Gender Assessment Data

Try it live!

Use Binder to explore and run this dataset interactively in your browser. Ideal for students and instructors to run, modify, and test the analysis in a live JupyterLab environment—no setup needed.

👉 Launch on Binder

About the Data

The Global Gender Assessment data set provides insights into how companies across various industries and regions perform on a wide range of gender equity measures. This data set includes scores on strategic actions, leadership, grievance mechanisms, pay equity, and many other factors related to gender based practices within corporate environments.

Collected in 2023, the data allows for comparative evaluation across countries, sectors, and ownership types (e.g., Public, Private, Government). Each record represents a company and its corresponding evaluation across 28 detailed gender related indicators, offering a comprehensive snapshot of corporate gender responsibility worldwide.

The data set can be used by researchers, policy analysts, and advocacy groups to track progress on corporate gender equity, identify gaps, and support data driven decision making in both public and private sectors.

Download

Download CSV

Metadata

CSV Name

genderassessment.csv

Data set Characteristics

Multivariate

Subject Area

Equity, Gender, Corporate Governance

Associated Tasks

Comparative Analysis, Score Aggregation, Visualization

Feature Type

Character, Numeric, Factor

Instances

2000 records

Features

30

Has Missing Values?

Yes

Variables

Variable Name Role Type Description Units Missing Values
company ID Character Name of the company – No
country Feature Character Country where the company is headquartered – Yes
region Feature Factor Geographical region (e.g., “North America”, “Europe & Central Asia”) – Yes
industry Feature Factor Industry sector (e.g., “Chemicals”, “Retail”, “Metals & Mining”) – Yes
ownership Feature Factor Ownership type (e.g., “Public”, “Private”, “Government”) – Yes
year Feature Numeric Year of assessment Year No
score Target Numeric Overall gender assessment score (out of 52.3) – Yes
percent_score Target Numeric Overall score expressed as a percentage % Yes
strategic_action Feature Numeric Score (out of 1) for a public commitment to gender equality and women’s empowerment – Yes
gender_targets Feature Numeric Score (out of 2) for target setting on gender equality and women’s empowerment – Yes
gender_due_diligence Feature Numeric Score (out of 2) for gender-responsive human rights due diligence process – Yes
grievance_mechanisms Feature Numeric Score (out of 3) for availability of grievance mechanisms, along with the collection, analysis, and monitoring of the associated data by sex – Yes
stakeholder_engagement Feature Numeric Score (out of 1) for surveys or other engagement mechanisms that specifically address gender equality & women’s empowerment issues – Yes
corrective_action Feature Numeric Score (out of 2) for corrective action processes related to gender-related issues – Yes
gender_leadership Feature Numeric Score (out of 4) for gender equality in leadership – Yes
development_recruitment Feature Numeric Score (out of 2) for availability of recruitment and career development opportunities and collection of the associated data by sex – Yes
employee_data_by_sex Feature Numeric Score (out of 4) for availability of employee data by sex – Yes
supply_chain_gender_leadership Feature Numeric Score (out of 1) for collection of leadership data in the supply chain by sex – Yes
enabling_environment_union_rights Feature Numeric Score (out of 1) for enabling environment for freedom of association and collective bargaining – Yes
gender_procurement Feature Numeric Score (out of 2) for of gender-responsive procurement – Yes
gender_pay_gap Feature Numeric Score (out of 3) for collection and analysis of gender pay gap data – Yes
carer_leave_paid Feature Numeric Score (out of 4) for paid leave policies for caregivers and the monitoring of the associated data – Yes
childcare_support Feature Numeric Score (out of 2) for availability of childcare support and family support – Yes
flex_work Feature Numeric Score (out of 4) for flexible work options and collection of the associated data by sex – Yes
living_wage_supply_chain Feature Numeric Score (out of 2) for enforcement of living wage in supply chain – Yes
health_safety Feature Numeric Score (out of 3) for workplace health and safety, including the availability of sex-specific health and safety information to employees – Yes
health_safety_supply_chain Feature Numeric Score (out of 2) for the statement and monitoring of supply chain health and safety – Yes
violence_prevention Feature Numeric Score (out of 1) for policies and measures to prevent gender-based violence – Yes
violence_remediation Feature Numeric Score (out of 2) for mechanisms for remediation after incidents of violence, along with the collection, analysis, and monitoring of the associated data by sex – Yes

Key Features of the Data set

Each row represents a single company’s gender assessment in a given year and includes organizational details, overall scores, and performance on various gender-related indicators:

  • Organizational details (company, country, region, industry, ownership, year) – Company name, headquarters location, region, industry sector, ownership structure, and assessment year.
  • Assessment scores (score, percent_score) – Overall gender assessment score and percentage for comparison.

