This presentation was made by Patrick Dunleavy, United Kingdom, at the 12th Annual Meeting on Performance and Results held at the OECD, Paris, on 24-25 November 2016
2. Six essential steps in measuring productivity in
the public sector
• Identify and count the agency’s ‘core’ outputs or activities
• Develop unit costs or activity accounting for core outputs
• Develop a ‘cost-weighted total output’ metric for each
agency
• Develop an accurate total inputs cost number for the agency
• Divide total outputs by total inputs to give a total factor
productivity (TFP) number for the organization – labour
productivity measures less useful given outsourcing
• Decide on a strategy for controlling quality issues
3. Two ways of controlling quality issues when
measuring agencies’ productivity
• For most central departments and agencies
– treat services’ core outputs as being of uniform quality over time
or across agencies, unless strong evidence suggests quality lapses
or variations
– in which case ‘conditionalize’ the data point(s) involved
• For most decentralized agencies, delivering personal services
– apply an additional quality-weighting to the ‘total outputs
weighted by unit costs’ metric
– e.g. using data on quality from sector auditors, complaints, public
opinion surveys
– choose an appropriate reweighting level
4. Citizen-facing ministries/ agencies with large
transactional loads
• Usually includes social security department/ agencies, taxation
department, national regulatory bodies – account for a high % of central
state staffing
• Individual services transactions are easily aggregated into total outputs
using cost-weighting
• Quality control is pretty standardized in these ‘machine bureaucracies’, but
allow for lapses etc. – first strategy in Quality slide above
• Large gains to be made from developing productivity paths for these
agencies, and comparing paths across countries (even in agencies do partly
different things) – easy for leaders/ managers/ stakeholders to learn
lessons
5. Large central departments delivering complex
outputs
• Examples include defence, foreign affairs
• Some progress made on getting to more final outputs e.g. ‘unit
availability days’ in defence, or end-activity data (like flying hours)
• But complex outputs mean no single metric (e.g. ‘teeth to tail ratio in
defence) works. We need baskets of metrics on outputs
• Perhaps next stage would be further developing disaggregated
measures and activity costing
• But in principle productivity paths over time and across countries
should be do-able and would be invaluable for policymakers and
managers
6. Measuring productivity in decentralized public
sector organizations
• Hospitals, schools, social care organizations, police forces, fire
services etc. all deliver services where quality is an important factor
• Ideally choose the fuller quality-weighting approach set out above,
reweighting total outputs to reflect quality at some appropriate level
• Large N of delivering agencies allows for multivariate regression
analysis (fitting a regression line and explaining outliers)
• Or a ‘data envelopment analysis’ (DEA) comparing agencies with the
same service mix on costs and efficiency
• Cross-national analyses of other large N settings can be instructive,
• And ultimately perhaps productivity paths
7. Improving productivity measures in national
statistics for large public service sectors
• Very aggregate numbers are produced at this level, e.g. for entire
public healthcare or public education systems. Strong focus only on
individually delivered services so far by NSAs (National Statistical
Authorities)
• Useful for macro-economic understanding of growth
• A lot of focus on fixing output levels (using some inputs and
sometimes outcome components also)
• Quality weighting of outputs is feasible for NSAs, but not common yet
• At this level of aggregation not much potential for lesson-drawing
that is directly helpful for leaders, managers or stakeholders
8. Conclusions and recommendations
• Some substantial progress has been made on activity costing, but
productivity as total outputs/total inputs is rarer
• Comparing TO/TI performance over time is valuable (and avoids a lot of
measurement difficulties). Comparative productivity paths for similar
central department/agency types could be very instructive in improving
performance
• Suggested priority order for seeking improvements:
1 Large central government transaction agencies;
2 Large central agencies producing complex outputs;
3 Analysis of large N networks of decentralized agencies delivering individual
services, then collective services
4 Further developing sectoral-level data in national accounts