The Data Liberation Initiative (DLI) is a Canadian program that provides Canadian post-secondary institutions affordable access to Statistics Canada data resources. This paper describes the project and key success factors so that other agencies may determine its applicability to their own situations.
Pricing
Should microdata be sold or provided free of charge?
If production of microdata for dissemination is not part of a survey budget, NSO might have little choice but to try to recover the cost of it. One approach data producers may consider is selling microdata products. As statistics offices have become more autonomous, they are often faced with the obligation to identify and implement income-generating activities. Microdata are one of their high-value products. While availability of microdata files enhances a survey’s value, there will be costs incurred in the documentation and anonymization process.
A complete microdata dissemination service includes other costs such as a review process for release of files, licensing of data files, operating a data enclave, supporting users, and maintaining infrastructure. Generally this requires dedicated staff and infrastructure. If these are not in the agency’s budget or the responsibility cannot be assigned to an existing part of the organization, it may be difficult to provide a sustainable, quality service. Selling data is one way of making the beneficiaries of the data service bear some of these costs.
Free or for a fee?
There is no definitive answer to this question. There are many reasons to minimize the charges for accessing microdata files; that is not to say there should be no charge at all.
Free
The main argument for not charging fees is that, in many cases, NSOs are in the process of strengthening their data dissemination practices for microdata files. They fear too high a price may be a barrier to attracting users.
- Selling data substantially reduces the number of potential users and hence the real value of the data.
- In developing countries in particular, it may be an obstacle to users with the most interest in the data, such as students, local research centers, and universities.
Another argument against fees is the cost of collecting them. It also matters whether the fees revert to the NSO or a central agency; this affects staff incentives to recover fees. To work properly, the system must be efficient.
- Selling imposes an obligation of quality and service.
- Selling generates minimal income. Much of the demand is from academia, which has limited resources and the option of conducting their research elsewhere with other data.
While experience in other countries—most of which are developed—shows some degree of cost recovery is possible, it is unlikely that all additional costs of microdata dissemination can be recovered. There are indications that aggressive cost recovery discourages the use of microdata files and, in the long run, reduces a survey’s potential value.
The most desirable approach for an NSO is to have all costs covered in the survey budget, which maximizes accessibility. Such costs can be borne by sponsors as a way of maximizing survey benefits. This is especially important in countries where researchers have limited finances and producers few resources for analysis.
For a fee
It is more than likely that charging for data files generates some income. But other factors, such as the following points, must also be considered when deciding whether or not to charge:
- Does the NSO have the legal right to charge a fee for their products?
- Which costs does the NSO wish to recover?
- Are these costs identifiable, i.e., will users understand and accept them?
- Can users afford to pay? Can a user consortium be formed to recover the costs up front? This entails identifying which costs should be recovered and dividing them among user organizations.
- Can fees be collected efficiently?
- Complex files freely available on a website may result in their access by people lacking the ability to use microdata, which can lead to increased demand for support.
Another important consideration when developing a pricing strategy is aligning it with the pricing policy/philosophy for other products such as paper publications and Internet access. Most NSO websites are freely available to users as there are few, if any, incremental costs involved. The same, however, is not true for paper products. If publication pricing is based on covering the incremental costs of producing and shipping additional copies, the same principle could be applied to microdata files and the additional cost of supporting or servicing additional users.
Some countries’ experience
There is a long history of cost recovery by NSOs around the world. In response to budget cuts and pressure from central agencies, it became common in the 1980s to shift the burden of cost of statistics from taxpayer to user. While a complete review of NSOs’ cost-recovery activities is beyond the scope of this document, here are a few observations:
- Statistics New Zealand (SNZ) undertook to recover 25 percent of its total budget from the sale of products and services. This happened at a time when New Zealand underwent a rigorous administrative review. While this undertaking staved off program cuts, it was unable to reach its target, and eventually it was judged inappropriate to do so. Instead of attempting to recover costs, NSZ has undertaken a cost-reduction strategy. SNZ’s website reveals that about 90 percent of their information is freely available; charges are levied for custom tabulations and detailed tables. There are fees for access to microdata files but these are not specified on the website.
- Statistics Canada implemented a comprehensive cost-recovery program in the 1980s in response to budgetary and political pressure. It undertook this challenge positively despite severe opposition from users. The program met with uneven success and many parts of it have been eliminated or greatly reduced. When charges for microdata products were increased substantially, Canadian researchers started using data from the United States, whose data were readily available through the Interuniversity Consortium for Political and Social Research (ICPSR); the Data Liberation Initiative was born when it became apparent that this was not serving Canadian objectives. The following quote provided the impetus for the DLI: “...the genuine exercise of democracy increasingly requires that citizens get access to complex information and have the skills required to understand it.” While there are pressures on Statistics Canada to reduce costs and increase income, he feels the outcome has been the restriction of “...access to information only to groups that have the solid ability to pay.” Bernard feels that this may “...hamper the participation in public debates of groups whose contribution is not backed up by much money” as well as “those who have no prospect of turning a profit or reaping some tangible and relatively immediate benefit from using it.” This, he states, is “...likely to lead, in the long run, to suboptimal development and less than full-blown democracy.” This led to creation of a network of 74 educational institutions with access to all public data from Statistics Canada. These data include approximately 300 public use microdata files and thousands of other files, databases, and geographical files. The subscription fee covers support costs as well as development of an enhanced technical infrastructure serving subscribers as well as the agency. It is not intended to cover cost of the data. The DLI access portal is not available to government departments, many of which have agreements with Statistics Canada to share the costs of certain surveys. Generally, their access to microdata data is covered by such agreements and thus they may not need this service.
Ernie Boyko and Wendy Watkins - November 2011