DepEd ICT Equipment Audit: 47,818 Public Schools Statistical Analysis 2023-2024
Comprehensive statistical analysis of ICT equipment distribution across 47,818 Philippine public schools. Covers computer inventory (2M devices), funding sources, Gini inequality metrics, and regional allocation patterns from DepEd's SY 2023-2024 audit.
This report provides a comprehensive statistical analysis of the ICT equipment
audit for Philippine public schools in School Year 2023-2024. All findings are
cross-referenced with supporting tables and visualizations throughout.
Executive Summary
The audit covers 47,818 public schools with ICT device coverage at 96.3% for any
device, 87.2% for any computer, and 94.3% for any ICT equipment. Total recorded
devices are 3,151,403, consisting of 2,072,008 computers and 1,079,395 ICT
equipment. The median device count per school is 24, indicating a skewed
distribution when compared with the mean (65.91).
Table 1. Executive Summary (Public Schools, SY 2023-2024).
| Metric | Value | Percent |
|---|
| Total Public Schools | 47,818 | - |
| Schools with Any ICT Device | 46,029 | 96.3% |
| Schools with Any Computer | 41,689 | 87.2% |
| Schools with Any ICT Equipment | 45,101 | 94.3% |
| Total ICT Devices | 3,151,403 | - |
| Total Computers | 2,072,008 | - |
| Total ICT Equipment | 1,079,395 | - |
| Median Devices per School | 24.0 | - |
Figure 1. Total ICT devices (computers and equipment). The total device
count equals the sum of computers and equipment in Table 1.
Figure 2. Schools with any ICT devices. Bars show yes/no/unknown responses,
consistent with the 96.3% coverage in Table 1.
Figure 3. Total ICT devices by school management. The distribution shows
variability and outliers across management types.
Data Coverage and Validation
The source dataset contains 60,167 records, filtered to 47,818 public school
records. No duplicate school IDs were detected and no missing school IDs were
recorded.
ICT Inventory: Computers
Academic computers dominate the computer inventory, with tablets as the largest
academic category (926,796). Desktops and laptops are the next largest academic
classes, while virtual terminals are exclusively academic in this dataset.
Table 2. Computers by Type (Academic vs Admin).
| Type | Academic Total | Admin Total |
|---|
| desktop | 261803 | 36666 |
| laptop | 231798 | 58487 |
| all_in_one | 55735 | 2231 |
| tablet | 926796 | 21222 |
| tablet_pc | 286627 | 5262 |
| virtual_terminal | 185381 | 0 |
Figure 4. Academic computers by type. Tablets are the largest academic
computer category, consistent with Table 2.
Figure 5. Admin computers by type. Admin counts are lower across all types
compared with academic devices in Table 2.
ICT Inventory: Equipment
Printers are the largest academic ICT equipment category (288,900), followed by
LED TVs and smart TVs. Lapel devices are present only in the academic category.
Table 3. ICT Equipment by Type (Academic vs Admin).
| Type | Academic Total | Admin Total |
|---|
| led_tv | 176177 | 10179 |
| smart_tv | 175919 | 12454 |
| network_switch | 30168 | 3384 |
| printer | 288900 | 78843 |
| projector | 53758 | 6568 |
| external_hard_drive | 56492 | 16104 |
| ups | 72076 | 8562 |
| router | 28997 | 10412 |
| lapel | 50402 | 0 |
Figure 6. Academic ICT equipment by type. Printers lead the academic
equipment inventory, consistent with Table 3.
Figure 7. Admin ICT equipment by type. Admin equipment is concentrated in
printers and smart/LED TVs, matching Table 3.
Funding Sources
DepEd Computerization Program (DCP) is the largest funding source for total
devices (1,310,038), followed by LGU-SEF (783,915) and DepEd (645,758).
Table 4. Funding Sources Summary.
| Funding Source | Total Devices | Academic Devices |
|---|
| deped_dcp | 1310038 | 1227518 |
| deped | 645758 | 554540 |
| lgu_sef | 783915 | 741450 |
| pta | 92507 | 89848 |
| private_donation | 90062 | 79848 |
| other_gov | 50840 | 43170 |
| other_funding | 178283 | 144655 |
Figure 8. Total devices by funding source. DepEd DCP, LGU-SEF, and DepEd are
the top three sources by volume (Table 4).
