What EPSO numerical reasoning actually tests
The numerical reasoning test gives you 10 questions in 20 minutes — exactly 2 minutes per question. Each question presents a data table, chart, or statistical dataset (always EU-themed: Eurostat data, member state comparisons, budget allocations) followed by a calculation question with five answer choices.
The mathematics is secondary school level. The real test is data extraction under pressure: can you identify the correct row, column, and unit from a complex table in under 30 seconds, then perform the right calculation in the remaining 90 seconds?
For AD5 2026, numerical reasoning is a pass/fail gate — you must score at least 50% (5/10) to advance, but it carries zero ranking weight. For AST competitions, it contributes to the combined CBT score. Regardless of weight, failing it eliminates you from the competition entirely.
The five mathematical templates
Over 90% of EPSO numerical questions follow one of five templates. Recognising the template within 15 seconds transforms a 2-minute puzzle into a 60-second calculation.
Template 1: Percentage change
The most frequent template. You are given two values (typically from different years or countries) and must calculate the percentage increase or decrease.
Formula: Percentage change = (New value - Old value) / Old value × 100
Key trap: Always divide by the original (earlier, base) value. If GDP went from 200 to 250, the change is (250-200)/200 = 25%, not (250-200)/250 = 20%. EPSO places both 25% and 20% among the answer options.
Variation: "By how many percentage points did the rate change?" This is not percentage change — it is simple subtraction. If unemployment went from 8.2% to 6.7%, the answer is 1.5 percentage points (not -18.3%).
Template 2: Ratio and proportion
Compare two quantities from the table. "How many times larger is Germany's export volume compared to Portugal's?" or "What is the ratio of renewable energy to fossil fuel production?"
Formula: Ratio = Value A / Value B
Key trap: The question phrasing determines the direction. "How many times larger is A than B" means A/B. "What fraction of A does B represent" means B/A. Reversing the division is the most common error — and EPSO always includes the reversed answer among the options.
Template 3: Weighted average
This template appears whenever you see per-capita data alongside population data. You cannot simply average per-capita values — you must weight each by its population.
Formula: Weighted average = Σ(value × weight) / Σ(weights)
Example: Average CO2 emissions per capita across France (4.5 tonnes, population 67M), Germany (7.3 tonnes, population 83M), and Poland (7.6 tonnes, population 38M):
(4.5×67 + 7.3×83 + 7.6×38) / (67+83+38) = (301.5 + 605.9 + 288.8) / 188 = 6.36 tonnes
Key trap: The simple (unweighted) average is (4.5+7.3+7.6)/3 = 6.47. This will always be one of the wrong answer options. The difference between weighted and unweighted may be small — but it is always an available distractor.
Template 4: Compound Annual Growth Rate (CAGR)
The most mathematically demanding template. Given a starting value, an ending value, and a time period, calculate the constant annual growth rate.
Formula: CAGR = (End/Start)1/n - 1
Where n = number of years (intervals, not data points).
Critical trap — counting years: 2018 to 2022 = 4 years, not 5. You count the gaps between years, not the years themselves. EPSO includes both the 4-year and 5-year calculations among the answer options.
Second trap — simple vs compound: Total growth of 32% over 4 years is NOT 8% per year. The compound rate is (1.32)1/4 - 1 = 7.2%. The simple rate (32/4 = 8%) will always appear as a distractor.
Template 5: Percentage of total
"What percentage of the total EU budget does agricultural spending represent?" Identify the part and the whole, divide, multiply by 100.
Formula: Part / Whole × 100
Key trap: The total may not be given directly — you may need to sum all rows. Candidates who use the wrong denominator (using one category instead of the total) generate an answer that EPSO has pre-calculated as a distractor.
Second trap: Dividing whole by part instead of part by whole. If research is 18.7B out of 140.6B total, the answer is 18.7/140.6 = 13.3%. The reverse calculation (140.6/18.7 = 7.52) is meaningless but appears as an option.
Data extraction: the real bottleneck
Most numerical reasoning errors are not calculation errors — they are data extraction errors. You read the wrong row, confuse two column headers, or miss a unit conversion.
Table reading protocol
- Read the question first. Identify which variable, which country/year, and which calculation.
