and are unlikely to appear without some supportive intervention.
Likewise, government focussed market engagement models require political support and a carefully
designed approach to gathering and disseminating data.
Table 1: Data provider models
Benefits
Drawbacks
Design Requirements
Archetype
Individual
data
generator
Individuals maintain control; approve use on
a case-by-case basis.
Incentives: Social, Financial (data products)
Potentially low data volumes due to
high transaction costs.
Problematic for individual to
establish data value.
Costs: Transaction costs relatively
high per exchange; limited
opportunities for economies of
scale in provision..
Continuum from case-by-case
negotiation to pre-approved sharing.
Mitigate transaction costs, e.g. by
efficient matching and contracting.
Need a good platform and institutional
facilitation; requirement for market
thickness.
Data bazaar (ad hoc sharing)
i.e. eBay-for-data.
Single organisation data
provider i.e. CSIRO.
Data characteristics: rival
(predominantly) and additive
data value.
Cooperative
aggregator
Facilitates value-adding agglomeration.
Potential for strong network externalities.
Incentives: Social, Data Complementarity
Requires a clearly defined
community with common sense of
purpose; coordinating diversity of
club members objectives/ambitions
can be an ongoing challenge.
Costs: Moderate transaction costs;
potentially high costs of
establishing cooperative from
scratch.
Gather and share (internally) data at
minimal cost to members; share
(externally) with authenticated users.
Initial requirement for support to
develop institutions; consider off-the-
shelf models.
Privacy maintenance (e.g. aggregation
or anonymization).
Ag Data Cooperatives, e.g.:
Grower Information Services
Cooperative, Agricultural Data
Coalition.
Data characteristics: non-rival
(generally within cooperative)
and data value comes from
aggregation complementarity.
Corporate
aggregator
Efficient (competitive market), innovation-
focussed.
Cultivates network externalities to sustain
revenue streams.
Transaction costs largely internalised.
Incentives: Financial (data products and
services)
No control over data once in private
ecosystem; possibility of emerging
monopoly power.
Costs: High infrastructure costs, low
marginal transaction costs
Gather data at minimal cost to
providers; share with authenticated
users.
Needs to develop market scale to
achieve provision scale economies.
Privacy maintenance (e.g. aggregation
or anonymization).
The Farmers Business
Network, Google, Facebook,
etc.
Data characteristics: rival
(potentially) with possibilities
for both additive and
complementary data value.
Public data
generator
Efficient, Public control (trust)
Can provide substantial scale economies and
minimise transaction costs.
Incentives: Votes, Public Good
Limited incentives for innovation or
efficient level of supply.
Principal-agent issues.
Costs: High infrastructure costs.
Open access or user pays, depending
on transaction costs.
BoM, ABS, GA, etc.
Data characteristics: non-rival
(generally) with possibilities
for both additive and
complementary data value.
5.2 Incentives for data sharing
Incentives are the motivating factors which drive data providers to share data, the absence of which
generally explain observed inhibitions. Incentives can take any form which represents positive value
to the individual in question. For example, an incentive could be a use value which is something
which is valuable directly or indirectly right now, or at any time in the future. Incentives of this type
could take the form of direct monetary payment (dollars-for-data) or payment by way of service
(service-for-data). Alternatively, an incentive could be a less tangible non-use value which reflects
good-will motives of contributing to the social good or bequeathing data to future generations
(karma-for-data). An example of the latter in the context of agricultural data could include donation
of soil moisture data to the rural fire service to assist in fire mitigation planning. Pro-sharing
behaviours are most likely driven by various combinations of incentives, for example, the
benevolent gifting of data by providers to public research organisations. This reflects incentives
associated with social-good motives, but may include the incentive associated with the possible
benefit of subsequent scientific discovery.
Collectively, an expression of the minimum value of required incentives (of any form) to motivate
an individual to share data is summarised by their willingness to accept (WTA). Willingness to accept
will be equal to the opportunity cost of sharing the data, which includes all the relevant costs and
risks of sharing the data plus any foregone benefit associated with its sharing. If the sharing exercise
is concerned with the secondary use of data, the bulk of the costs of sharing are likely to be
transaction costs associated with (1) market information and search, (2) negotiation and
contracting, and (3) policing and enforcement of agreements. Foregone benefits of sharing data are
harder to define, but we could consider concerns around privacy.