Gender Assessment Score Components

The Overall Gender Assessment Score is a weighted composite score derived from six thematic areas, as shown below:

Theme Category Weight (%) Description
A Governance and Strategy 20% Integration of gender equality in corporate governance and strategic frameworks.
B Representation 17.5% Gender balance in leadership, workforce, and management.
C Compensation and Benefits 17.5% Pay equity, benefits, and family-friendly policies.
D Health and Well-being 17.5% Worker well-being, safe working conditions, and health services.
E Violence and Harassment 17.5% Policies and mechanisms to prevent and respond to gender-based violence.
F Marketplace and Community 10% Gender-inclusive procurement, community programs, and marketing practices.

The final overall score is computed out of 35, based on these weighted components, and scaled to percentage (out of 100%) for easier comparison.

Purpose and Use Cases

This data set supports analysis of:

  • Gender equity performance across industries, regions, and ownership types.

  • Temporal trends in corporate gender assessment scores with a focus on changes over time to track progress or setbacks.

  • The influence of gender-related policies and corporate governance structures by examining how internal practices and ownership models affect outcomes.

  • The relationship between regulatory environments and gender equity outcomes exploring how regional or national policy contexts align with performance.

Case Study

Objective

Gender equality in the workplace is a critical goal for sustainable and inclusive development. Companies are increasingly being evaluated not only on financial metrics but also on their social and governance performance, including how they support gender equity. This data set offers a standardized assessment of a company’s gender-related policies and practices across the world.

To explore these dynamics, we focus on the following key questions:

What is the average Overall Gender Assessment Score by industry, and which industries have the highest and lowest average scores?

Does ownership type affect gender assessment performance?

By analyzing the score variable across industry and ownership categories, and exploring key gender-related indicators, we aim to identify structural patterns and gaps in corporate gender performance.

Analysis

Loading Libraries

library(diversedata)
library(tidyverse)  # includes ggplot2, readr, dplyr
library(janitor)    # for data cleaning
library(knitr)  # For kable() function
library(car)    # For leveneTest()
library(ggthemes)   # extra ggplot themes
library(ggalt)  # For geom_dumbbell
library(FSA) # For Dunn's test
import diversedata as dd
import pandas as pd
import altair as alt
import statsmodels.formula.api as smf
from scipy.stats import shapiro, levene, kruskal
import scikit_posthocs as sp
from IPython.display import Markdown

