Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
Definition in One Paragraph
The Technology Acceptance Model (TAM) is an information systems framework developed by Fred D. Davis (1986, 1989) to explain and predict how individuals accept and adopt computing technology. Adapted from the Theory of Reasoned Action, TAM posits that usage is driven by behavioral intention, determined by two beliefs: perceived usefulness (the degree to which a user believes a system will enhance job performance) and perceived ease of use (the degree to which a user expects the system to be free of effort). Researchers employ TAM to evaluate digital systems by operationalizing variables into survey scales, estimating path coefficients, and determining whether resistance stems from functionality or usability barriers.
Formula with Every Symbol Explained
In structural equation modeling (SEM), TAM is specified through three simultaneous regression equations:
1. Perceived Usefulness: PU = β₀₁ + β₁₁ · PEOU + γ₁ · X_ext + ε₁
- PU: Endogenous latent variable for Perceived Usefulness.
- β₀₁: Intercept; β₁₁: Standardized path coefficient from PEOU to PU.
- PEOU: Exogenous latent variable for Perceived Ease of Use.
- γ₁: Coefficient vector for external design features (X_ext); ε₁: Disturbance error term.
2. Behavioral Intention: BI = β₀₂ + β₂₁ · PU + β₂₂ · PEOU + ε₂
- BI: Endogenous latent variable for Behavioral Intention to Use.
- β₀₂: Intercept; β₂₁: Path coefficient from PU to BI; β₂₂: Path coefficient from PEOU to BI; ε₂: Disturbance term.
3. Actual System Usage: Actual Use = β₀₃ + β₃₁ · BI + ε₃
- Actual Use: Quantified usage behavior; β₃₁: Path coefficient from intention to behavior; ε₃: Disturbance term.
Mediation Formula (Total Effect):
Davis (1989) demonstrated that PEOU influences intention both directly and indirectly through PU: Total Effect(PEOU → BI) = β₂₂ + (β₁₁ × β₂₁). Evaluating direct versus mediated pathways helps researchers classify types of variables and isolate whether interface refinement directly motivates usage or functions primarily by unlocking system utility.
Worked Example with Real Numbers and the Arithmetic Shown
Consider an implementation study evaluating electronic lab notebook (ELN) adoption among N = 250 laboratory scientists, applying sample size determination formulas for structural path modeling.
Variables are scored on a 1–7 Likert scale using Davis’s validated items. Sample statistics: PEOU (Mean = 4.85, SD = 1.10), PU (Mean = 5.20, SD = 1.15), BI (Mean = 5.05, SD = 1.25). Observed correlations: r(PEOU, PU) = 0.480, r(PU, BI) = 0.600, r(PEOU, BI) = 0.420.
Step 1: PEOU → PU Path and Variance Explained
In standardized bivariate regression: β₁₁ = r(PEOU, PU) = 0.480.
Variance explained: R²_PU = (0.480)² = 0.2304 (23.0%).
Step 2: Direct Paths Predicting Behavioral Intention
Using simultaneous regression for standardized variables (z_BI = β₂₁ · z_PU + β₂₂ · z_PEOU):
Denominator = 1 – [r(PEOU, PU)]² = 1 – (0.480)² = 0.7696.
β₂₁ (PU → BI) = [r(PU, BI) – r(PEOU, BI) · r(PEOU, PU)] / 0.7696 = [0.600 – (0.420 × 0.480)] / 0.7696 = 0.3984 / 0.7696 = 0.5177 ≈ 0.518.
β₂₂ (PEOU → BI) = [r(PEOU, BI) – r(PU, BI) · r(PEOU, PU)] / 0.7696 = [0.420 – (0.600 × 0.480)] / 0.7696 = 0.1320 / 0.7696 = 0.1715 ≈ 0.172.
Step 3: Variance Explained in Intention (R²_BI)
R²_BI = β₂₁ · r(PU, BI) + β₂₂ · r(PEOU, BI) = (0.5177 × 0.600) + (0.1715 × 0.420) = 0.3106 + 0.0720 = 0.3826 ≈ 0.383 (38.3%).
Step 4: Indirect and Total Effects
Indirect Effect = β₁₁ × β₂₁ = 0.480 × 0.5177 = 0.2485 ≈ 0.249.
