📄 Abstract
Human resource development has reported the return on training investment for two decades and has never reported the return on the next rupee of it. Every published evaluation framework computes an average — total benefit divided by total cost — while every budget decision a learning function actually makes is marginal. This study closes that gap. Using 9,340 training completions across 41 Indian organisations between 2022 and 2026, and exploiting 78 scored eligibility thresholds that generate discontinuous variation in training dosage, it estimates a causal return function for each of three programme architectures: conventional cohort-based delivery, adaptive modular delivery, and AI-personalised delivery. The fuzzy regression discontinuity first stage is strong, with dosage rising by ₹11,840 at the cutoff. Return functions of the bounded form R(I) = A[1 − e^(−λI)] fit well, and the two parameters carry the architecture effect: the value ceiling A rises from ₹53,940 under cohort delivery to ₹159,350 under AI-personalisation, while the saturation rate λ falls from 0.069 to 0.019. At the sample mean dosage of ₹24,000 per employee per year, average returns are 1.82, 2.14 and 2.47 across the three designs — a spread of 1.36 to one that would justify only a modest reallocation. Marginal returns at the same dosage are 0.71, 1.38 and 1.94 — a spread of 2.73 to one. Average return therefore understates the difference between architectures by a factor of two, and conceals the finding that matters most: the median cohort programme is spending 26 per cent beyond its point of exhaustion, where the last rupee committed returns 71 paise. Points of exhaustion differ sharply, at ₹19,040, ₹33,420 and ₹58,180. Marginal returns are not equalised within organisations, as allocative efficiency would require; reallocating dosage alone raises total return by 5.0 per cent at unchanged budget, and shifting a fifth of participants from cohort to personalised delivery raises it by 12.3 per cent. Notably, all three architectures return between ₹3.09 and ₹3.73 on the first rupee: they differ not in whether training works but in how quickly each design exhausts what it has to teach.
🏷️ Keywords
📚 How to Cite:
Dr. N Subbukrishna Sastry, Dr. Manjula Mallya M , RESKILLING PROGRAMME DESIGN AND THE MARGINAL RETURN TO TRAINING CAPITAL , Volume 13 , Issue 9, September 2026, EPRA International Journal of Environmental Economics, Commerce and Educational Management(ECEM) , Pages: 69 - 89 ,