1. Data Cleaning & Processing

gender_data <- genderassessment |> 
  clean_names() |> 
  drop_na(score) |> 
  mutate(
    industry = as.factor(industry),
    ownership_type = as.factor(ownership)
  )
kable(head(gender_data))
company country region industry ownership year score percent_score strategic_action gender_targets gender_due_diligence grievance_mechanisms stakeholder_engagement corrective_action gender_leadership development_recruitment employee_data_by_sex supply_chain_gender_leadership enabling_environment_union_rights gender_procurement gender_pay_gap carer_leave_paid childcare_support flex_work living_wage_supply_chain health_safety health_safety_supply_chain violence_prevention violence_remediation ownership_type
3M United States North America Chemicals Public 2023 11.3 22 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 2 0 1.0 2 1.0 0 Public
Asos United Kingdom Europe & Central Asia Apparel & Footwear Public 2023 16.9 32 1 0 0 2 0 0 3 1 0 0 1 0 0 0 1 1 2 0.5 2 0.5 0 Public
A.P. Moller - Maersk Denmark Europe & Central Asia Freight & logistics Public 2024 10.9 21 1 1 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 1.0 2 1.0 0 Public
ABB Switzerland Europe & Central Asia Capital Goods Public 2023 12.8 25 1 1 1 2 0 0 0 1 0 0 0 0 0 1 0 0 0 1.0 2 1.0 0 Public
AbbVie United States North America Pharmaceuticals & Biotechnology Public 2023 15.4 30 1 0 0 2 0 0 2 1 0 0 0 1 0 0 2 1 0 1.0 2 1.0 0 Public
Abercrombie & Fitch United States North America Apparel & Footwear Public 2023 10.0 19 0 1 0 2 0 0 1 0 0 0 0 0 0 0 0 1 0 1.0 2 1.0 0 Public 
gender_data = dd.load_data("genderassessment").dropna(subset='score')
gender_data['industry'] = pd.Categorical(gender_data['industry'])
gender_data['ownership'] = pd.Categorical(gender_data['ownership'])
Markdown(gender_data.head().to_markdown())
company country region industry ownership year score percent_score strategic_action gender_targets gender_due_diligence grievance_mechanisms stakeholder_engagement corrective_action gender_leadership development_recruitment employee_data_by_sex supply_chain_gender_leadership enabling_environment_union_rights gender_procurement gender_pay_gap carer_leave_paid childcare_support flex_work living_wage_supply_chain health_safety health_safety_supply_chain violence_prevention violence_remediation
0 3M United States North America Chemicals Public 2023 11.3 22 1 0 0 0 0 0 0 0 0 0 1 1 0 0 0 2 0 1 2 1 0
1 Asos United Kingdom Europe & Central Asia Apparel & Footwear Public 2023 16.9 32 1 0 0 2 0 0 3 1 0 0 1 0 0 0 1 1 2 0.5 2 0.5 0
2 A.P. Moller - Maersk Denmark Europe & Central Asia Freight & logistics Public 2024 10.9 21 1 1 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 1 2 1 0
3 ABB Switzerland Europe & Central Asia Capital Goods Public 2023 12.8 25 1 1 1 2 0 0 0 1 0 0 0 0 0 1 0 0 0 1 2 1 0
4 AbbVie United States North America Pharmaceuticals & Biotechnology Public 2023 15.4 30 1 0 0 2 0 0 2 1 0 0 0 1 0 0 2 1 0 1 2 1 0

2. Exploratory Data Analysis

Average Overall Assessment Score by Industry

industry_stats <- gender_data |> 
  group_by(industry) |> 
  summarise(
    avg_score = mean(score, na.rm = TRUE),
    median_score = median(score, na.rm = TRUE),
    sd_score = sd(score, na.rm = TRUE),
    n = n()
  ) |> 
  arrange(desc(avg_score))