- Read the table title and units. "GDP in millions of euros" versus "GDP in billions" is a factor-of-1000 difference.
- Trace row and column with your finger (or cursor). Complex tables with 6+ columns cause mis-reads.
- Extract the exact values onto your scratchpad before calculating. Never calculate from memory.
Common unit traps
- Mixing thousands and millions (table says "in thousands" but you read the number as absolute)
- Percentages vs percentage points ("increased from 4% to 6%" is 2 percentage points, not 50%)
- Per capita vs absolute values (the question asks for total, but the table shows per-capita)
- Currency mismatch (one column in EUR, another in USD with a conversion rate in the footnote)
Calculator strategy
EPSO provides an on-screen calculator. Use it strategically — not for every operation.
- Use the calculator for: division with decimals, CAGR exponents, weighted sums with large numbers
- Mental arithmetic for: simple percentages (10%, 25%, 50%), rounding checks, order-of-magnitude estimates
- Estimation first: Before calculating precisely, estimate the answer range. If your calculation falls outside the estimate, you likely have a data extraction error.
Example: "What is the percentage change from 487 to 612?" Before calculating: 612 is about 25% more than 487 (because 25% of 500 is 125, and 487+125 = 612). Scan the options — if one says approximately 25.7%, that is likely correct. Verify with the calculator: (612-487)/487 = 25.67%. Confirmed.
Time management: the 120-second budget
| Phase | Time | Action |
|---|---|---|
| 1. Read question | 15 sec | Identify template type and required data |
| 2. Extract data | 30 sec | Find correct row/column, note values and units |
| 3. Estimate | 10 sec | Rough mental estimate of answer range |
| 4. Calculate | 45 sec | Apply formula, use calculator if needed |
| 5. Verify | 20 sec | Check result against estimate and answer options |
If you cannot identify the template within 15 seconds, skip and return. Spending 4 minutes on one question steals time from two easier ones.
Answer option analysis
EPSO answer options are not random — they are pre-computed wrong answers based on predictable mistakes. If you understand the traps, you can sometimes eliminate options without completing the full calculation.
- One option is the correct answer
- One option uses the wrong base (divides by the new value instead of the old)
- One option confuses simple and compound rates
- One option reverses the numerator and denominator
- One option uses the wrong number of years or wrong units
This means you can often narrow the field to 2-3 options before calculating, then verify precisely.
Competition-specific format
| Competition | Questions | Time | Role |
|---|---|---|---|
| AD5 (generalist) | 10 | 20 min | Pass/fail gate (no ranking weight) |
| AD7 (specialist) | 10 | 20 min | Pass/fail gate |
| AST | 10 | 20 min | Combined CBT score |
| CAST (FG III-IV) | 10 | 20 min | Combined CBT score |
Preparation strategy
Week 1-2: Learn the five templates. For each, do 20 untimed questions. Focus on accuracy, not speed. Identify which templates you find hardest.
Week 3-4: Timed practice. Start with 3 minutes per question, reduce to 2 minutes, then target 1:45. Log every error by category: data extraction, calculation, or template identification.
Week 5-6: Full mock tests (10 questions / 20 minutes) under exam conditions. No pausing, no going back, strict timer. Your target accuracy is 80%+ (8/10).
Final week: Focus on your weakest template. If CAGR consistently trips you, spend two days exclusively on compound growth questions. One template mastered is worth more than broad superficial review.
The three calculation shortcuts
- The 10% anchor. To find any percentage of a number, start with 10%. 10% of 847 = 84.7. Need 30%? Triple it: 254.1. Need 5%? Halve it: 42.35. This is faster than multiplying by 0.30 on the calculator.
- The complement method. "What percentage is NOT agricultural spending?" If agricultural is 34.2%, the complement is 100 - 34.2 = 65.8%. Faster than summing all other categories.
- The ratio shortcut. If A = 450 and B = 150, the ratio A/B = 3. But if A = 447 and B = 153, estimate: A/B is roughly 450/150 = 3, then check which option is closest to 2.92.
Want structured preparation?
Our training programs cover exactly the skills and techniques described in this article.
Start your preparation