3.4 Rates, Staffing Metrics & Weighted Averages

Standardized Operational Rates and Public Metrics

In public sector administration, raw counts rarely provide an adequate basis for policy decisions or performance appraisal. A county reporting 8,000 social housing support applications appears to face higher demand than an adjacent county reporting 3,000 applications. However, if their populations are 400,000 and 100,000 respectively, standardizing the figures reveals that the smaller county experiences a higher rate of demand (30 per 1,000 residents) than the larger county (20 per 1,000 residents). Per-capita rates account for population size, but differences in eligibility, case mix, period or service coverage can still affect comparisons.

Per-Capita Rates on a Common Scale

A rate per a fixed population scale expresses the frequency of an administrative event relative to a uniform population denominator: $\text{Rate per } k \text{ Citizens} = \left(\frac{\text{Total Events or Cost}}{\text{Total Population}}\right) \times k$

Where $k$ is a standard scaling factor, such as 1,000, 10,000, or 100,000 citizens. In Irish local government and civil service audits, per capita metrics benchmark expenditure on public amenities, planning applications, and healthcare utilization.

---

Staffing Metrics and Workload Dynamics

Public service workforce management utilizes Whole-Time Equivalent (WTE) metrics to standardize full-time and part-time personnel. Key operational metrics include:

1. Caseload-to-Staff Ratio: $\text{Caseload Ratio} = \frac{\text{Active Caseload}}{\text{Total WTE Staff}}$ This gives active files per WTE, which differs from files per person when staff work different hours.

2. Operational Throughput Rate (Processing Velocity): $\text{Processing Speed} = \frac{\text{Applications Processed}}{\text{Caseworkers} \times \text{Working Days}}$ For equal working days and constant staff numbers, this gives average output per officer-day. Use total officer-days or hours when staffing varies.

3. Backlog Clearance Velocity: When new applications arrive continuously while a team processes existing files, backlog clearance is determined by the net clearance rate: $\text{Net Daily Clearance Rate} = \text{Gross Daily Processing Capacity} - \text{Daily Inflow of New Cases}$ $\text{Days to Eliminate Backlog} = \frac{\text{Current Backlog Size}}{\text{Net Daily Clearance Rate}}$

Use the clearance-time formula only for a positive net rate under constant inflow and capacity. At zero net clearance the existing backlog stays constant; at a negative net rate it grows. Round up if the question requires a whole number of complete working days.

---

Simple Average vs. Weighted Average

The distinction between a simple arithmetic average and a weighted average depends on the quantity the question asks you to average.

The Mathematical Fallacy of the Simple Average

A simple average assigns identical importance (equal weight) to every data point, regardless of the size of the unit it represents: $\bar{x} = \frac{\sum x_i}{n}$

In contrast, a weighted average scales each value according to its proportion of the total volume, headcount, or financial expenditure: $\bar{x}_w = \frac{\sum (w_i \times x_i)}{\sum w_i}$ For a pooled mean or rate, $w_i$ must represent its appropriate denominator, such as employees for average salary or applications for an approval rate. Use non-negative weights with a positive sum. Arbitrary weights do not recover the pooled quantity.

Detailed Worked Scenario: Regional Intreo Processing Speeds

Consider four regional Intreo service centers assessing jobseeker claims. Each office reports its headcount of caseworkers and its average daily processing speed per caseworker:

Now compare the two calculations:

• Simple Average of Rates: $\bar{x} = \frac{14.0 + 18.0 + 22.0 + 26.0}{4} = \frac{80.0}{4} = 20.0 \text{ claims per officer/day}$

• True Weighted Average: $\bar{x}_w = \frac{\sum (w_i \times x_i)}{\sum w_i} = \frac{840 + 540 + 440 + 260}{120} = \frac{2{,}080}{120} \approx 17.33 \text{ claims per officer/day}$

Why the Simple Average Fails

The simple average (20.0) creates a substantial distortion, overstating true system productivity by over 15%. This occurs because Dublin Central employs half of all caseworkers (60 out of 120) but operates at the lowest rate (14.0). Simple averaging treats Waterford South's 10 caseworkers as equally significant as Dublin's 60 caseworkers.

---

Simpson’s Paradox: A Reversal after Aggregation

Simpson’s paradox can occur when a comparison within each subgroup reverses in the pooled data because the groups have different weights. A larger aggregate gap in the same direction as every subgroup comparison is not, by itself, a reversal.

Hypothetical Visa Application Example

Two divisions process Routine and Complex applications:

Alpha has the higher approval rate in both categories: 90% exceeds 85%, and 50% exceeds 45%. Yet Alpha has the lower overall rate: 58% is below 77%. The comparison reverses when the categories are pooled.

The weights explain the reversal. Routine applications, which have higher approval rates in both divisions, are only 20% of Alpha’s workload but 80% of Beta’s. The aggregate rates are:

$\text{Alpha}=0.20\times90\%+0.80\times50\%=58\%$ $\text{Beta}=0.80\times85\%+0.20\times45\%=77\%$

A simple average of the category percentages would instead give 70% for Alpha and 65% for Beta. Those are equally weighted category means, not either division’s actual pooled approval rate.

This example describes approval rates, not proof that one division’s staff are more efficient or that case mix causes a particular decision. Check denominators and subgroup comparisons before drawing conclusions about overall performance. Where relevant information is missing, state the limit rather than infer a cause.

Reference: UC Berkeley statistical teaching on confounding and Simpson’s paradox. The numerical example here is original hypothetical teaching data.