Oxford CSAE Presentation Strategy

Converting a 60+ Minute Talk to 20 Minutes

Venue: Centre for the Study of African Economies, Oxford
Time Constraint: 20 minutes (likely 15 presentation + 5 Q&A, or similar)
Audience: Development economists, empirical microeconomists, Africa specialists

---

Part I: What Makes This Work Novel for CSAE

Primary Contributions (Rank by Novelty)

1. Measurement Innovation: Scheduled vs Unscheduled Absence
   • The literature conflates these (World Bank SDI, Muralidharan et al. 2017)
   • You show that 48% of “absent” teachers are at school but not timetabled
   • This is genuinely new and policy-relevant: administrative reforms ≠ behavioural interventions
   • CSAE will appreciate this distinction because it reframes the policy problem
2. Within-School Slack > Spatial Misallocation
   • The misallocation literature (Walter 2020, Fagernäs & Pelkonen 2017) focuses on between-school STR variation
   • Your finding: 18.8% ECS reduction from within-school labour slack vs 5.8 pp from spatial pooling
   • This directly challenges the dominant policy focus on redistribution between schools
3. TCR as a “Remarkably Understudied” Metric
   • Bennell (2022) laments the gap; you fill it at unprecedented scale
   • 5.4 million learners with reconstructed class-level timetables
   • First rigorous large-scale TCR analysis in primary education anywhere
4. The Causal Mechanism: Resource Slack → Expanded Slack
   • IV results showing δ_FD < 0 (more teachers → more inefficiency)
   • This explains why input-based interventions disappoint (Hanushek 1986, Glewwe & Kremer 2009)
   • Novel mechanism linking teacher hiring to persistent crowding

Secondary Contributions (Mention, Don’t Dwell)

• Nested optimisation framework (methodologically neat but not the headline)
• Fiscal leakage quantification (R22.3bn—striking but derivative of main finding)
• Provincial heterogeneity (useful for policy targeting but not academically novel)

---

Part II: Critical Assessment of Current Slides

Strengths

1. Clear conceptual framework: The STR = SCR / TCR identity is elegant and understandable
2. Compelling puzzle: “STR = 32.5, yet ECS = 43.3” is a great hook
3. Strong data: 5.4M learners is impressive; administrative data eliminates reporting bias
4. Multiple identification strategies: Demographic shocks + IV with PPN
5. Policy relevance: Clear, actionable implications

Weaknesses Requiring Attention

1. Too Many Ideas for 20 Minutes

The current deck tries to do everything: - Measurement framework - Descriptive decomposition (6 scenarios) - Variance decomposition (LMG) - Demographic shock elasticities - IV identification - Equity analysis (quintiles, race) - Fiscal costing - Policy design

Solution: Pick ONE identification strategy and go deep. The IV is stronger; demographic shocks are correlational.

2. Rhetorical Slides That May Not Land

• “Asymmetry of Endurance” (if 6-year-olds can handle it, teachers can)—this is clever but risks sounding flippant to serious economists
• “The Time Budget: Adequate Prep Exists”—this reads as advocacy, not analysis

Solution: Cut these entirely. Let the data speak. CSAE audiences are allergic to normative framing.

3. IV Section Needs Tightening

• Currently spreads across 7+ slides
• The threats to validity slide lists manipulation concerns but doesn’t provide formal tests
• “I test for the above and find that almost all learners begin in Grade 1 or R and exit in Grade 7”—this is asserted but not shown

Solution: One slide on instrument, one on specification, one on results with robustness summary. Show the balance tests.

4. Scenario Decomposition is Overkill

• Six scenarios (S0–S6) with representative school examples
• For a 20-minute talk, audiences won’t retain this granularity

Solution: Collapse to three buckets: (1) Technical efficiency (S1–S2), (2) Within-school allocation (S3–S4), (3) Spatial (S5–S6). Present as waterfall chart only.

5. Equity Analysis is Distracting

• The quintile and race heatmaps are interesting but tangential
• “Efficiency reforms alone cannot close equity gaps” is an important caveat but doesn’t advance the core argument

Solution: Relegate to backup. Mention in passing (“efficiency gains are proportional; relative gaps persist”).

6. Text Density

Several slides have 100+ words. CSAE talks should be visually driven with minimal text.

Solution: Ruthlessly cut. One claim per slide. Let figures do the work.

---

Part III: Proposed 20-Minute Structure

Recommended Slide Count: 10–12 slides (max)

#	Slide	Time	Content
1	Title	0:30	Name, affiliation, title
2	Puzzle	1:30	STR = 32.5, ECS = 43.3. Teachers employed but not deployed. One equation: STR = SCR / TCR.
3	Literature Gap	1:30	Three bullets: (1) Misallocation literature ignores within-school, (2) TCR understudied, (3) Input elasticities disappointing—we explain why
4	Data	1:00	5.4M learners, 9,512 schools, reconstructed timetables. Novel: class-level deployment from admin data.
5	Measurement Framework	2:00	Scheduled vs unscheduled absence distinction. SDI conflation. One figure: SDI breakdown.
6	Descriptive Results	2:00	Waterfall chart: S0 → S4 → S5. Key: Labour slack (−18.8%) > spatial (−5.8 pp). TCR = 1.22 nationally.
7	Identification Challenge	1:00	Why STR is endogenous: reverse causality, management quality, omitted variables.
8	IV Strategy	2:30	PPN instrument. Grade-weight variation. Show the formula. One slide.
9	First Stage & Validity	2:00	F > 60. Balance tests (show table). Pre-trends falsification.
10	IV Results	2:00	δ_FD = −4.5. Interpretation: resource pressure → reduced slack. This is the key causal finding.
11	Implications	2:00	R22.3bn fiscal leakage. Hiring absorbed into slack. Policy: deployment incentives, not fiscal expansion.
12	Conclusion	1:30	Three sentences. Main finding. Contribution. Policy relevance.