In economic terms, privacy may be understood as a desire to maintain an information asymmetry.
In negotiations, asymmetric information can have strategic value. For example, in a price
negotiation for some good a buyer might be quoted a higher price if the seller knows the buyer is
particularly desperate to acquire the good. Likewise, if an individual’s personal information becomes
public they may be permanently disadvantaged in negotiations. Hence revealing personal
information can be costly, and most individuals will have a high WTA. However, if an individual can
be assured that their data will only be used for a very specific purpose these concerns may be
assuaged – hence privacy assurances will reduce their WTA. There are also less tangible aspects of
privacy such as the discomfort of revealing personal information, but these are even more difficult
to value (Acquisti et al. 2016).
From the perspective of a data provider, the appearance of the WTA-sharing disbenefit relationship
is likely to be downward sloping in the direction of lower sharing disbenefits (Schudy and Utikal
2017) (Figure 6), which is determined by the relationship of transaction costs, risks and foregone
benefits to the perceived sharing disbenefit of the data. For example, transaction costs are likely to
steadily decrease in the direction of lower sharing disbenefits because there is a lesser requirement
for the data provider to invest time in negotiation and contracting to control how data are used. In
other words, the foregone benefits of poorly controlled subsequent use are small and possibly
negligible, so data providers are less likely to be concerned about investing in policing and enforcing
terms for sharing data. Collectively, this is likely to result in WTA values which are relatively small
for data which is assessed by the data provider to have low sharing disbenefits. On the other hand,
Cultivating Trust: Towards an Australian Agricultural Data Market | 23
careful negotiation, contracting and enforcement are required for data where sharing disbenefits
are high because there are potentially significant risks of foregone benefits to be experienced for
uncontrolled subsequent use. This is likely to result in WTA values which are increasing in the
direction of greater assessed sharing disbenefit. In short, to the extent that privacy concerns and
sharing disbenefits are related, the greater the importance of managing privacy, the higher we
would expect to see the associated WTA value of the data provider.
Figure 6: Willingness to accept (WTA) vs Willingness to Pay (WTP) subject to data sharing disbenefit
From the perspective of the data demanders (i.e. service providers, and data/service users), there
is a willingness to pay (WTP) function, which is the market-making partner function of WTA. Data
demanders have WTP functions which reflect the possible value they could derive from accessing
the collections held by data providers. Typically, this value will reflect an assessment of the collective
use values of acquiring and employing a given data collection, for which a large component is often
given by the time-savings of not having to collect the data directly. However, in instances where the
phenomena of interest is held within (or as a part of) a private property arrangement, a portion of
this value can reflect the fact that it is not possible to acquire the data by alternative means. This is
particularly evident where data is substantially private in nature and cannot otherwise be obtained
except by direct acquisition from the data provider.
In some cases, data demanders will have higher WTP for data which data providers are unwilling to
provide. For example, a landholder might have a high WTA for their land or soil quality data because
of the risk that others might use the data to adversely value their land. Data demanders might have
a higher value for the same reason – because they want to understand the value of land in particular
areas. One such representation is presented in Figure 6, here the data demanders place higher
(lower) valuations on data collections which have a relatively higher (low) sharing disbenefit to the
data provider. Similar patterns may be observed with data-for-service organisations like Facebook
or Google. The greater the level of detail about a user (i.e. the more privacy dense information) that
can be obtained through their service platforms, the greater the possibility to on-sell targeted
advertising access. This equates to a higher WTP for access to that data. In other cases, the data
demander’s WTP may be invariant to the data provider’s assessment of sharing disbenefit which
means the WTP in Figure 6 would appear as a flat line.
