6.1 Measures of Central Tendency & Spread

Section 6.1: Measures of Central Tendency & Spread

The Core Measures of Central Tendency & Dispersion

In Civil Service operational delivery, performance reporting, and policy analysis, numerical data is rarely uniform. Whether monitoring citizen wait times on telephone helplines, evaluating clearance rates for immigration tribunals, or analysing salary distributions across ministerial departments, analysts must synthesise thousands of individual data points into concise, representative summaries. These summary figures are known as measures of central tendency (which describe where the center of the data lies) and measures of dispersion or spread (which quantify the extent to which individual observations deviate from that center).

The Civil Service Numerical Test (CSNT) evaluates your command of four foundational statistical concepts:

1. The Arithmetic Mean

The mean (denoted as $\bar{x}$) is the most familiar measure of central tendency. It is calculated by dividing the sum of all observed values by the total count of observations:

$\bar{x} = \frac{\sum x}{n}$

Where:

• $\sum x$ represents the sum of all individual data values ($x_1 + x_2 + \dots + x_n$).
• $n$ represents the total number of observations in the dataset.

Key Operational Characteristic: Because every single observation contributes directly to $\sum x$, the mean is mathematically inclusive. However, this total inclusion makes the mean highly vulnerable to extreme values (outliers) or asymmetric distributions.

2. The Median

The median is the physical middle value of a dataset that has been arranged in ascending (or descending) order. It partitions the dataset into two equal halves: exactly 50% of observations fall at or below the median, and 50% fall at or above it.

To locate the median position ($P_{\text{median}}$) in an ordered list of $n$ items:

$P_{\text{median}} = \frac{n + 1}{2}$

• Odd Number of Observations ($n$ is odd): The median is the single value located at the integer position $\frac{n + 1}{2}$. For example, if $n = 7$, the position is $(7 + 1) / 2 = 4$; the 4th value in the ordered list is the median.
• Even Number of Observations ($n$ is even): The formula yields a fractional position ending in $.5$. The median is the arithmetic mean of the two central values located at positions $\frac{n}{2}$ and $\frac{n}{2} + 1$. For example, if $n = 8$, the position is $(8 + 1) / 2 = 4.5$; the median is $\frac{\text{Value}_4 + \text{Value}_5}{2}$.

Key Operational Characteristic: The median is resistant to outliers. Changing the magnitude of the highest or lowest numbers in the dataset does not alter the median, provided their rank order relative to the center remains unchanged.

3. The Mode

The mode is the data value that appears with the highest frequency in a dataset. A dataset may have:

• A single mode (unimodal).
• Two distinct modes (bimodal).
• Three or more modes (multimodal).
• No mode, if every observation appears with equal frequency.

Key Operational Characteristic: The mode is the only measure of central tendency that can be used with non-numerical, qualitative, or categorical data. For example, if the Driver and Vehicle Licensing Agency (DVLA) records citizen enquiry channels as Online Portal, Telephone, Post, and Email, the "modal contact method" identifies the single most common channel utilized by the public.

4. The Range

The range is the simplest metric of statistical spread or dispersion. It measures the absolute distance between the two extreme endpoints of the dataset:

$\text{Range} = x_{\max} - x_{\min}$

Key Operational Characteristic: The range indicates the overall boundary span of the observations. However, because it relies exclusively on the two most extreme values, it reveals nothing about how the remaining data points are distributed between those boundaries.

---

Mean vs. Median: The Strategic Choice in Public Administration

A recurring theme across CSNT questions is evaluating whether the mean or the median provides the more accurate, defensible representation of a public operational dataset.

Outlier Vulnerability and Skewness

Consider an operational immigration tribunal chamber where 5 caseworkers resolve cases. Four caseworkers resolve routine asylum status updates in 12, 14, 15, and 17 weeks. A fifth caseworker is assigned a high-profile judicial review involving foreign diplomatic entities that takes 92 weeks to conclude.

Let us evaluate the two measures for this 5-case dataset:

• Mean: $\bar{x} = \frac{12 + 14 + 15 + 17 + 92}{5} = \frac{150}{5} = 30.0\text{ weeks}$.
• Median: Ordered list: $[12, 14, 15, 17, 92]$. The middle value ($n = 5$, position 3) is 15.0 weeks.

Reporting that the "average case takes 30 weeks" is deeply misleading. Four out of the five cases were concluded in 17 weeks or fewer. The single 92-week outlier pulls the arithmetic mean upward by 15 weeks, creating an artificial impression of systemic departmental delay. The median of 15.0 weeks accurately reflects the typical casework duration experienced by the public.

Why the UK Civil Service and ONS Report Median Earnings

In the UK Annual Survey of Hours and Earnings (ASHE) published by the Office for National Statistics (ONS), headline earnings figures for public sector workers are universally presented as medians, not means. In wage distributions, a small fraction of senior civil servants and agency directors earn substantial salaries (e.g., Permanent Secretaries earning £160,000 to £200,000+), while the vast majority of operational delivery staff sit in administrative officer (AO) and executive officer (EO) bands (earning £24,000 to £34,000). Calculating the mean wage pulls the figure sharply upwards, distorting the standard remuneration of the civil service workforce. The median provides an unskewed, robust benchmark.