Figure 9. Academic devices by funding source. Academic totals mirror the
ranking in Table 4, with DepEd DCP leading.
Devices by School Level
Elementary schools hold the largest academic device inventory (1,206,312),
followed by JHS (1,010,163) and SHS (664,554). Admin devices follow the same
ranking.
Table 5. Devices by School Level.
| Level | Academic Devices | Admin Devices |
|---|
| ES | 1206312 | 139766 |
| JHS | 1010163 | 90459 |
| SHS | 664554 | 40149 |
Note: Schools offering multiple levels (e.g., integrated ES+JHS schools) are
counted in each applicable level row. Totals therefore represent level-specific
device allocations, not mutually exclusive school groups.
Figure 10. Academic devices by school level. ES has the largest academic
device count, consistent with Table 5.
Figure 11. Admin devices by school level. ES leads admin device totals,
followed by JHS and SHS (Table 5).
Distribution and Inequality
Device counts are highly skewed: the median is 24 while the mean is 65.91, with
an upper tail reaching 32,216 devices in a single school. Inequality metrics
show high concentration for computers (Gini 0.8186) and overall devices (Gini
0.6938).
Table 6. Device Distribution Summary.
| Metric | Value |
|---|
| Min | 0 |
| Mean | 65.91 |
| Median | 24 |
| Std | 252.94 |
| P10 | 4 |
| P25 | 11 |
| P75 | 56 |
| P90 | 145 |
| Max | 32216 |
Figure 12. Distribution of total ICT devices per school. The long right tail
supports the skew seen in Table 6.
Figure 13. Funding diversity distribution (Shannon entropy). This reflects
variation in funding mix across schools.
Table 7. Inequality Metrics.
| Metric | Value | 95% CI (Bootstrap) |
|---|
| Gini - Computers | 0.8186 | [0.8142, 0.8230] |
| Gini - Equipment | 0.5364 | [0.5298, 0.5430] |
| Gini - All Devices | 0.6938 | [0.6884, 0.6992] |
Note: Bootstrap confidence intervals computed with 1,000 resamples using the
bias-corrected and accelerated (BCa) method.
Figure 14. Lorenz curve for device allocation. The curvature indicates
unequal distribution, consistent with Gini values in Table 7.
Correlation and Multivariate Structure
Pearson correlations highlight how core indicators move together across schools,
while PCA provides a low-dimensional view of the multivariate structure.
Figure 15. Pearson correlation matrix for core ICT indicators.
Figure 16. PCA projection (2 components) summarizing multivariate structure.
Clustering and Segmentability
Cluster validation metrics show modest separation for both k-means and
hierarchical clustering, suggesting some structure but overlapping groups.
Clustering was performed on a stratified random sample of 5,000 schools
(≈10% of total) for computational efficiency; results are representative
of the full dataset.
Table 10. Cluster Validation Metrics.
| Method | Silhouette | Davies-Bouldin | Calinski-Harabasz | Sample Size |
|---|
| kmeans | 0.3687 | 0.992 | 1804.48 | 5000 |
| hierarchical | 0.3375 | 1.0118 | 1576.83 | 5000 |
Figure 17. Cluster validation metrics. Silhouette values are moderate,
indicating partial but not strong separation.
Distribution Diagnostics and Fits
Normality diagnostics show strong departures from normality for all key metrics,
which justifies the use of count and non-parametric models in subsequent
analyses. Distribution fit selection (AIC) identifies zero-inflated and negative
binomial models for most count indicators, while funding entropy and HHI are
best fit by gamma distributions.
Table 11. Distribution Fit Summary.
