6.2 Weighted Averages & Composite Group Statistics

Section 6.2: Weighted Averages & Composite Group Statistics

The Weighted Mean Formula and Rationale

In Civil Service governance, executive agencies, regional directorates, and operational divisions vary dramatically in scale. A frontline operational directorate in HM Revenue & Customs (HMRC) may employ 15,000 caseworkers, while an internal policy advisory unit in HM Treasury may comprise only 75 specialists. When consolidating performance indicators—such as staff engagement, operational compliance, or processing accuracy—evaluating departments on equal footing is statistically invalid.

To compute an accurate overall metric across disparate groups, each group's average must be multiplied by a factor representing its size, volume, or relative importance. This calculation is the weighted mean (denoted $\bar{x}_w$):

$\bar{x}_w = \frac{\sum (w_i \cdot x_i)}{\sum w_i}$

Where:

• $x_i$ is the value or average of subgroup $i$.
• $w_i$ is the weight (headcount, budget, transaction volume, or percentage importance) assigned to subgroup $i$.
• $\sum (w_i \cdot x_i)$ is the sum of the weighted values across all subgroups.
• $\sum w_i$ is the total sum of all weights.

If the weights are expressed as proportions or percentages that sum to 100% (or $1.00$), the formula simplifies directly to:

$\bar{x}_w = \sum (W_i \cdot x_i) \quad \text{where } \sum W_i = 1.00$

---

The Fatal Mistake: The Fallacy of Averaging Averages

The single most pervasive error encountered on the CSNT is the unweighted averaging of averages. Candidates frequently calculate the simple arithmetic mean of two or more group percentages, completely ignoring the relative sizes of the cohorts.

Realistic Government Scenario

Consider two executive agencies within a ministerial department reporting their annual staff satisfaction survey scores:

• Agency A: Has 1,000 staff and reports an 80.0% satisfaction score.
• Agency B: Has 200 staff and reports a 50.0% satisfaction score.

What is the department's true overall staff satisfaction level?

The Erroneous Approach (Averaging Averages):

$\text{Unweighted Average} = \frac{80.0\% + 50.0\%}{2} = \frac{130.0\%}{2} = 65.0\% \quad \text{[FATAL ERROR]}$

The Correct Approach (Weighted Mean):

To find the true percentage of satisfied staff across the entire department, we must calculate the absolute count of satisfied personnel:

• Satisfied staff in Agency A: $1,000 \times 0.80 = 800\text{ employees}$
• Satisfied staff in Agency B: $200 \times 0.50 = 100\text{ employees}$
• Total satisfied staff: $800 + 100 = 900\text{ employees}$
• Total staff across both agencies: $1,000 + 200 = 1,200\text{ employees}$

Applying the weighted mean formula: $\bar{x}_w = \frac{(1,000 \times 80.0) + (200 \times 50.0)}{1,000 + 200} = \frac{80,000 + 10,000}{1,200} = \frac{90,000}{1,200} = 75.0\% \quad \text{[CORRECT]}$

Why This Matters

Agency A is five times larger than Agency B ($1,000 : 200 = 5 : 1$). Therefore, Agency A represents $\frac{5}{6}$ (approx. 83.3%) of the department's workforce, while Agency B represents only $\frac{1}{6}$ (approx. 16.7%). The department-wide satisfaction score must be anchored much closer to Agency A's 80.0% than to Agency B's 50.0%. The naive calculation of 65.0% understates departmental satisfaction by 10 full percentage points.

[!WARNING] CSNT Rule of Thumb: You may ONLY take the simple arithmetic mean of subgroup averages if all subgroups contain the exact same number of observations ($w_1 = w_2 = \dots = w_k$). In every other instance, you must compute the weighted mean.

---

Multi-Department Composite Performance Scoring

Central government frameworks frequently evaluate public bodies using composite indexes, where multiple performance criteria are combined into a single score based on predefined policy weights.

