Digital Transformation Measurement Framework of Agribank - Vietnam

This thesis does two things that are usually kept apart. It builds a measurement instrument for digital transformation at Agribank - Vietnam Bank for Agriculture and Rural Development, the country’s largest state-owned commercial bank by branch count - and then it subjects that instrument to an econometric test to see whether it measures what it claims to measure. The instrument is DBI 2026 (Digital Banking Index 2026). It is organised around seven pillars and 192 KPIs, of which 67 are measured at branch level and 125 at Head Office. The seven pillars are Digital Infrastructure (HTS), Digital Technology (CNS), Digital Data (DLS), Digital Operations (VHA), Digital Culture (VHS), Digital Customer Experience (TNS), and Digital Business Performance (HQ). Each pillar is tied back to a recognised management theory and, where the law speaks to it, to a specific Vietnamese regulatory text: HTS to the Resource-Based View (Barney, 1991) and Circular 50/2024/TT-NHNN; CNS and DLS kept apart so that technology and data can be funded as separate resources; VHA, VHS, and TNS to the internal-process, learning-and-growth, and customer perspectives of the Balanced Scorecard (Kaplan and Norton, 1992); and HQ as the outcome pillar that ties capability spending back to economic value, which is now what Resolution 57-NQ/TW formally demands. Scoring runs on a 0-5 maturity scale and mixes data pulled straight from the bank’s systems with survey-style organisational assessments. The test uses branch-level panel data from Agribank’s own DBI evaluation system - 170 Type-I branches observed over four years, N = 680. I estimate three models rather than one, and the reason is that a single model would leave too many objections open. A Two-Way Fixed Effects (TWFE) regression is the baseline. A Fractional Logit with Correlated Random Effects (Papke and Wooldridge, 1996; Wooldridge, 2019) checks whether the bounded nature of a revenue-share variable changes the answer. A Hansen (1999) threshold regression checks whether the relationship bends rather than runs straight. Three results are worth stating up front. A one-point gain in the aggregate DX_Score goes with a digital revenue share roughly 8.6 percentage points higher (TWFE, p < 0.001), and the Fractional Logit marginal effect of 8.3 points lands almost on top of it - so the bound on the dependent variable is not driving the headline. Within the six capability pillars, CNS and TNS do most of the work (2.76 and 2.28 points per pillar point, both p < 0.001), while VHA is flat and behaves like a hygiene factor rather than a growth lever. The threshold sits at about 2.92, near the Maturity Level 3 boundary, with the marginal effect rising from 7.27 points below it to 9.27 above. Put together, the contribution is a measurement system that rests on management theory, fits the current Vietnamese regulatory frame (Resolution 57-NQ/TW, Circular 50/2024/TT-NHNN, Law 71/2025/QH15), and has been tested with a method that other Vietnamese commercial banks can borrow once their own KPI measurement matures.