| Metric | Best Distribution | AIC | N | Parameters | Sample Note |
|---|
| computers_academic_total | zero_inflated_negbin | 374076 | 47817 | -14.6770, 3.7072, 4.1642 | count_data |
| computers_admin_total | zero_inflated_negbin | 192272 | 47817 | -14.1559, 0.9518, 2.3363 | count_data |
| computers_total | zero_inflated_negbin | 403422 | 47817 | -15.6256, 3.7689, 3.1731 | count_data |
| equipment_academic_total | negbin | 382166 | 47817 | 2.9709 | count_data |
| equipment_admin_total | zero_inflated_negbin | 213141 | 47817 | -16.2124, 1.1197, 1.6659 | count_data |
| equipment_total | negbin | 395793 | 47817 | 3.1168 | count_data |
| devices_total | negbin | 496894 | 47817 | 4.1882 | count_data |
| device_type_diversity | poisson | 247056 | 47818 | 1.7350 | count_data |
| funding_entropy | gamma | 55077.7 | 39964 | 213.1536, -5.8541, 0.0330 | positive_only |
| funding_hhi | gamma | -11852.1 | 46029 | 3.6827, 0.1420, 0.1221 | full_sample |
Table 12. Normality Diagnostics.
| Metric | Shapiro p | DAgostino p | Lilliefors p |
|---|
| computers_academic_total | 0 | 0 | 0.001 |
| computers_admin_total | 0 | 0 | 0.001 |
| computers_total | 0 | 0 | 0.001 |
| equipment_academic_total | 0 | 0 | 0.001 |
| equipment_admin_total | 0 | 0 | 0.001 |
| equipment_total | 0 | 0 | 0.001 |
| devices_total | 0 | 0 | 0.001 |
| device_type_diversity | 0 | 0 | 0.001 |
| funding_entropy | 0 | 0 | 0.001 |
| funding_hhi | 0 | 0 | 0.001 |
Figure 18. Q-Q plot for computers_total. Deviations from the line confirm
non-normality noted in Table 12.
Figure 19. Q-Q plot for device_type_diversity. The distribution departs from
normality, consistent with Table 12.
Figure 20. Q-Q plot for devices_total. The long tail aligns with the skew in
Table 6 and non-normal results in Table 12.
Figure 21. Q-Q plot for equipment_total. The pattern reflects non-normal
distributions in Table 12.
Figure 22. Q-Q plot for funding_entropy. The distribution deviates from a
straight line, consistent with Table 12.
Figure 23. Discrete fit for computers_academic_total (zero-inflated negative
binomial; Table 11). The fitted PMF overlays observed counts.
Figure 24. Discrete fit for computers_admin_total (zero-inflated negative
binomial; Table 11).
Figure 25. Discrete fit for computers_total (zero-inflated negative
binomial; Table 11).
Figure 26. Discrete fit for device_type_diversity (Poisson; Table 11).
Figure 27. Discrete fit for devices_total (negative binomial; Table 11).
Figure 28. Discrete fit for equipment_academic_total (negative binomial;
Table 11).
Figure 29. Discrete fit for equipment_admin_total (zero-inflated negative
binomial; Table 11).
Figure 30. Discrete fit for equipment_total (negative binomial; Table 11).
Figure 31. Distribution fit for funding_entropy (gamma; Table 11).
Figure 32. Distribution fit for funding_hhi (gamma; Table 11).
Statistical Tests
Tests indicate a statistically significant association between computer
availability and school management, but the effect size is small (Cramer's V
0.0273). Academic device totals differ substantially between SHS-offering and
non-SHS schools (|rank-biserial r| = 0.7236). Total device counts by management
are not statistically significant at the 0.05 level after corrections.
Table 8. Hypothesis Tests with Effect Sizes.
| Test | Statistic | p-value | p-adj FDR | p-adj Bonferroni | Effect Size | Effect Label | Assumptions |
|---|
| Chi-square: Computer Availability vs Management | 35.6812 | 0 | 0 | 0 | 0.0273 | Cramer's V | low_expected_cells=3 |
| Mann-Whitney U: Academic Devices by SHS Offer | 2.67406e+08 | 0 | 0 | 0 | -0.7236 | Rank-biserial r | non-normal or small sample |
| Kruskal-Wallis: Total Devices by Management | 3.1214 | 0.0773 | 0.0773 | 0.2318 | 0 | Epsilon squared | non-normal or heteroscedastic |
Table 13. Post-hoc Dunn Test.
| Group | DOST | DepEd | Other GA |
|---|
| DOST | 1 | 0.231745 | 0.705341 |
| DepEd | 0.231745 | 1 | 0.158185 |
| Other GA | 0.705341 | 0.158185 | 1 |
Interpretation: All pairwise p-values in Table 13 are above 0.05, indicating no
statistically significant pairwise differences in the post-hoc comparisons.