Cultivating Trust: Towards an Australian Agricultural Data Market | 24
5.3 Approaching markets for sharing data
Markets exist in more areas of our lives than we might initially appreciate. Any instance where WTA
and WTP exist there is the potential for a market, which under our definition need not necessarily
be mediated by money – exchanges may be driven by reciprocity or good-will. While the potential
exists, the acts of exchange may not; we need WTA and WTP to coincide for a market to result in
exchange.
Consider the bundle of service offerings from organisations like Facebook or Google (data
demanding service providers), together with their service users (data providers) they are engaged
in a data market. The provision of freely accessible services to users in exchange for user data,
represents an implicit expression of their WTP for that data. In turn, we could think that users are
continually making an implicit evaluation about their WTA services in exchange for their data. In
reality, the average user is unlikely to be keenly aware of this exchange, but this also provides us
with important signals of their underlying WTA. When these two sides of the market meet and the
WTP exceeds a given user’s WTA, then the user will engage in pro-sharing behaviours (for example,
Figure 6). In other words, when the value of services offered by Facebook exceed a user’s WTA, they
will engage with the service and share their data. Conversely, anti-sharing behaviours will be evident
where WTA exceeds WTP because the reward on offer isn’t sufficient to induce the user to share
their data.
All of this becomes more interesting when we consider that changes to institutional arrangements
will impact upon the expressions of WTA and WTP. For instance, changes in the way in which data
providers and data demanders (service providers and data/service users) approach the market will
generally have implications for the resulting patterns of sharing behaviours. In Figure 6, we have
illustrated the plausible interaction of an individual data provider with some data demander, but
we could just as easily consider the implications of the formation of a cooperative (club) of data
providers. The WTA of a cooperative would differ from that of an individual due to the likely
presence of scale economies in the underlying transactions costs. Each of the sources of transaction
costs (market search and information acquisition, negotiation and contracting, and policing and
enforcement of agreements) have elements which are likely to be characterised by economies of
scale in their provision, meaning the more of the activity is undertaken the lower the average cost
of its provision.
The implication of data cooperatives being able to take advantage of scale economies in transaction
costs is to reduce the WTA values of their data generating members (Figure 7). This changes the
division between the pro and anti-sharing behavioural regions, meaning data providers are likely to
be more willing to share data collections. Simply, this is because WTP now exceeds WTA for a greater
range of data collections.
Cultivating Trust: Towards an Australian Agricultural Data Market | 25
Figure 7: changing institutions
5.4 Balancing privacy
Concerns over privacy will arise where there exists data that a data provider is either compelled to,
or wishes to, keep private. This includes personal information as well as organisational data that
may be confidential or commercial-in-confidence. Whatever the underlying motivations of data
providers, any benefits which could be enjoyed are likely to be moderated by perceptions of privacy
risks. There are likely to be trade-offs between efficiency and value, both for individual data
providers and also the broader information market. Privacy protections are likely to increase the
willingness of data providers to participate, but may also diminish the subsequent value of the data.
There are also legal requirements. The rules which govern the rights and responsibilities of
interaction in a market in Australia are defined by a plethora of Commonwealth and State
legislation. Consider the Privacy Act 1988, which details the limits of an organisation’s ability to
collect and utilise the personal information of users
1
. This provides data providers some assurance
of the protections they are afforded, meaning some perceived data sharing risks cannot eventuate.
Compared to cases in which there are no privacy protections, this likely has the effect of lowering
the WTA of data providers, encouraging them to engage in greater levels of pro-sharing behaviours
(for example, Figure 7). However, this effect may be moderated by the possibility of lower WTP by
data demanders because the possibilities for use of the data are now legally attenuated.
Clearly, there are reasons why all data cannot or should not be shared openly and without regard
to the privacy concerns of data providers. There may be benefits of moving to more stringent privacy
rights regimes which provide stronger incentives for the sharing of data by providers. However, this
may come at the possible cost of reduced innovation, service provision and productivity which
would otherwise represent a source of value being returned to data providers. Analogues from the
management of intellectual property rights suggests that weak applications of such rights can be
1
A more stringent and data focussed counterpart exists in the EU (Regulation (EU) 2016/679 and Directive (EU) 2016/680)
Cultivating Trust: Towards an Australian Agricultural Data Market | 26
research and innovation enhancing, albeit that with increasing stringency they become innovation
inhibiting (Tabarrok 2013). While there is no prescribed method for achieving an ideal balance
among these competing positions, any resulting set of arrangements should be characterised by the
following (Productivity Commission 2016):
1. Effectiveness: providing sufficient incentive to generate new data, while maintaining the
socially/economically desirable sharing of data.
