1. An Investigation into the Impact of
the Medium-Term Expenditure
Framework on Budgetary Outcomes
Jim Brumby
World Bank
2. Aim of the Research
Medium Term Expenditure Frameworks (MTEFs) are
fiscal instruments aimed at enhancing fiscal
performance. This study analyzes the impact of MTEFs
on several dimensions of fiscal performance:
1. Fiscal discipline, that indicates how well spending
relates to revenues.
2. Allocative efficiency, that measures the extent to
which resources flow toward the most needed public
projects.
3. Technical efficiency, that shows the extent of waste of
resources allocated to a given project.
3. Types of MTEFs
Different types of MTEFs can be distinguished:
1. MT fiscal frameworks (MTFFs) provide a macro-fiscal
basis for budget formulation. It is likely to impact
fiscal discipline more than efficiency.
2. MT budget frameworks (MTBFs) in addition specify
spending agency expenditure ceilings based on top-
down resource availability and bottom-up resource
needs. This is likely to impact most on allocative
efficiency.
3. MT performance frameworks (MTPFs) in addition
focus on spending program inputs, outputs and
outcomes. This would impact most on technical
efficiency.
4. Cumulative Number of Countries with MTEFs at end of 3-year
period (line graph)
0
100
120
140
20
40
60
80
1990
MTEF
MTFF
1991-93
MTBF
1994-96
MTPF
1997-99
MTEF
2000-02
MTFF
2003-05
MTBF
Global MTEF Adoption
MTPF
2006-08
0
10
20
30
40
50
60
70
Number of Countries adopting MTEF over 3-year period (bar
graph)
5. Stylized Facts – Event Study MTEFs
Fiscal Discipline
3
Central Government Balance (% of
2 -0.4
1
0
GDP)
-1 t-3 t-2 t-1 t t+1 t+2 t+3
Allocative Efficiency -2
-3
80
-4 -3.1
Health Spending Volatility
70 -5
86.6
60 mtef lower bound upper bound
before ave after ave
50
40
30 94.2
20
t-3 t-2 t-1 t t+1 t+2 t+3
mtef lb_mtff_es
ub_mtff_es before ave
after ave
6. Stylized Facts – Event Study MTEFs
Fiscal Discipline
3
Central Government Balance (% of
2 -0.4
1
0
GDP)
-1 t-3 t-2 t-1 t t+1 t+2 t+3
Allocative Efficiency -2
-3
80
-4 -3.1
Health Spending Volatility
70 -5
60 mtef lower bound upper bound
50.3
before ave after ave
50
40
48.9
30
20
t-3 t-2 t-1 t t+1 t+2 t+3
mtef lb_mtff_es
ub_mtff_es before ave
after ave
7. Econometric Results - Summary Table
allocative
efficiency - health
allocative efficiency – spending as a technical efficiency – technical efficiency –
fiscal discipline – health spending share of overall health spending ‘s health spending’s
overall balance as volatility (- effect government effect on life effect on child
share of GDP means improvement) spending expectancy mortality
(1) (2) (3) (4) (5)
MTFF 0.845** -2.660*** 0.352*** 0.105 0.0778
(0.419) (0.911) (0.136) (0.125) (0.289)
MTBF 0.986* -2.948*** 0.429** 0.0725 -0.751**
(0.524) (0.966) (0.176) (0.173) (0.333)
MTPF 2.816*** -2.191 1.038*** 0.513*** 0.107
(0.959) (1.554) (0.376) (0.186) (0.411)
Notes: Robust standard errors clustered by country are in parentheses, *** p<0.01, ** p<0.05, * p<0.1. The results
reported are based on the AB System GMM estimations with year effects and IV. All the specifications include a set of
regressors and the constant term.
8. Key Messages
• MTEFs improve fiscal discipline measured as overall fiscal balance. The effect is significant
and increasing as the country moves from an MTFF to an MTPF.
• MTEFs improve allocative efficiency measured as the volatility of health expenditures to total
expenditures. As expected, the effect is larger as the country goes from an MTFF to an MTBF.
