Salary Analysis: Spanish Data Roles Job Market 2025

Salary analysis is one of the most sought-after insights from any job market EDA — and one of the most methodologically fraught. This page documents how salary_clean was constructed, what coverage looks like across sources, and how salary distributions vary by role, geography, and data source. It also explains the key caveats that must be kept in mind when interpreting any numeric salary figures from this dataset.

Coverage Summary

Salary data is not universally available across the three sources that make up the unified dataset. After cleaning and midpoint conversion, 69.46% of the 1,542 unified offers include a usable salary_clean value. The remaining 30.54% have no salary information.

69.46% — With Salary

Offers where salary_clean was successfully extracted and converted to a comparable EUR numeric value.

30.54% — No Salary

Offers where salary was absent, described qualitatively (e.g., “competitive”), or could not be reliably parsed.

df_jobs (international/US salaries) significantly inflates the overall salary distribution compared to df_tecno (Spanish market offers). df_tecno (Tecnoempleo) has approximately 78% null salary rate — the vast majority of its listings do not publish a salary figure, which is a known characteristic of Spanish job boards. By contrast, df_jobs (the international source) has near-complete salary coverage but contains predominantly non-Spanish, US-market salaries that are substantially higher than Spanish equivalents. Any aggregate salary statistic computed over the full unified dataset is therefore weighted towards df_jobs and should not be read as representative of Spanish market compensation. Always stratify by source_dataset before drawing salary conclusions.

The `salary_clean` Methodology

Raw salary data arrives in a variety of formats across the three sources: annual ranges in EUR, hourly rates, ambiguous “up to X” figures, and sometimes salaries in USD or GBP. The cleaning pipeline harmonises these into a single salary_clean column using the following logic:

Parse the raw salary string

Extract numeric values from salary text. Handles formats such as €40,000 - €55,000/year, $100k+, up to 60.000€, and similar patterns.

Convert to annual EUR

Hourly and monthly rates are annualised using standard multipliers. Non-EUR currencies are converted using approximate exchange rates at the time of data collection.

Compute the midpoint for ranges

For offers stating a salary range (e.g., €100,472 - €200,938), salary_clean is set to the arithmetic midpoint — in this example, €150,705. This is an approximation of expected compensation, not a guaranteed figure.

Flag outliers

The salary_clean_outlier boolean column is set to True for values that fall outside a statistical threshold (typically defined using IQR or z-score methods). These outliers are excluded from comparative analyses to prevent distortion.

salary_clean is an estimate, not an exact salary. For range-based offers, the actual salary offered to a candidate could be anywhere within the published range. The midpoint convention is a reasonable central estimate but should not be treated as a precise figure.

Source Differences

The three data sources have structurally different salary profiles, and this difference must be understood before making any cross-source comparisons.

df_jobs — International source (data_science_job_posts_2025.csv)

942 records total; approximately 143 explicitly Spanish entries.
Near-complete salary coverage — the majority of records include a parseable salary field.
International scope means many salaries are from the US, UK, or other high-wage markets.
USD-denominated salaries converted to EUR will appear inflated relative to Spanish-market offers.
Effect on aggregates: This source pulls the overall salary distribution significantly upward.

df_tecno — Spanish market source (tecnoempleo_spain_2026.csv)

Approximately 600 records, exclusively Spanish market.
~78% null salary rate — most Tecnoempleo listings omit salary information entirely.
Where salary is present, figures reflect the actual Spanish labour market and are generally lower than df_jobs equivalents.
Effect on aggregates: Under-represented in salary analyses due to high null rate, despite being the most representative Spanish market source.

df_scraping — Adzuna scrape

Additional Spanish recent job offers, scraped from Adzuna.
Salary coverage varies; more complete than df_tecno but less so than df_jobs.
Focuses on current Spanish market offers, complementing the other two sources.

Direct salary comparisons between df_jobs and df_tecno should be made with extreme caution. The two sources differ not only in geographic scope but in how salaries are reported. Comparing aggregate statistics across these sources without stratification by source_dataset will produce misleading results.