kable(industry_stats, digits = 2, caption = "Average Score by Industry")
Average Score by Industry
industry avg_score median_score sd_score n
Personal & Household Products 14.69 15.60 4.64 20
Electronics 12.63 13.30 5.44 55
Pharmaceuticals & Biotechnology 12.37 12.90 5.80 29
Apparel & Footwear 11.49 10.80 5.98 68
Banks 10.73 11.20 5.80 150
IT Software & Services 10.24 10.25 5.85 34
Retail 9.71 10.45 6.40 18
Insurance 9.60 10.20 5.26 71
Motor Vehicles & Parts 9.54 10.70 6.23 45
Oil & Gas 9.50 10.30 5.93 97
Capital Goods 9.38 9.80 5.99 25
Chemicals 9.23 9.80 4.71 70
Telecommunications 9.09 10.00 5.70 83
Entertainment 9.00 9.00 6.51 2
Payments 9.00 9.00 9.62 2
Food Retailers 8.93 9.10 5.39 76
NA 8.28 8.40 9.58 5
Containers & Packaging 7.53 8.00 4.59 23
Traditional Asset Managers 7.51 8.20 5.32 63
Food Production 7.50 7.10 5.53 206
Development Finance Institutions 7.43 7.60 4.33 15
Utilities 6.95 7.30 5.64 125
Conglomerates 6.94 7.75 6.33 18
Postal & Courier Activities 6.94 6.90 3.35 21
Metals & Mining 6.86 7.35 5.62 114
Construction Materials & Supplies 6.81 6.90 5.33 46
Paper & Forest Products 6.66 6.20 4.04 29
Freight & logistics 6.60 6.50 3.91 33
Waste Management 6.51 7.25 5.70 18
Real Estate 6.38 6.00 5.16 147
Investment Consultants 6.38 5.80 6.22 5
Construction & Engineering 5.64 5.35 5.07 56
Passenger Transport 4.89 2.90 5.21 101
Agricultural Products 4.24 4.25 3.60 32
Alternative Asset Managers 3.63 2.90 3.84 21
Sovereign Wealth Funds 3.18 0.50 4.80 19
Pension Funds 2.97 1.30 3.62 58 
industry_stats = gender_data.groupby(by="industry", observed=False)["score"].agg(
    ["mean", "median", "std", "count"]
)
industry_stats.columns = ["avg_score", "median_score", "sd_score", "n"]
industry_stats = industry_stats.round(2).sort_values(by=["avg_score"])
Markdown(industry_stats.to_markdown())
Average Score by Industry
industry avg_score median_score sd_score n
Pension Funds 2.97 1.3 3.62 58
Sovereign Wealth Funds 3.18 0.5 4.8 19
Alternative Asset Managers 3.63 2.9 3.84 21
Agricultural Products 4.24 4.25 3.6 32
Passenger Transport 4.89 2.9 5.21 101
Construction & Engineering 5.64 5.35 5.07 56
Investment Consultants 6.38 5.8 6.22 5
Real Estate 6.38 6 5.16 147
Waste Management 6.51 7.25 5.7 18
Freight & logistics 6.6 6.5 3.91 33
Paper & Forest Products 6.66 6.2 4.04 29
Construction Materials & Supplies 6.81 6.9 5.33 46
Metals & Mining 6.86 7.35 5.62 114
Postal & Courier Activities 6.94 6.9 3.35 21
Conglomerates 6.94 7.75 6.33 18
Utilities 6.95 7.3 5.64 125
Development Finance Institutions 7.43 7.6 4.33 15
Food Production 7.5 7.1 5.53 206
Traditional Asset Managers 7.51 8.2 5.32 63
Containers & Packaging 7.53 8 4.59 23
Food Retailers 8.93 9.1 5.39 76
Payments 9 9 9.62 2
Entertainment 9 9 6.51 2
Telecommunications 9.09 10 5.7 83
Chemicals 9.23 9.8 4.71 70
Capital Goods 9.38 9.8 5.99 25
Oil & Gas 9.5 10.3 5.93 97
Motor Vehicles & Parts 9.54 10.7 6.23 45
Insurance 9.6 10.2 5.26 71
Retail 9.71 10.45 6.4 18
IT Software & Services 10.24 10.25 5.85 34
Banks 10.73 11.2 5.8 150
Apparel & Footwear 11.49 10.8 5.98 68
Pharmaceuticals & Biotechnology 12.37 12.9 5.8 29
Electronics 12.63 13.3 5.44 55
Personal & Household Products 14.69 15.6 4.64 20

Visualization: Average Overall Gender Assessment by Industry Group

gender_data_grouped <- gender_data |> 
  mutate(industry_group = case_when(
    industry %in% c("Apparel & Footwear", "Food Production", "Food Retailers", 
                    "Retail", "Personal & Household Products", "Agricultural Products") ~ "Consumer Goods",
    
    industry %in% c("Banks", "Insurance", "Pension Funds", "Sovereign Wealth Funds", 
                    "Alternative Asset Managers", "Traditional Asset Managers", 
                    "Investment Consultants", "Payments") ~ "Finance",
    
    industry %in% c("Pharmaceuticals & Biotechnology") ~ "Healthcare",
    
    industry %in% c("IT Software & Services", "Telecommunications", "Electronics") ~ "Technology",
    
    industry %in% c("Oil & Gas", "Utilities") ~ "Energy & Utilities",
    
    industry %in% c("Chemicals", "Capital Goods", "Freight & logistics", 
                    "Passenger Transport", "Metals & Mining", "Containers & Packaging", 
                    "Motor Vehicles & Parts", "Postal & Courier Activities", 
                    "Waste Management", "Paper & Forest Products") ~ "Industrials",
    
    industry %in% c("Construction & Engineering", "Construction Materials & Supplies", "Real Estate") ~ "Real Estate",
    
    industry %in% c("Development Finance Institutions") ~ "Public Sector Finance",
    
    industry %in% c("Entertainment") ~ "Entertainment",
    
    TRUE ~ "Other"
  ))

gender_by_group <- ggplot(gender_data_grouped, aes(x = reorder(industry_group, score, FUN = median), y = score)) +
  geom_boxplot(width = 0.5, fill = "skyblue", outlier.alpha = 0.4) +  # width < 1 reduces box width
  coord_flip() +
  labs(
    title = "Overall Average Gender Assessment Score by Industry Group",
    x = "Industry Group", y = "Score"
  ) +
  theme_minimal()

gender_by_group

Show grouping code 
gender_data_grouped = gender_data.copy()
industry_group_mapping = {
    "Apparel & Footwear": "Consumer Goods",
    "Food Production": "Consumer Goods",
    "Food Retailers": "Consumer Goods",
    "Retail": "Consumer Goods",
    "Personal & Household Products": "Consumer Goods",
    "Agricultural Products": "Consumer Goods",