Structure of a Performance Index

Consider the Civil Service Operational Capability Framework, which rates departmental executive agencies on a 0–100 scale across four core operational pillars:

1. Service Level Agreement (SLA) Clearance: Weight = 40% ($0.40$)
2. Budgetary Discipline & Value for Money: Weight = 30% ($0.30$)
3. Public Transparency & Governance: Weight = 20% ($0.20$)
4. Digital Transformation Progress: Weight = 10% ($0.10$)

$\text{Composite Score} = (0.40 \cdot S_{\text{SLA}}) + (0.30 \cdot S_{\text{Budget}}) + (0.20 \cdot S_{\text{Gov}}) + (0.10 \cdot S_{\text{Digital}})$

Worked Example 1: Comparing Departmental Composite Scores

Two executive agencies within the Home Office achieve the following audited scores across the four pillars:

Notice that if we calculate the simple arithmetic mean of each agency's scores:

• Agency Alpha: $\frac{85 + 70 + 60 + 90}{4} = \frac{305}{4} = 76.25$
• Agency Beta: $\frac{65 + 90 + 80 + 70}{4} = \frac{305}{4} = 76.25$

Both agencies appear to have identical overall performance under a naive unweighted evaluation. However, when we apply the ministerial weights:

Agency Alpha Composite Score: $\text{Score}_{\text{Alpha}} = (0.40 \times 85.0) + (0.30 \times 70.0) + (0.20 \times 60.0) + (0.10 \times 90.0)$ $\text{Score}_{\text{Alpha}} = 34.0 + 21.0 + 12.0 + 9.0 = 76.0$

Agency Beta Composite Score: $\text{Score}_{\text{Beta}} = (0.40 \times 65.0) + (0.30 \times 90.0) + (0.20 \times 80.0) + (0.10 \times 70.0)$ $\text{Score}_{\text{Beta}} = 26.0 + 27.0 + 16.0 + 7.0 = 76.0$

Now, suppose the Permanent Secretary issues an emergency directive restructuring the weights to prioritize frontline delivery during a crisis: SLA Clearance is elevated to 50%, Budgetary Discipline is reduced to 20%, Public Transparency remains 20%, and Digital Transformation is 10%.

Let us recalculate the updated composite scores:

• Agency Alpha: $(0.50 \times 85) + (0.20 \times 70) + (0.20 \times 60) + (0.10 \times 90) = 42.5 + 14.0 + 12.0 + 9.0 = 77.5$
• Agency Beta: $(0.50 \times 65) + (0.20 \times 90) + (0.20 \times 80) + (0.10 \times 70) = 32.5 + 18.0 + 16.0 + 7.0 = 73.5$

Under the revised priority framework, Agency Alpha outscores Agency Beta by 4.0 points (77.5 vs 73.5) because Agency Alpha excelled in the highest-weighted priority (SLA Clearance).

---

Grouped Frequency Distributions and Class Midpoint Estimation

When government databases aggregate large volumes of continuous numerical records—such as telephone wait times at the Department for Work and Pensions (DWP) or passport application processing days—the data is frequently published in grouped frequency tables (class intervals) rather than raw individual lists.

Because the exact individual values inside each interval are hidden, you cannot calculate the precise mean. Instead, you must calculate an estimated mean using the midpoint of each class interval.

The Midpoint Method Steps:

1. Identify the Class Midpoint ($m_i$): Calculate the average of the lower and upper boundaries of each interval: $m_i = \frac{\text{Lower Bound} + \text{Upper Bound}}{2}$
2. Multiply Midpoint by Class Frequency ($f_i \cdot m_i$): Treat all observations in that class as if they sit exactly at the midpoint.
3. Sum the Products and Divide by Total Frequency ($N$): $\text{Estimated Mean } \bar{x} = \frac{\sum (f_i \cdot m_i)}{\sum f_i} = \frac{\sum (f_i \cdot m_i)}{N}$

Worked Example 2: DWP Contact Centre Telephone Wait Times

A management information report logs the telephone wait times (in minutes) for 240 citizens contacting a disability benefits helpline:

Calculating the Estimated Mean: $\text{Estimated Mean Wait Time} = \frac{1,760.0}{240} = 7.33\text{ minutes} \quad (7\text{ minutes } 20\text{ seconds})$

Assumption Note: This estimation assumes that the individual wait times within each interval are evenly distributed around the midpoint. If the distribution within an interval is heavily skewed towards one boundary, the estimated mean will carry a slight estimation error.