Geographic Salary Variation

Salary distributions differ meaningfully by location. Madrid, as the dominant city in the dataset, shows a broad salary range reflecting the concentration of both local SMEs and large multinational employers. Initial analysis from Block 2 of Notebook 04 (02b_salario_rol.png) indicates:

Madrid: Highest median salary among Spanish cities, driven by multinational presence.
Barcelona: Second highest, with a notable startup and tech company premium.
Other cities: Generally lower median salaries, with smaller sample sizes making statistical inference less reliable.

Geographic salary analysis is most meaningful when restricted to offers from Spanish-market sources (df_tecno, df_scraping) and when the salary_clean_outlier flag is used to exclude extreme values. Using the full unified dataset without filtering will inflate apparent salaries for all cities due to international offers.

Salary by Role Family

The Block 2 visualisation (02b_salario_rol.png) also disaggregates salary by job_family. Preliminary patterns from the EDA include:

data_science_ai: Widest salary range; highest upper quartile, reflecting both junior ML roles and senior AI research positions.
data_engineering: Consistently strong median salary; less variance than data science.
analytics / BI: Lower median than engineering and science roles; compressed range.
data_management / governance: Present in smaller numbers; salary patterns less reliable due to sample size.

When analysing salary by role family, always stratify by source_dataset first. Because data_science_ai roles are over-represented in df_jobs (the international source), naive aggregation will make this family appear disproportionately well-compensated relative to what the Spanish market alone would show.

Working with Salary Data

The following code demonstrates how to load the cleaned dataset, check salary coverage, remove outliers, and produce a source-stratified salary summary:

import pandas as pd
from pathlib import Path

df = pd.read_csv(Path('data/clean') / 'jobs_all_clean.csv')

# Salary coverage
print(f"Salary coverage: {df['salary_clean'].notna().mean():.2%}")

# Filter out outliers for cleaner analysis
df_clean = df[~df['salary_clean_outlier']]
print(df_clean['salary_clean'].describe())

# Salary by source
print(df_clean.groupby('source_dataset')['salary_clean'].describe())

Recommended filtering workflow for salary analysis

For the most interpretable salary analysis of the Spanish market specifically, apply the following filters before aggregating:

import pandas as pd
from pathlib import Path

df = pd.read_csv(Path('data/clean') / 'jobs_all_clean.csv')

# Step 1: Keep only Spanish-market sources
df_es = df[df['source_dataset'].isin(['df_tecno', 'df_scraping'])]

# Step 2: Remove statistical outliers
df_es = df_es[~df_es['salary_clean_outlier']]

# Step 3: Drop rows with no salary data
df_es = df_es.dropna(subset=['salary_clean'])

# Step 4: Analyse
print(f"Remaining records: {len(df_es)}")
print(df_es['salary_clean'].describe())
print(df_es.groupby('job_family')['salary_clean'].median().sort_values(ascending=False))

Note that after applying all three filters to Spanish-market sources with high null salary rates, the resulting sample may be small. Interpret results accordingly.

Salary Visualisation Reference

Notebook 04 produces two salary-specific visualisations:

02_analisis_salarial.png — Overall salary distribution histogram and box plot for salary_clean, with and without outlier exclusion.
02b_salario_rol.png — Salary distribution broken down by job_family, showing median, IQR, and range for each role category.

Both visualisations are generated after applying the salary_clean_outlier filter to produce readable distributions. The raw (unfiltered) distribution is extremely right-skewed due to international USD-denominated offers at the high end.

Introducción

Notebooks

Datos y Datasets

Análisis y Resultados

Salary Analysis: Spanish Data Roles Job Market 2025

Coverage Summary

69.46% — With Salary

30.54% — No Salary

The `salary_clean` Methodology

Source Differences

Geographic Salary Variation

Salary by Role Family

Working with Salary Data

Salary Visualisation Reference

Build docs developers (and LLMs) love

Introducción

Notebooks

Datos y Datasets

Análisis y Resultados

Documentation Index

​Coverage Summary

69.46% — With Salary

30.54% — No Salary

​The salary_clean Methodology

​Source Differences

​Geographic Salary Variation

​Salary by Role Family

​Working with Salary Data

​Salary Visualisation Reference

Build docs developers (and LLMs) love

Coverage Summary

The `salary_clean` Methodology

Source Differences

Geographic Salary Variation

Salary by Role Family

Working with Salary Data

Salary Visualisation Reference