Total: ~18 minutes (leaves buffer for transitions)

---

Part IV: Specific Recommendations

Slides to CUT Entirely

1. “The Time Budget: Adequate Prep Exists” (#slide-prep-time)
2. “Asymmetry of Endurance” (#slide-endurance)
3. All quintile/race heatmaps (move to backup)
4. All individual scenario school examples (S0–S6 appendix slides)
5. “Scenario Definitions” table (#slide-scenario-definitions)
6. “What explains reducibility?” variance decomposition (#slide-25b)
7. Full alternative explanations table (#backup-12)—summarise verbally
8. Demographic shock elasticities section (slides 22–24)—you don’t have 20 mins for both identification strategies

Slides to CONDENSE

1. Literature slide: Currently too dense. Three bullets maximum.
2. IV section: Combine specification + threats into one slide.
3. Robustness: One sentence (“17 specification checks, all robust”) + backup reference.

Slides to ADD

1. Balance test visual: Show that instrument is orthogonal to pre-determined covariates. This is what CSAE will scrutinise.
2. Pre-trends graph: If you have it, show it. If not, generate it.
3. First-stage scatter: Predicted STR vs actual STR. Visual credibility.

Visual Improvements

1. Waterfall chart: Good. Keep.
2. Cleveland dot plot (provincial descriptives): Good but may be cut for time.
3. Fiscal leakage bar chart: Good. Keep but present quickly.
4. Heatmaps: Too granular for 20 minutes. Cut or simplify.

---

Part V: Anticipating CSAE Questions

Methodological Scrutiny (Prepare Thoroughly)

1. “Why is the PPN instrument valid? Schools could manipulate grade composition.”
   • Prepare: Show entry/exit patterns (almost all enter Gr 1/R, exit Gr 7)
   • Have backup slide with formal test
2. “What about time-varying management quality?”
   • Prepare: Acknowledge limitation. Emphasise within-school FD absorbs time-invariant quality. Argue time-varying management unlikely to correlate with cohort-driven PPN shifts.
3. “Your FD and FE estimates have opposite signs. Explain.”
   • Prepare: FD isolates behavioural response; FE mixes selection. Have clear 2-sentence explanation ready.
4. “What about teacher unions? Isn’t this about bargaining power?”
   • Prepare: Acknowledge political economy mechanism. Emphasise paper is about measurement and mechanisms, not union-busting.
5. “External validity? This is South Africa.”
   • Prepare: Cite Tanzania TCR = 2.5 (worse than SA). Argue SA has timetabling software, so technical efficiency is high—other SSA countries likely have both technical and allocative inefficiency.

Policy Questions

6. “What’s the actual policy lever?”
   • Prepare: Contact-time norm enforcement; district-level hiring; performance management. Be specific.
7. “Why hasn’t this been done before?”
   • Prepare: Data availability. LURITS class identifiers only recently linked. What isn’t measured can’t be managed.

---

Part VI: Presentation Style for CSAE

Do

• Present with confidence but intellectual humility
• Acknowledge limitations before they’re raised
• Use precise language (“correlates with” vs “causes”)
• Engage with questions; don’t be defensive
• Have backup slides ready but don’t volunteer them

Don’t

• Oversell policy implications
• Make normative claims about teacher effort
• Rush through identification—this is where credibility is established
• Read from slides
• Apologise for complexity

---

Part VII: Next Steps

Immediate Actions

1. Create oxford_slides.qmd with streamlined structure above
2. Generate balance test figure
3. Generate first-stage scatter plot
4. Prepare 5 backup slides only (IV robustness, equity results, fiscal assumptions, SDI comparison, secondary extension)
5. Practice to 18 minutes exactly

Quality Checks

• Run spell-check (British English)
• Verify all figures render
• Test on projector (colour contrast)
• Prepare one-page handout with key equations and results

---

Summary: The 20-Minute Pitch

“South African primary schools employ enough teachers for classes of 32, but learners experience classes of 43. I show this gap is driven by within-school labour under-utilisation—teachers are timetabled for only 82% of instructional periods. Using an IV strategy based on bureaucratic teacher allocation rules, I find that resource pressure causally reduces inefficiency: schools absorb additional teachers into lighter workloads rather than smaller classes. This represents R22 billion in annual fiscal leakage and explains why teacher hiring interventions consistently disappoint. The solution is deployment reform, not fiscal expansion.”

Word count: 89 words. Delivery time: ~40 seconds.

This is your elevator pitch. Memorise it.