What Evidence-Based Policymaking Doesn’t Do

When it comes to improving citizen satisfaction with the delivery of government services, evidence-based approaches have been elevated to the status of motherhood and apple pie. If the government would institute these methods, we are told, the outlook for remarkable improvement is a near certainty. Look no further than the title of the recently-released report of the Congressionally-mandated commission on this issue: “The Promise of Evidence-Based Policy Making.” The allure of the purported benefits of an evidence-based methodology has even inspired a bipartisan effort to codify the commission’s recommendations.  

It sounds reasonable enough. Who would be against strengthening the government’s capacity to have actionable evidentiary data to understand program effectiveness and improve policies and outcomes? One can imagine the use of the evidence-based methodology in almost any federal agency. An obvious application would focus on customer satisfaction with government service providers. Data about customer satisfaction might include:

• Time and number of contacts to set up an appointment
• Wait time on the phone
• Distance traveled to meet a service agent
• Wait time before engagement with a service agent upon arrival at the office
• Courtesy and expertise of the service agent
• Number of agents needed to resolve the issue
• Time for the issue to be resolved

Based on the “evidence” gathered, modifications would be implemented to improve service under the current delivery model. And that’s precisely the problem. The evidence-based approach does not focus on using this data, combined with other trends and anticipated behaviors, to inform the design of delivery models for the future. The result is a false sense of comfort about service delivery by improving programs that may actually be headed for obsolescence. What the evidence-based approach community gets right is the need to achieve the goals and objectives of government programs. But its devotion to improving existing programs misses the need to anticipate and migrate to successful program models for the future.

Fortunately, this shortfall can be addressed by shifting the focus to predictive analysis. For purposes of this discussion, predictive analysis is considered a forecasting methodology that uses data to identify patterns that can be leveraged to anticipate future behaviors, preferences and needs. Predictive analysis distinguishes itself from data-based improvement of existing programs in multiple dimensions, such as:

• Promotion of new design vs. repair of old methodologies
• Incorporation of future trends and demand vs. embedded parameters and assumptions
• Focus on proactive vs. reactive measures
• Central incorporation of emerging technology and capabilities vs. restrictions of increasingly obsolete processes.

The insights from such analyses can inform key elements in designing government programs of the future, including product emphasis, demand volume, work effort and service delivery methodology. Instead of gathering data about a dated service program like the customer satisfaction example above, predictive analysis would help the agency anticipate the customer’s preferences for how they are served going forward by understanding characteristics like:

• Mobile access (at home using secure personal technology, not at a remote office)
• On-demand service (24/7, without phone calls and appointments)
• Ease and speed of service (timely, accurate answers without multiple interactions)
• Focus (likely questions and products of interest)

The agency would use those insights to build a dynamic service model and capacity that allow it to meet the expectations of the customer of the future, as well as the present. This type of thinking would also help the agency understand in advance the likely consequences of its policies and programs, prepare for emerging levels of demand and inform its strategic planning process.

Government is such a target-rich environment it almost cries out for the use of predictive analysis. Consider the benefit to public-facing agencies with product and service offerings; defense and intelligence programs; safety, security and disaster recovery programs; government insurance, loan and other financial services; and financial payment and transactional fraud prevention.

These are only a few of the most obvious applications. Just think of the difficulties experienced by customers of government programs that might have been forestalled if a predictive methodology had been the established practice across government 10 years ago.  

The government can and should begin to establish robust predictive analysis capabilities now.  Some logical steps to get started include:

• Capturing insights of predictive analysis practitioners currently in government
• Engaging outside experts to accelerate adoption and to leverage private sector expertise
• Selecting appropriate predictive tools, techniques and supporting technology
• Choosing a small number of programs to which predictive analysis can be readily applied to demonstrate the value
• Establishing a community of experts who can rotate through agencies to implement predictive analysis programs
• Identifying advocates to increase understanding across government
• Convincing agency leaders, OMB, Congress, GAO and other stakeholders to support the use of predictive analytics

Now is the time for all government program and policy offices to start thinking about the future. They can begin by adopting these practical steps. Congress can do its part by modifying its draft legislative framework for the evidence-based approach to direct agencies to implement predictive analysis. Over a century ago, forward thinkers embraced the opportunity to focus on developing the automobile rather than wasting valuable resources on improving the horse and carriage. Let’s hope the government has as much foresight. The people it serves are depending on it.

Linda M. Springer is a former U.S. Office of Management and Budget Senior Advisor and U.S. Office of Personnel Management Director.