TAM, UTAUT or UTAUT2: which framework should be used to study technology acceptance?

TAM focuses on perceived usefulness and ease of use. UTAUT adds organisational context, facilitating conditions and moderators. UTAUT2 extends the framework to consumers with hedonic motivation, price value and habit. Select the framework before analysis from the population and decision, never from the most favourable result.

• Decision maker: which framework fits the decision?
• Practitioner: which items, groups and occasions should be collected?
• Analyst: measurement, structure, uncertainty and robustness.

A team is preparing a digital service launch. It must distinguish employees’ initial acceptance, consumers’ voluntary use and repeated use. These populations do not imply the same constructs or decisions.

Within a declared population, context and horizon, which constructs are associated with intention and self-reported use, and with what uncertainty?

Adding every construct increases respondent burden, overlap risk and analytical degrees of freedom. Choosing the framework with the highest R² after estimation turns comparison into opportunistic selection.

• Population, technology, voluntary or mandatory setting, date and horizon.
• Items from the selected scale, documented translation and identical response options.
• Intention measured before self-reported use when use is the target.

η = Bη + Γξ + ζ

η denotes endogenous constructs, ξ exogenous constructs, B relations among endogenous constructs, Γ effects of exogenous constructs on η, and ζ the residual. Moderation enters only through prespecified interaction terms; the measurement model links latent constructs to their items.

1. Declare the framework and exclusions.
2. Test the measurement model before the structural model.
3. Estimate relations, intervals and diagnostics on the declared specification.
4. Compare only prespecified models on the same observations.

In the Hong Kong UTAUT2 study of experienced mobile Internet users, 4,127 valid first-wave respondents become 2,220 at wave two and 1,512 in the final sample. Mean age is about 31 and only one technology is studied. Reported R² rises from 0.56 to 0.74 for intention and from 0.40 to 0.52 for self-reported use. The 0.18 and 0.12 differences have no published uncertainty interval; they describe that sample and specification, not universal superiority.

• Items represent constructs in the studied population.
• Compared groups share sufficiently invariant measurement.
• A cross-sectional association proves neither temporality nor causality.

Document missing data, reliability, convergent and discriminant validity, invariance, intervals, holdout or later-wave stability, and sensitivity to removed items. Unavailable remains unavailable, never zero.

A coefficient or R² describes the studied specification. Read it with uncertainty, measurement quality and scope. It does not automatically rank product initiatives.

The team may select TAM for parsimonious acceptance diagnosis, UTAUT for organisational deployment, or UTAUT2 for consumer use where pleasure, price and habit are substantively required. The final choice remains human and documented.

Do not use these frameworks alone to estimate causal effects, forecast aggregate demand, value a technology financially or replace usage research. Alternatives include experiments, longitudinal panels, choice models, qualitative research or diffusion models depending on the decision.

The reference file recalculates only the two reported R² differences, without refitting the model or fabricating individual data.

CSV · Python 3.13.7

# R 4.5.1: transform(read.csv("msc-p027-tam-utaut.csv"), delta=r2_utaut2-r2_utaut)
* SPSS 31: COMPUTE delta = r2_utaut2 - r2_utaut.
* SAS 9.4: data result; set input; delta=r2_utaut2-r2_utaut; run;

• Decision, population, technology, horizon and target behaviour.
• Selected framework, included and excluded constructs, item provenance and analysis plan.
• Results with uncertainty, diagnostics, limitations and reassessment rule.

E3 evidence: three foundational full texts verified at claim level. The comparison remains a methodological synthesis; it proves neither commercial performance nor causality.

• Davis (1989), TAM ↗
• Venkatesh et al. (2003), UTAUT ↗
• Venkatesh, Thong & Xu (2012), UTAUT2 ↗