2. Efficiency: ensuring data is being generated by those who can do so at lowest cost and is
finding its way to those who can use it to generate the highest returns.
3. Adaptability: maintain sufficient certainty in the arrangements for data providers, service
provider and data/service users, whilst maintaining capacity for those arrangements to
evolve in response to social, legal and technical changes.
4. Accountability: establish transparent and open dealings among data providers, service
provider and data/service users without discouraging the generation or use of data.
Cultivating Trust: Towards an Australian Agricultural Data Market | 27
6 Towards an Australian agricultural data
market
Digital technologies, particularly the expanding numbers of sensors installed in everything from
tractors to toasters, are generating ever more data. Much of this data has value in helping
individuals and societies make better decisions. Realising these values requires data to be shared.
The optimal level of sharing depends on the nature of the data. Some, like adolescent poetry, is best
not shared at all. Data may be useful to a few others, or more widely, in which case open access
provision should yield the greatest overall benefits. Poor verse aside, a communal approach to data
would be desirable, but is in practice difficult to achieve. In the absence of incentives people are
unlikely to go to the trouble of generating and sharing the ideal amounts of data.
Furthermore, most people are wary of sharing information, particularly about themselves or their
business. This is entirely rational, as there are many circumstances in which our information can be
used against us (although it is also clear that many of us are quite inconsistent in how and when we
apply our preferences for privacy). Assurances about how personal data will, and will not, be used
require trust, both in the organisation doing what is says and in its ICT infrastructure working as it
should. Neither is a given. Increasing awareness of the value of data may also make people reluctant
to share it without the right incentive.
Making the best use of data will therefore require institutions to facilitate careful sharing and re-
use. For many goods and services markets provide the most efficient allocation mechanism. Where
data are widely dispersed the transaction costs of directly engaging with a market are likely to
exceed the benefits for most potential participants. Transaction costs can be minimised through
vertical integration along the data supply chain, for example through a single company installing,
owning and operating the data generation (e.g. sensors) and aggregation infrastructure.
Alternatively, a company might collect data from others in exchange for services, for example
providing analytics or equipment maintenance.
In agriculture, and beyond, many companies are aware the potential value of data and are seeking
to capture as much as possible. While this model supports aggregation and re-use, it concentrates
data, and hence market power, in the hands of a few. With increasing returns to scale for digital
products and services, data monopolists (which are likely to be multinational companies) could
disintermediate other service providers (e.g. agronomists) and capture much of the market.
Individual farmers would have little power in such marketplaces. Data generators such as farmers
and agronomists may be better served by alternative models such as data cooperatives in which
they retain more control over their data. Establishing cooperatives will not be straightforward as
they will require appropriate technical infrastructure and underlying social, organisational and legal
arrangements.
There is a need for further research to inform the various design elements that data cooperatives
are likely to require, which could then be tailored to each specific circumstance. For example, there
are a range of methods for protecting privacy which might reduce data generators’ inhibitions about
sharing (see O’Keefe and Rubin 2015). There may also be a role for industry bodies or public agencies
to act as data aggregators. Ultimately a combination of social and technical innovations will be
required to ensure that data is a source of empowerment, rather than marginalisation, for
Australian agriculture.
Cultivating Trust: Towards an Australian Agricultural Data Market | 28
References
Acquisti, A., Taylor, C., and Wagman, L. (2016). The economics of privacy. Journal of Economic
Literature 54, 442-492.
Baca, M.R. (2006) Barriers to innovation: intellectual property transaction costs in scientific
collaboration. Duke Law & Technology Review, 4.