• While the volatility drops with after the introduction of the MTEFs, the spending in health
increases consistent with expectations. Results can vary with the estimator used, but always
positive when significant.
• Results on technical efficiency measured as health spending’s impact on output of the health
sector—whether life expectancy or infant mortality—are mixed. If we use life expectancy as a
measure of service delivery, results show a positive effect of the MTPF, but no effect from the
introduction of other frameworks (as expected). But when considering child mortality, the
only significant coefficients are negative and relative to MTBF.
Editor's Notes
Lines show totals numbers in place, bars show adoptions.
The graphs show the data from all countries that have adopted an MTEF in their different types (going from an MTFF to an MTBF is also counted as MTEF adoption). The time of introduction is put at t. The black line shows fiscal discipline (measured as the overall fiscal balance of the central government) and allocative efficiency (measured as the volatility in health spending as ratio of overall government spending) before the introduction of the MTEF (periods t-3 to t-1) and after introduction (t+1 to t+3). The red dotted lines show the 95 percent confidence interval. The blue dotted lines show the average for all the years before and after the introduction (time t). For fiscal discipline, this graph suggests that there is a clear improvement in the fiscal balance after MTEF introduction. The average fiscal balance of the central government was -3.1 percent of GDP before the MTEF introduction and -0.4 percent of GDP after.For allocative efficiency, the pattern is not so clear, but still suggestive of improvement after MTEF introduction. The measure of allocative efficiency is the deviations from the trend value in health spending as a percent of overall government spending, which is a proxy for health spending volatility. One would expect the introduction of MTEF to initially lead to a re-allocation among sectors, with health spending being a likely beneficiary. So, in the short term, health spending volatility may go up. Then, over the longer term, a more stable path of health spending would suggest better allocative efficiency. The chart shows that immediately after the MTEF introduction (in time t) there is not a clear change in health spending volatility, but in the longer term volatility drops, as expected (from an index of 94.2 before the MTEF introduction to 86.6 in the period after).
The graphs show the data from all countries that have adopted an MTEF in their different types (going from an MTFF to an MTBF is also counted as MTEF adoption). The time of introduction is put at t. The black line shows fiscal discipline (measured as the overall fiscal balance of the central government) and allocative efficiency (measured as the volatility in health spending as ratio of overall government spending) before the introduction of the MTEF (periods t-3 to t-1) and after introduction (t+1 to t+3). The red dotted lines show the 95 percent confidence interval. The blue dotted lines show the average for all the years before and after the introduction (time t). For fiscal discipline, this graph suggests that there is a clear improvement in the fiscal balance after MTEF introduction. The average fiscal balance of the central government was -3.1 percent of GDP before the MTEF introduction and -0.4 percent of GDP after.For allocative efficiency, the pattern is not so clear, but still suggestive of improvement after MTEF introduction. The measure of allocative efficiency is the deviations from the trend value in health spending as a percent of overall government spending, which is a proxy for health spending volatility. One would expect the introduction of MTEF to initially lead to a re-allocation among sectors, with health spending being a likely beneficiary. So, in the short term, health spending volatility may go up. Then, over the longer term, a more stable path of health spending would suggest better allocative efficiency. The chart shows that immediately after the MTEF introduction (in time t) there is not a clear change in health spending volatility, but in the longer term volatility drops, as expected (from an index of 50.3 before the MTEF introduction to 48.9 in the period after).
The results are interpreted in the next slide.Regressions were run using different estimators; we present here the theoretically most powerful specification which resolves endogeneity issues (through instrumentalization and lagged variables) and model specification issues (through using panel data with country and year effects). Among the regressors were inflation, GDP growth, trade openness, population, OECD aid flows, and dummies for IMF program, conflicts, oil exporters, and HIPC Initiative.
We ran various types of regressions using different estimators and other changes in the model to ensure our results were robust. We note here the major results from the robustness analysis (e.g., in the third bullet).We are currently revisiting the technical efficiency results—there are numerous technical challenges here that we are working through.