    "Banks": "Finance",
    "Insurance": "Finance",
    "Pension Funds": "Finance",
    "Sovereign Wealth Funds": "Finance",
    "Alternative Asset Managers": "Finance",
    "Traditional Asset Managers": "Finance",
    "Investment Consultants": "Finance",
    "Payments": "Finance",

    "Pharmaceuticals & Biotechnology": "Healthcare",

    "IT Software & Services": "Technology",
    "Telecommunications": "Technology",
    "Electronics": "Technology",

    "Oil & Gas": "Energy & Utilities",
    "Utilities": "Energy & Utilities",

    "Chemicals": "Industrials",
    "Capital Goods": "Industrials",
    "Freight & logistics": "Industrials",
    "Passenger Transport": "Industrials",
    "Metals & Mining": "Industrials",
    "Containers & Packaging": "Industrials",
    "Motor Vehicles & Parts": "Industrials",
    "Postal & Courier Activities": "Industrials",
    "Waste Management": "Industrials",
    "Paper & Forest Products": "Industrials",

    "Construction & Engineering": "Real Estate",
    "Construction Materials & Supplies": "Real Estate",
    "Real Estate": "Real Estate",

    "Development Finance Institutions": "Public Sector Finance",

    "Entertainment": "Entertainment"
}

# apply groupings
gender_data_grouped["industry_group"] = (
    gender_data_grouped["industry"].map(industry_group_mapping).fillna("Other")
)
industry_group_order_by_median = (
    gender_data_grouped.groupby(by="industry_group", observed=False, as_index=False)["score"]
    .median()
    .sort_values(by="score", ascending=False)["industry_group"]
    .to_list()
)

alt.Chart(gender_data_grouped).mark_boxplot(size=20).encode(
    x=alt.X("score").title("Score"),
    y=alt.Y("industry_group").sort(industry_group_order_by_median).title("Industry Group"),
).properties(
    title="Overall Average Gender Assessment Score by Industry Group",
    width=600,
    height=400,
)

Comparison of Average Overall Gender Assessment Scores Between 2023 vs 2024 by Region

region_wide <- gender_data |> 
  filter(year %in% c(2023, 2024)) |>  # Ensure only these years are included
  group_by(region, year) |> 
  summarise(avg_score = mean(score, na.rm = TRUE)) |> 
  pivot_wider(
    names_from = year, 
    values_from = avg_score,
    names_prefix = ""  # Removes automatic "year" prefix
  ) |> 
  ungroup()

ggplot(region_wide, aes(x = `2023`, xend = `2024`, y = reorder(region, `2024`))) +
  geom_dumbbell(
    size = 1.2,  # thinner connecting line
    colour = "#a3c4dc",       # light blue line
    colour_x = "#87CEEB",     # skyblue for 2023
    colour_xend = "#00008B"   # dark blue for 2024
  ) +
  # Dummy points to create legend
  geom_point(aes(x = `2023`, color = "2023"), size = 2) +
  geom_point(aes(x = `2024`, color = "2024"), size = 2) +
  scale_color_manual(
    name = "Year",
    values = c("2023" = "#87CEEB", "2024" = "#00008B")
  ) +
  labs(
    title = "Change in Average Score by Region (2023 to 2024)",
    x = "Average Score", y = "Region"
  ) +
  theme_minimal() +
  theme(axis.title.y = element_blank())

region_wide = (
    gender_data.loc[gender_data["year"].isin([2023, 2024])]
    .groupby(["region", "year"], as_index=False)["score"]
    .mean()
    .rename(columns={"score": "avg_score"})
)

region_order = (
    region_wide[region_wide["year"] == 2024]
    .sort_values(by="avg_score", ascending=False)["region"]
    .to_list()
)

lines = alt.Chart(region_wide).mark_line(size=3, opacity=0.4).encode(
    x=alt.X("avg_score"),
    y=alt.Y("region").sort(region_order),
    detail=alt.Detail("region"),
)

points = alt.Chart(region_wide).mark_circle(opacity=1, size=30).encode(
    x=alt.X("avg_score").title("Average Score"),
    y=alt.Y("region").sort(region_order).title(None),
    color=alt.Color("year:N").title("Year")
)