---

Finding Missing Values Using Target Means (Inversion Technique)

On the CSNT, questions frequently present a scenario where several known performance scores are given, and you must calculate what score an additional branch, team, or caseworker must attain to achieve an overall target average.

The Inversion Formula

To find an unknown missing value, invert the basic arithmetic mean formula:

1. Calculate the Total Required Points: Multiply the target mean by the total number of observations ($n$, including the missing observation): $\text{Total Required} = \text{Target Mean} \times n$
2. Calculate the Current Sum: Sum all known observations: $\text{Current Sum} = \sum_{i=1}^{n-1} x_i$
3. Isolate the Missing Value: Subtract the current sum from the total required: $\text{Missing Value} = \text{Total Required} - \text{Current Sum}$

Worked Example 1: Executive Agency Customer Satisfaction Target

An Executive Agency evaluates customer satisfaction across 5 regional operational hubs. Scores for 4 of the hubs have been validated:

• Hub North: 78%
• Hub Midlands: 82%
• Hub South: 71%
• Hub East: 85%

The Agency Chief Executive has established a mandatory target mean satisfaction score of 80.0% across all 5 regional hubs. What minimum satisfaction score must Hub West achieve to secure this target?

Step 1: Calculate the total required percentage points for all 5 hubs: $\text{Total Required} = 80.0 \times 5 = 400.0\text{ points}$

Step 2: Calculate the sum of the 4 known hubs: $\text{Current Sum} = 78 + 82 + 71 + 85 = 316.0\text{ points}$

Step 3: Solve for the required score of Hub West: $\text{Score}_{\text{West}} = 400.0 - 316.0 = 84.0\%$

Hub West must achieve at least 84.0% customer satisfaction.

---

Effect of Adding, Removing, or Transforming Observations

Understanding how dynamic changes in a dataset affect central tendency is crucial for CSNT multi-step questions.

1. Adding a New Observation

• Impact on the Mean:
  • If the new value $x_{\text{new}} > \bar{x}_{\text{old}}$, the mean increases.
  • If the new value $x_{\text{new}} < \bar{x}_{\text{old}}$, the mean decreases.
  • If the new value $x_{\text{new}} = \bar{x}_{\text{old}}$, the mean remains unchanged.
• Impact on the Median: Adding a value shifts the median position by 0.5 steps along the ordered distribution. An extreme value, regardless of whether it is 100 or 1,000,000, shifts the median by at most one position or into an adjacent midpoint.

2. Linear Transformations of Data

When operational policies apply uniform mathematical adjustments across an entire dataset, the statistical measures transform predictably:

• Adding a constant $c$ to every value (e.g., adding a flat £500 cost-of-living stipend to all salaries): $\bar{x}_{\text{new}} = \bar{x}_{\text{old}} + c$ $\text{Median}_{\text{new}} = \text{Median}_{\text{old}} + c$ $\text{Range}_{\text{new}} = \text{Range}_{\text{old}} \quad (\text{Range is completely unaffected})$
• Multiplying every value by a constant $k$ (e.g., applying a 5% across-the-board percentage increase, where $k = 1.05$): $\bar{x}_{\text{new}} = k \cdot \bar{x}_{\text{old}}$ $\text{Median}_{\text{new}} = k \cdot \text{Median}_{\text{old}}$ $\text{Range}_{\text{new}} = k \cdot \text{Range}_{\text{old}}$

Worked Example 2: Outlier Correction in Tribunal Backlogs

An immigration tribunal chamber records the appeals backlog resolution times (in weeks) for 8 cases: $14, 16, 18, 19, 21, 22, 24, 66$

Initial Dataset Calculations ($n = 8$):

• Ordered Array: $[14, 16, 18, 19, 21, 22, 24, 66]$
• Median Position: $\frac{8 + 1}{2} = 4.5$. The 4th value is 19; the 5th value is 21. $\text{Median} = \frac{19 + 21}{2} = 20.0\text{ weeks}$
• Sum: $\sum x = 14 + 16 + 18 + 19 + 21 + 22 + 24 + 66 = 200.0\text{ weeks}$
• Mean: $\bar{x} = \frac{200}{8} = 25.0\text{ weeks}$
• Range: $66 - 14 = 52.0\text{ weeks}$

Scenario Adjustment: A forensic case audit reveals that the 66-week case was recorded under clerical error; its true resolution time was 22 weeks.

Recalculation with Corrected Value:

• New Ordered Array: $[14, 16, 18, 19, 21, 22, 22, 24]$
• New Sum: $200 - 66 + 22 = 156.0\text{ weeks}$
• New Mean: $\bar{x}_{\text{new}} = \frac{156}{8} = 19.5\text{ weeks}$
• New Median: The 4th value remains 19; the 5th value is now 21. $\text{Median}_{\text{new}} = \frac{19 + 21}{2} = 20.0\text{ weeks}$
• New Range: $24 - 14 = 10.0\text{ weeks}$

Strategic Takeaway: Correcting the clerical outlier caused the mean to fall by 5.5 weeks (from 25.0 to 19.5) and the range to collapse by 42.0 weeks (from 52.0 to 10.0). However, the median remained completely unchanged at 20.0 weeks. This demonstrates the robust resilience of the median in operational reporting.