Outlier Summary
Outliers are present under multiple detection rules, reflecting the long-tailed
distribution of device counts.
Table 9. Outlier Summary.
| Method | Count |
|---|
| z_score_outliers | 1576 |
| iqr_outliers | 9837 |
| isolation_forest_outliers | 2302 |
| dbscan_outliers | skipped_large_sample |
Regression Summary
Regression models ran successfully on 47,817 schools using level offerings and
management indicators as predictors. Coefficients indicate associations, not
causal effects.
- OLS log devices: offers_jhs and offers_shs are positive and significant; offers_es
is negative and significant; mgmt_DepEd is positive (p=0.010), while mgmt_Other GA
is not significant.
- Poisson GLM: offers_jhs and offers_shs are positive; offers_es is negative; mgmt_Other GA
is negative and significant; mgmt_DepEd is not significant.
- Negative Binomial GLM: matches the Poisson sign pattern, with mgmt_Other GA
negative and significant (p=0.020); mgmt_DepEd is not significant.
- Logit (any device): offers_es/jhs/shs and mgmt_DepEd are positive and significant;
mgmt_Other GA is not significant.
Given the over-dispersion in device counts (Table 6) and the distribution-fit
selection (Table 11), the Negative Binomial model is the preferred count
specification for device totals.
Table 14. Regression Results (Coefficients and IRR/OR).
| Model | Term | Coef | Exp Label | Exp(Coef) | Pct Change (approx) | p-value |
|---|
| OLS log(devices_total+1) | const | 0.8851 | Multiplier | 2.4233 | 142.33% | 0.3067 |
| OLS log(devices_total+1) | offers_es | -0.2191 | Multiplier | 0.8032 | -19.68% | 0 |
| OLS log(devices_total+1) | offers_jhs | 0.3971 | Multiplier | 1.4876 | 48.76% | 0 |
| OLS log(devices_total+1) | offers_shs | 1.1977 | Multiplier | 3.3126 | 231.26% | 0 |
| OLS log(devices_total+1) | mgmt_DepEd | 2.2438 | Multiplier | 9.4291 | 842.91% | 0.0095 |
| OLS log(devices_total+1) | mgmt_Other GA | -0.4744 | Multiplier | 0.6223 | -37.77% | 0.8729 |
| Poisson GLM | const | 4.2785 | IRR | 72.1314 | NA | 0 |
| Poisson GLM | offers_es | -0.7912 | IRR | 0.4533 | NA | 0 |
| Poisson GLM | offers_jhs | 0.3672 | IRR | 1.4437 | NA | 0 |
| Poisson GLM | offers_shs | 0.703 | IRR | 2.0198 | NA | 0 |
| Poisson GLM | mgmt_DepEd | 0.0614 | IRR | 1.0633 | NA | 0.8714 |
| Poisson GLM | mgmt_Other GA | -3.4137 | IRR | 0.0329 | NA | 0.0077 |
| Negative Binomial GLM | const | 4.2083 | IRR | 67.2444 | NA | 0 |
| Negative Binomial GLM | offers_es | -0.7427 | IRR | 0.4758 | NA | 0 |
| Negative Binomial GLM | offers_jhs | 0.4632 | IRR | 1.5892 | NA | 0 |
| Negative Binomial GLM | offers_shs | 0.6946 | IRR | 2.003 | NA | 0 |
| Negative Binomial GLM | mgmt_DepEd | 0.0648 | IRR | 1.0669 | NA | 0.8633 |
| Negative Binomial GLM | mgmt_Other GA | -2.187 | IRR | 0.1123 | NA | 0.0203 |
| Logit (Any Device) | const | -2.0474 | OR | 0.1291 | NA | 0.001 |
| Logit (Any Device) | offers_es | 0.9261 | OR | 2.5247 | NA | 0 |
| Logit (Any Device) | offers_jhs | 0.4868 | OR | 1.6271 | NA | 0.0013 |
| Logit (Any Device) | offers_shs | 1.1851 | OR | 3.271 | NA | 0 |
| Logit (Any Device) | mgmt_DepEd | 4.2851 | OR | 72.6079 | NA | 0 |
| Logit (Any Device) | mgmt_Other GA | 0.7484 | OR | 2.1136 | NA | 0.6828 |
Interpretation: Exp(Coef) is the incidence rate ratio (IRR) for Poisson/Negative
Binomial models, the odds ratio (OR) for the logit model, and a multiplicative
factor for the log-transformed OLS model. The OLS percent-change column is an
approximate change in devices_total + 1.
