Monday, January 28, 2013

Reforms are incremental and probabilistic than binary and definitive?

It frustrates me when people talk about the success or failure of reforms based on their simplified assessments of whether an ideal desired outcome has been achieved or not. Accordingly, education reforms are deemed to have failed unless learning outcomes have improved dramatically. Or structural reforms have failed if fiscal deficit is not eliminated or inflation not brought below 3-4 percent. Or social safety net reforms have failed if they have not succeeded in eliminating leakages. All these betray a cognitive bias that anchors any discussion about reforms into neatly defined solutions in our mental space.  

This world-view of reforms is based on two important assumptions. One, the reforms' outcome matrix is binary - either success or failure. Two, it is possible to achieve success through a complementary set of initiatives that constitute the reforms. I believe both these assumptions are flawed and reveal a failure to appreciate the difficulty of achieving change in complex systems.

Here is an alternative narrative, which can be applied to analyzing many transformations. Consider any reform as consisting of the dynamic interaction of two changes - technical and systemic/behavioural - in an evolutionary mode. The former involves deploying specific inputs, technology, and processes to prepare the ground for alignment of incentives with desired outcomes. I believe that though we cannot ex-ante identify all the ideally required elements of the technical change, it is possible to short-list a broad set of elements. The latter involves the response of interacting stakeholders to these technical changes. In complex social systems, this evolution happens in a highly context-specific, protracted, and non-linear manner, punctuated with multiple equilibrium, and through an iterative process of changes. As they say, two steps forward, one step backward, and another sideward!

A more relevant framework for evaluating the success of any such change maybe incremental and probabilistic than binary and definitive. In particular, I can think of three touchstones for any such reform process. One, do the components of the technical change contain the basic requirements necessary to achieve such transformations? Two, does the first iteration of reforms improve outcomes from business as usual? Three, are the elements of the transformation project cost-effective, compared to other alternatives?

Let me illustrate with the case of education. Consider a program to improve student learning outcomes. A basic requirement to achieve this objective is the presence of an adequate number of schools (to enable access) and teachers, minimum physical infrastructure, reasonably regular attendance of teachers and students, availability of adequate learning materials, a mechanism to measure student-wise learning outcomes embedded in an institutional framework that would optimize incentives of all stakeholders. To this extent, any reform program that seeks to establish schools, appoint teachers, build physical infrastructure, give text-books, nudges and forces teachers and students into attending school, use computers to capture student learning levels data and run a monitoring system (however flawed), is not only an improvement from business as usual but also form essential elements in any effort to improve learning outcomes.

How stakeholders respond to these technical changes has social and political dimensions, which are not readily amenable to the desired trajectory of change. Therefore, for example, if the prevailing political dynamics rule out the adoption of the evidently more cost-effective and incentive compatible arrangement of contract teachers instead of regular teachers, then we should go ahead with the latter, irrespective of the nature of systemic distortions. In fact, the dynamics of systemic/behavioral changes most often push the system into sub-optimal outcomes in such transformations. Fortunately, we can avoid systems remaining entrapped in these sub-optimal outcomes for long if they can embrace an iterative approach to such transformations.

An institutionalized iterative process would provide feedback from the emergent behavioral and systemic failures. This feedback could be used to re-engineer processes so as to re-align incentives and even modify the nature of technical change, wherever required and feasible. The principle behind this iterative process should be that the proposed change would increase the likelihood of movement in the direction of the ultimate objective. I believe that this whole evolutionary process can be best managed through a collaborative process of deep-dive problem solving and experimental research within public systems.


shashank rai said...

Totally agree with you. I work in the education sector with the govt. in India and there are so many problems that you see and one can only hope to correct them one by one. To look at yearly data and say that work done is useless, is highly harsh on people working with these systems with all the socio-political dynamics involved.

But I have a question. How does one treat a legal systemic failure when working in the public systems? For e.g., ever since the law was passed that no student can be failed in a class till class X, the students as well as teachers have started taking it easy and learning levels have suffered. Not only that, schools have started admitting students as per age and as per their learning levels. Specifically in public systems, how would you deal with such a problem and how would you motivate the teachers?

Anonymous said...

I agree that the incremental and probabilistic approach is more pragmatic. The field of technology has however benefited from the disruptive approach (Theory of Regulation, French school) and it should be applicable in many fields of public management as well.

Public systems - particularly the social sector - face difficulties in monitoring and measuring the success of processes (as against outcomes). Within the field of education, for instance (and I do not have much experience in this domain) how do we measure the success of a system that obviates exams until class X? Not through outcomes, but would it be possible through monitoring processes, and if so what are the tools?