(lines + points).properties(
    width=400, height=300, title="Change in Average Score by Region (2023 to 2024)"
)

3. Gender Assessment Score by Ownership Type

In addition to industry-based analysis, we examine whether ownership structure affects the overall gender performance scores. The following box plot illustrates how scores vary by ownership category

filtered_data <- gender_data |>
  subset(!is.na(score) & !is.na(ownership))

ggplot(filtered_data, aes(x = ownership, y = score, fill = ownership)) +
  geom_boxplot() +
  scale_fill_manual(values = c(
    "Public" = "#1f77b4",     # blue
    "Private" = "#6baed6",    # lighter blue
    "Government" = "#08306b"  # dark blue
  )) +
  labs(title = "Score by Ownership Type", x = "Ownership", y = "Score") +
  theme_minimal() +
  theme(legend.position = "none")

filtered_data = gender_data.dropna(subset=['score', 'ownership'])

alt.Chart(filtered_data).mark_boxplot(size=80).encode(
    x=alt.X('ownership').title('Ownership').axis(alt.Axis(labelAngle=0)),
    y=alt.Y('score').title('Score'),
    color=alt.Color('ownership', legend=None)
).properties(width=450, height=350, title="Score by Ownership Type")

Hypothesis Testing: Gender Assessment Scores by Ownership Type

We aim to test whether companies with different ownership types have significantly different mean gender assessment scores.

Hypotheses

  • Null Hypothesis (H₀):
    The mean score is the same across all ownership groups.
    \[ \mu_{\text{public}} = \mu_{\text{Private}} = \mu_{\text{Government}} \]

  • Alternative Hypothesis (Hₐ):
    Not all group means are equal.

Assumption checks for ANOVA

Before applying ANOVA to compare gender assessment scores across ownership types, it is essential to verify that the assumptions underlying the test are met.

1. Independence of Observations

  • Observations within and across groups must be independent: This assumption is satisfied by the data set design, provided that each company appears only once. This assumption is generally satisfied by the data set design, provided that each company appears only once and that there is no clustering of related firms or repeated measurements. Since our data set includes one record per company without repeated measures, we consider the independence assumption to be reasonably met.

2. Normality of Residuals

ANOVA assumes that the residuals are normally distributed. We assess this using the Shapiro-Wilk test:

anova_model <- aov(score ~ ownership, data = gender_data)
shapiro.test(residuals(anova_model))

    Shapiro-Wilk normality test

data:  residuals(anova_model)
W = 0.98844, p-value = 1.423e-11
anova_model = smf.ols(
    "score ~ C(ownership)",   # C() specifies a categorical variable
    data=gender_data
).fit()
residuals = anova_model.resid
shapiro_test_stat, shapiro_p_value = shapiro(residuals)

print(f"W = {shapiro_test_stat:.4f}, p-value = {shapiro_p_value:.4e}")
W = 0.9884, p-value = 1.4227e-11

Inference

Since the p-value is less than 0.05, we reject the null hypothesis of normality. The residuals are not normally distributed, violating the normality assumption for ANOVA.

3. Homogeneity of Variances

Another key assumption is that the variance of scores is roughly equal across the ownership types. We use Levene’s Test to assess this:

leveneTest(score ~ ownership, data = gender_data)
Warning in leveneTest.default(y = y, group = group, ...): group coerced to
factor.
Levene's Test for Homogeneity of Variance (center = median)
        Df F value    Pr(>F)    
group    2  25.558 1.097e-11 ***
      1993                      
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
# create a list of scores per ownership and convert that series of lists into a lists of lists
groups = gender_data.groupby('ownership', observed=False)['score'].apply(list).to_list()

# perform Levene's test (* unpacks a list)
levene_test_stat, levene_p_value = levene(*groups)

print(f"F = {levene_test_stat:.4f}, p-value = {levene_p_value:.4e}")
F = 25.5580, p-value = 1.0972e-11

Inference: As the p-value is well below 0.05, we reject the null hypothesis of equal variances. The assumption of homogeneity of variance is violated.