Figure 33. Regression summary (IRR/OR, 95% CI). The Negative Binomial model
summarizes device count associations, while the Logit model summarizes
device-availability odds. Values above 1 indicate higher counts or odds.
## ols_log_devices
OLS Regression Results
==============================================================================
Dep. Variable: devices_total R-squared: 0.235
Model: OLS Adj. R-squared: 0.235
Method: Least Squares F-statistic: 2733.
Date: Mon, 05 Jan 2026 Prob (F-statistic): 0.00
Time: 00:24:08 Log-Likelihood: -75797.
No. Observations: 47817 AIC: 1.516e+05
Df Residuals: 47811 BIC: 1.517e+05
Df Model: 5
Covariance Type: HC3
=================================================================================
coef std err z P>|z| [0.025 0.975]
---------------------------------------------------------------------------------
const 0.8851 0.866 1.022 0.307 -0.812 2.582
offers_es -0.2191 0.035 -6.255 0.000 -0.288 -0.150
offers_jhs 0.3971 0.027 14.809 0.000 0.345 0.450
offers_shs 1.1977 0.036 33.575 0.000 1.128 1.268
mgmt_DepEd 2.2438 0.866 2.592 0.010 0.547 3.940
mgmt_Other GA -0.4744 2.966 -0.160 0.873 -6.287 5.338
==============================================================================
Omnibus: 2886.230 Durbin-Watson: 1.291
Prob(Omnibus): 0.000 Jarque-Bera (JB): 5198.286
Skew: -0.460 Prob(JB): 0.00
Kurtosis: 4.327 Cond. No. 280.
==============================================================================
Notes:
[1] Standard Errors are heteroscedasticity robust (HC3)
## glm_poisson_devices
Generalized Linear Model Regression Results
==============================================================================
Dep. Variable: devices_total No. Observations: 47817
Model: GLM Df Residuals: 47811
Model Family: Poisson Df Model: 5
Link Function: Log Scale: 1.0000
Method: IRLS Log-Likelihood: -2.6035e+06
Date: Mon, 05 Jan 2026 Deviance: 4.9707e+06
Time: 00:24:08 Pearson chi2: 1.92e+07
No. Iterations: 7 Pseudo R-squ. (CS): 1.000
Covariance Type: HC3
=================================================================================
coef std err z P>|z| [0.025 0.975]
---------------------------------------------------------------------------------
const 4.2785 0.383 11.168 0.000 3.528 5.029
offers_es -0.7912 0.067 -11.855 0.000 -0.922 -0.660
offers_jhs 0.3672 0.061 5.983 0.000 0.247 0.487
offers_shs 0.7030 0.078 8.970 0.000 0.549 0.857
mgmt_DepEd 0.0614 0.379 0.162 0.871 -0.682 0.805
mgmt_Other GA -3.4137 1.281 -2.665 0.008 -5.925 -0.903
=================================================================================
## glm_negbin_devices
Generalized Linear Model Regression Results
==============================================================================
Dep. Variable: devices_total No. Observations: 47817
Model: GLM Df Residuals: 47811
Model Family: NegativeBinomial Df Model: 5
Link Function: Log Scale: 1.0000
Method: IRLS Log-Likelihood: -2.3376e+05
Date: Mon, 05 Jan 2026 Deviance: 69886.