Total Effect = Direct + Indirect = 0.1715 + 0.2485 = 0.4200 (matches r(PEOU, BI) = 0.420). The indirect effect (0.249) is substantially larger than the direct effect (0.172), confirming that ease of use matters primarily because it unlocks utility.
Interpretation (What the Result Means)
Evaluating TAM findings requires benchmarking empirical coefficients against established standards:
| Metric | Recommended Threshold | Methodological Interpretation |
|---|---|---|
| Cronbach’s α / Composite Reliability | ≥ 0.70 (ideal ≥ 0.80) | Confirms internal consistency across multi-item survey scales. |
| Average Variance Extracted (AVE) | ≥ 0.50 | Establishes convergent validity; construct explains over half of indicator variance. |
| Path Coefficient (β) | |β| ≥ 0.20, p < 0.05 | Indicates a practically meaningful, statistically significant causal relationship. |
| Variance Explained (R²_PU) | 0.20 to 0.40 | Proportion of perceived usefulness accounted for by ease of use and system design. |
| Variance Explained (R²_BI) | 0.35 to 0.55 | Substantial explanatory power for predicting technology adoption intentions in standard TAM. |
Standard TAM accounts for 35%–45% of variance in adoption intentions. Extended frameworks like TAM2 (Venkatesh & Davis, 2000) and UTAUT (Venkatesh et al., 2003) incorporate social and organizational moderators, explaining up to 70% of variance.
Assumptions and Common Mistakes
Applying TAM requires adhering to key structural principles:
Methodological Assumptions
- Linearity: Latent construct relationships are assumed to follow linear causal trajectories.
- Voluntary Context: Original TAM assumes users possess discretion over adoption; in mandatory settings, intention correlates less strongly with usage.
- Multivariate Normality: Maximum likelihood SEM assumes normality; non-normal data requires robust estimators or PLS-SEM.
- Discriminant Validity: Discriminant validity must be confirmed (HTMT ratio < 0.85) to ensure indicators do not cross-load.
Common Mistakes
- Conflating Intention with Usage: Treating intention as an uncritical surrogate for adoption without verifying sustained system usage over time.
- Common Method Bias: Measuring beliefs and intentions in a single cross-sectional survey without procedural controls (e.g., temporal separation).
- Deleting the PEOU → PU Path: Modeling PU and PEOU as independent parallel predictors, ignoring ease of use’s theoretical role as an antecedent to usefulness.
- Citation Errors: Citing Davis (1989) Management Science for scale validation, or citing Davis (1989) MIS Quarterly for the TRA comparative analysis.
How to Report It (APA Example)
When reporting TAM structural equation modeling findings, adhere to APA 7th edition guidelines.
APA 7th Results Narrative Example
“A structural equation model evaluated Technology Acceptance Model predictors of electronic lab notebook adoption (N = 250). Both scales demonstrated high internal consistency: Perceived Usefulness (α = .94) and Perceived Ease of Use (α = .91). The model accounted for 38.3% of the variance in Behavioral Intention (R² = .383) and 23.0% of the variance in Perceived Usefulness (R² = .230). Perceived Ease of Use significantly predicted Perceived Usefulness (β = .480, SE = .055, t = 8.65, p < .001). Behavioral Intention was strongly predicted by Perceived Usefulness (β = .518, SE = .058, t = 8.93, p < .001) and moderately predicted by Perceived Ease of Use (β = .172, SE = .058, t = 2.96, p = .003). Bootstrap mediation analysis (5,000 resamples) confirmed a significant indirect effect of Perceived Ease of Use on Behavioral Intention through Perceived Usefulness (β_indirect = .249, 95% CI [.174, .331], p < .001).”
Davis (1989) Citation Disambiguation
- Scale Validation Paper: Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008 (Cite for 6-item scales, psychometrics, and definitions).
- Model Comparison Paper: Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management Science, 35(8), 982–1003. https://doi.org/10.1287/mnsc.35.8.982 (Cite when comparing TAM against TRA).
How to Calculate It in Excel, R, or Python
Researchers calculate TAM regression path models using spreadsheets or programming languages, often combined with modeling in R techniques.