Since both the normality of residuals and homogeneity of variance assumptions are violated, a non-parametric alternative to ANOVA is more appropriate.

4. Kruskal-Wallis test

To evaluate whether the score differs across different types of business ownership, we use the Kruskal–Wallis test.

Hypotheses

  • Null Hypothesis (H₀):
    The median score is the same across all ownership groups.
    \[ \tilde{\mu}_{\text{public}} = \tilde{\mu}_{\text{Private}} = \tilde{\mu}_{\text{Government}} \]

  • Alternative Hypothesis (Hₐ):
    Not all group medians are equal.

Assumptions of the Kruskal-Wallis Test

  • Ordinal or Continuous Dependent Variable: The variable score is numeric, satisfying this assumption.

  • Independent Groups: Each business (or observation) belongs to only one ownership type. This ensures group independence.

  • Independent Observations: Each row in the data set represents a unique business entity, so the observations are independent.

kruskal.test(score ~ ownership, data = gender_data)

    Kruskal-Wallis rank sum test

data:  score by ownership
Kruskal-Wallis chi-squared = 517.7, df = 2, p-value < 2.2e-16
kruskal_test_stat, kruskal_p_value = kruskal(*groups)

print(
    f"Kruskal-Wallis chi-squared = {kruskal_test_stat:.4f}, p-value = {kruskal_p_value:.4e}"
)
Kruskal-Wallis chi-squared = 517.7035, p-value = 3.8204e-113
  • The p-value is small (< 0.05), meaning we reject the null hypothesis (H₀) that all ownership groups have the same median score.
  • There are statistically significant differences in median score between at least two ownership types.

Since the test indicates that not all ownership groups are equal, we need to conduct a post-hoc test to identify which specific groups differ.

5. Dunn’s Post-Hoc Test (with Bonferroni Correction)

To determine which ownership groups differ, we conduct Dunn’s test with Bonferroni adjustment for multiple comparisons.

# Run Dunn's test with Bonferroni correction
dunnTest(score ~ ownership, data = gender_data, method = "bonferroni")
Warning: ownership was coerced to a factor.
Warning: Some rows deleted from 'x' and 'g' because missing data.
Dunn (1964) Kruskal-Wallis multiple comparison
  p-values adjusted with the Bonferroni method.
            Comparison          Z      P.unadj        P.adj
1 Government - Private  -4.232728 2.308732e-05 6.926197e-05
2  Government - Public -18.867414 2.113945e-79 6.341836e-79
3     Private - Public -16.052311 5.507422e-58 1.652227e-57
gender_data = gender_data.dropna(subset="ownership")

adjusted_p_values = sp.posthoc_dunn(
    gender_data, val_col="score", group_col="ownership", p_adjust="bonferroni"
)

Markdown(adjusted_p_values.to_markdown())
Government Private Public
Government 1 6.9262e-05 6.34184e-79
Private 6.9262e-05 1 1.65223e-57
Public 6.34184e-79 1.65223e-57 1

Inference

Pairwise comparisons using Dunn’s test (with Bonferroni correction) indicate statistically significant differences in gender assessment scores across all ownership types:

  • Government-owned companies have significantly lower scores than both Private and Public companies.

  • Private companies also score significantly lower than Public companies.

Discussion

This analysis highlights notable disparities in gender assessment performance across industries, ownership types, and regions. Industries demonstrated varying levels of engagement with gender equity, with the Personal & Household Products sector achieving the highest average scores. In contrast, the Pension Funds industry consistently lagged behind, pointing to potential structural or cultural barriers that may impede progress in specific sectors.

Ownership structure played a critical role in performance. Government-owned companies significantly outperformed both privately held and publicly traded firms. In contrast, publicly traded companies, despite their visibility and access to resources, showed relatively poor performance.

Despite the insights provided, this analysis faces some limitations. While average scores are useful for summarizing overall trends, they can hide meaningful differences within subgroup like firm size or revenue.

Attribution

Data sourced from World Benchmarking Alliance, available under a Creative Commons Attribution 4.0 International License. Original data set: 2024 Gender Assessment data set.