Time: 00:24:08 Pearson chi2: 2.39e+05
No. Iterations: 10 Pseudo R-squ. (CS): 0.4589
Covariance Type: HC3
=================================================================================
coef std err z P>|z| [0.025 0.975]
---------------------------------------------------------------------------------
const 4.2083 0.378 11.136 0.000 3.468 4.949
offers_es -0.7427 0.058 -12.731 0.000 -0.857 -0.628
offers_jhs 0.4632 0.043 10.760 0.000 0.379 0.548
offers_shs 0.6946 0.064 10.823 0.000 0.569 0.820
mgmt_DepEd 0.0648 0.376 0.172 0.863 -0.673 0.802
mgmt_Other GA -2.1870 0.942 -2.320 0.020 -4.034 -0.340
=================================================================================
## logit_any_device
Logit Regression Results
==============================================================================
Dep. Variable: has_any_device No. Observations: 47817
Model: Logit Df Residuals: 47811
Method: MLE Df Model: 5
Date: Mon, 05 Jan 2026 Pseudo R-squ.: 0.01187
Time: 00:24:08 Log-Likelihood: -7539.5
converged: True LL-Null: -7630.0
Covariance Type: HC3 LLR p-value: 3.096e-37
=================================================================================
coef std err z P>|z| [0.025 0.975]
---------------------------------------------------------------------------------
const -2.0474 0.622 -3.294 0.001 -3.266 -0.829
offers_es 0.9261 0.182 5.094 0.000 0.570 1.282
offers_jhs 0.4868 0.151 3.227 0.001 0.191 0.782
offers_shs 1.1851 0.136 8.686 0.000 0.918 1.452
mgmt_DepEd 4.2851 0.595 7.205 0.000 3.119 5.451
mgmt_Other GA 0.7484 1.832 0.409 0.683 -2.842 4.338
=================================================================================
Key Findings
- Coverage and scale:
- 47,818 public schools were analyzed, with 96.3% reporting any ICT device
(Table 1, Figure 2).
- Inventory composition:
- Tablets are the largest academic computer category (926,796), and printers
are the largest academic equipment category (288,900) (Tables 2-3,
Figures 4 and 6).
- Funding structure:
- DepEd DCP provides the largest share of total devices (1,310,038), followed
by LGU-SEF and DepEd (Table 4, Figures 8-9).
- Level allocation:
- Elementary schools hold the highest academic device totals (1,206,312)
followed by JHS and SHS (Table 5, Figure 10).
- Distribution and inequality:
- Device counts are highly skewed (median 24 vs mean 65.91) with a maximum of
32,216, and Gini values indicate high concentration, especially for computers
(Tables 6-7, Figures 12 and 14).
- Statistical evidence:
- Computer availability differs by management with a small effect size, while
academic devices differ strongly by SHS offering (Table 8).
- Distribution modeling:
- Count indicators are best fit by negative binomial or zero-inflated models,
confirming over-dispersion and excess zeros (Table 11, Figures 23-30).
- Regression associations (non-causal):
- The preferred Negative Binomial model shows higher device counts for schools
offering JHS and SHS (IRR > 1) and lower counts for ES-only offerings (IRR < 1);
“Other GA” management is associated with lower counts (Table 14, Figure 33).
- The Logit model shows higher odds of having any device for schools offering
ES/JHS/SHS and for DepEd-managed schools; “Other GA” is not significant (Table 14).
Appendix A: Chart Index
Appendix B: Statistical Methods Summary
- Descriptive statistics: mean, median, mode, standard deviation, IQR, percentiles.
- Distribution fits:
- Continuous: normal, lognormal, gamma (AIC-based).
- Count models: Poisson, negative binomial, zero-inflated Poisson/NegBin.
- Normality diagnostics: Shapiro-Wilk, D'Agostino, Lilliefors.
- Hypothesis tests: chi-square, t-test/Welch, Mann-Whitney U, ANOVA, Kruskal-Wallis.
- Post-hoc testing: Dunn test with Bonferroni adjustment.
- Multiple testing corrections: FDR and Bonferroni.
- Correlation analysis: Pearson and Spearman.
- Multivariate analysis: PCA, factor analysis.
- Clustering validation: k-means and hierarchical metrics.
- Outlier detection: z-score, IQR, Isolation Forest, DBSCAN.
- Inequality metrics: Gini coefficient, Lorenz curve.
Technical Notes
Data Sources:
- Source: Department of Education (DepEd) ICT equipment data, SY 2023-24.
- Sector filter: Public schools only (
sector == "Public").
- School Year: 2023-24.
- Data Files:
Analytical Methods:
- All analyses were run on the cleaned ICT dataset with public-sector filtering.
- Missing values can represent non-applicability or unavailable data; the
missingness report should be consulted when interpreting totals.
- Discrete distribution fits are reported for count indicators; continuous
distribution fits are reported for funding entropy and HHI.