Bansler, J.P., and Havn, E. (2004) Exploring the role of network effects in IT implementation,
Information Technology & People, 17, 268-285.
Box P., Simons B., Cox S., Maguire S, (2015) A Data Specification Framework for the Foundation
Spatial Data Framework. CSIRO, Australia.
Buchanan, J.M. (1965) An Economic Theory of Clubs. Economica, 32, 1-14.
Cabral, L.M.B. (1990) On the adoption of innovations with ‘network’ externalities, Mathematical
Social Sciences, 19(3), 299-308.
Gans, J. (2013). Information Wants to be Shared. Harvard Business Press.
Houghton, J., and Gruen, N. (2014) Open Research Data. Report to the Australian National Data
Service (ANDS).
Jakku, E., Taylor, B., Fleming, A., Mason, C. and Thorburn, P. (2016) Big data, trust and collaboration:
exploring the socio-technical enabling conditions for big data in the grains industry. EP164134.
CSIRO, Brisbane.
Mickey, R. (2010) The impact of a seller’s eBay reputation on price. The American Economist, 55:162-
169.
O'Keefe, C. M., and Rubin, D. B. (2015) Individual privacy versus public good: protecting
confidentiality in health research, Statistics in Medicine, 34, 3081-3103.
Poppe, K., Wolfert, S., Verdouw, C., and Renwick, A. (2015), A European Perspective on the
Economics of Big Data, Farm Policy Journal, 12, 11-19.
Productivity Commission (2016) Intellectual property arrangements. Draft Report, Canberra.
Quiggin, J. (2010) The value of public sector information for cultural institutions. Report to Gov 2.0
Taskforce.
Randall, A. (1983). The problem of market failure, Natural Resources, 23, 131-148.
Resnick, P., Zeckhauser, R., Swanson, J. and Lockwood, K. (2006) The value of reputation on eBay: a
controlled experiment. Experimental Economics, 9:79-101.
Rosen, S. (1983) Specialization and Human Capital, Journal of Labor Economics, University of Chicago
Press, 1, 43-49.
Sanderson, T., Reeson, A. and Box, P. (2017) Understanding and unlocking the value of public
research data: OzNome social architecture report. EP168075. CSIRO, Canberra.
Schudy, S., and Utikal, V. (2017) ‘You must not know about me’ – On the willingness to share person
data, Journal of Economic Behaviour and Organisation, 141, 1-13.
Cultivating Trust: Towards an Australian Agricultural Data Mark et | 29
Standifrid, S.S. (2001) Reputation and e-commerce: eBay auctions and the asymmetrical impact of
positive and negative ratings. Journal of Management, 27:279-295.
Tabarrok, A. (2013) Innovation Policy and Australia. Available at http://grattan.edu.au/wp-
content/uploads/2014/05/908_tabarrok_alliance_21.pdf.
World Bank (2014) Open Data for Economic Growth. Washington, DC. Available at
https://openknowledge.worldbank.org/handle/10986/19997.
Cultivating Trust: Towards an Australian Agricultural Data Market | 30
CONTACT US
t 1300 363 400
+61 3 9545 2176
e csiroenquiries@csiro.au
w www.csiro.au
AT CSIRO, WE DO THE
EXTRAORDINARY EVERY DAY
We innovate for tomorrow and help
improve today – for our customers, all
Australians and the world.
Our innovations contribute billions of
dollars to the Australian economy
every year. As the largest patent holder
in the nation, our vast wealth of
intellectual property has led to more
than 150 spin-off companies.
With more than 5,000 experts and a
burning desire to get things done, we are
Australia’s catalyst for innovation.
CSIRO. WE IMAGINE. WE COLLABORATE.
WE INNOVATE.
FOR FURTHER INFORMATION
Todd Sanderson
t +61 2 6216 7003
e todd.sanderson@csiro.au
w www.data61.csiro.au
Andrew Reeson
t +61 2 6216 7323
e Andrew.reeson@csiro.au
w www.data61.csiro.au
Paul Box
t +61 2 9325 3122
e paul.j.box@csiro.au
w www.csiro.au