Excel Implementation
Using standardized columns (PEOU in Column A, PU in Column B, BI in Column C):
# 1. PEOU -> PU Path: =SLOPE(B2:B251, A2:A251)
# 2. R-squared for PU: =RSQ(B2:B251, A2:A251)
# 3. Multiple Regression for BI (LINEST array): =LINEST(C2:C251, A2:B251, TRUE, TRUE)
R Implementation
In R, use lavaan for structural equation modeling:
library(lavaan)
tam_model <- '
PU =~ PU1 + PU2 + PU3 + PU4 + PU5 + PU6
PEOU =~ PEOU1 + PEOU2 + PEOU3 + PEOU4 + PEOU5 + PEOU6
BI =~ BI1 + BI2 + BI3
PU ~ a * PEOU
BI ~ b * PU + c * PEOU
indirect := a * b
total := c + (a * b)
'
fit <- sem(tam_model, data = df, std.lv = TRUE)
summary(fit, fit.measures = TRUE, standardized = TRUE, rsquare = TRUE)
Python Implementation
In Python, calculate path equations and mediation effects using statsmodels:
import statsmodels.formula.api as smf
model_pu = smf.ols('PU ~ PEOU', data=df).fit()
model_bi = smf.ols('BI ~ PU + PEOU', data=df).fit()
beta_a = model_pu.params['PEOU']
beta_b = model_bi.params['PU']
beta_c = model_bi.params['PEOU']
indirect = beta_a * beta_b
total = beta_c + indirect
print(f"Indirect: {indirect:.3f}, Total: {total:.3f}")
FAQ
What is a technology acceptance model?
A Technology Acceptance Model is a theoretical framework explaining how users accept and adopt technology. It models adoption as a causal chain where system features influence cognitive beliefs (perceived usefulness and ease of use), determining behavioral intentions and actual usage.
What is the TAM model used for?
TAM is used to predict technology adoption rates, identify usability bottlenecks prior to deployment, and evaluate whether user resistance stems from deficient functionality or interface complexity.
Who proposed the TAM theory?
Fred D. Davis proposed TAM in his 1986 MIT Sloan doctoral dissertation, validating its scales in MIS Quarterly (1989) and evaluating it against the Theory of Reasoned Action in Management Science (1989).
What are the main components of the TAM model?
The primary components are External Variables, Perceived Usefulness (PU), Perceived Ease of Use (PEOU), Attitude Toward Using (ATU), Behavioral Intention to Use (BI), and Actual System Use, organized across independent vs. dependent variables.
How does TAM differ from UTAUT?
TAM focuses on two core cognitive predictors in voluntary settings. UTAUT unifies eight models into four constructs (Performance Expectancy, Effort Expectancy, Social Influence, Facilitating Conditions) with four demographic moderators, explaining up to 70% of variance.
Where can researchers find the original Technology Acceptance Model questionnaire?
The original validated 12-item questionnaire is published in Davis’s 1989 MIS Quarterly paper (Vol. 13, No. 3, pp. 319–340), containing six items measuring Perceived Usefulness and six items measuring Perceived Ease of Use across 7-point Likert anchors.
Sources
- Davis, F. D. (1986). A technology acceptance model for empirically testing new end-user information systems: Theory and results (Doctoral dissertation, MIT Sloan). DSpace@MIT. https://dspace.mit.edu/handle/1721.1/15192 (Accessed 2026-11-08).
- Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008 (Accessed 2026-11-08).
- Davis, F. D., Bagozzi, R. P., & Warshaw, P. R. (1989). User acceptance of computer technology: A comparison of two theoretical models. Management Science, 35(8), 982–1003. https://doi.org/10.1287/mnsc.35.8.982 (Accessed 2026-11-08).
- Marikyan, D., & Papagiannidis, S. (2023). Technology Acceptance Model: A review. In S. Papagiannidis (Ed.), TheoryHub Book. Newcastle University. https://open.ncl.ac.uk/theories/1/technology-acceptance-model/ (Accessed 2026-11-08).
- Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the Technology Acceptance Model: Four longitudinal field studies. Management Science, 46(2), 186–204. https://doi.org/10.1287/mnsc.2000.0200 (Accessed 2026-11-08).
- Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), 425–478. https://doi.org/10.2307/30036540 (Accessed 2026-11-08).







