DATA VISUALIZATION — COMPLETE NOTES

1. Introduction to Data Visualization

Data visualization is the graphical representation of information using charts, graphs, maps, and other visual tools. It helps audiences
understand complex datasets quickly by revealing patterns, trends, and relationships hidden in raw data.

Why it matters (especially in Economics & Social Sciences)

 Large and complex datasets require visualization for meaningful interpretation.

 Makes reports clearer, more appealing, and more persuasive.

 Helps communicate evidence effectively to both technical and non-technical audiences.

 Supports storytelling, research communication, and policy analysis.

Brief History

 Early examples include maps, astronomical charts, and the first statistical graphics (17th–19th centuries).

 19th–20th centuries: development of principles involving the use of position, color, shape, and invention of many chart types.

 Recent decades: rapid growth due to computing, interactive visualization, and user-friendly software.

What Makes a Visualization Effective?

1. Informative – Conveys key insights clearly.

2. Efficient & Clear – Uses correct scales, labels, chart types; avoids ambiguity.

3. Attractive – Aesthetic but not distracting; draws attention to important points.

4. Interactive (modern) – Allows filtering, drilling down, exploring data.

Relevance for Economics and Policy

 Helps analyse trends (GDP, inflation), compare groups (rural/urban), and examine relationships (education–income).

 Clear visuals strengthen arguments in research papers, presentations, and policy briefs.

2. Data Types

Understanding data types is essential for correct analysis, visualization, and interpretation.

A. Main Classification

1. Quantitative (Numerical) Data

 Represent amounts or magnitudes.

 Arithmetic operations (mean, ratios) are meaningful.

 Examples: income, age, marks, GDP, price.

Subtypes:

1. Continuous

o Can take any value in an interval.

o Examples: height, wage rate, interest rate.

o Visuals: histogram, density plot, line chart, boxplot.

2. Discrete

o Countable values, often integers.

o Examples: number of children, number of firms.

 o Visuals: bar chart, dot plot.

2. Qualitative (Categorical) Data

 Represent groups or labels.

 Usually analysed using counts or percentages.

 Examples: gender, occupation, state, sector.

Subtypes:

1. Nominal

o No natural order.

o Examples: region, religion, brand.

o Visuals: bar/pie chart.

2. Ordinal

o Ordered categories; distances not defined.

o Examples: satisfaction (low/medium/high), education level.

o Visuals: ordered bar chart, stacked bars.

B. Measurement Scales

1. Nominal – Only labels; no ordering.

2. Ordinal – Ordered categories; no equal spacing.

3. Interval – Equal intervals, arbitrary zero (example: °C temperature).

4. Ratio – True zero; all arithmetic meaningful (example: income, quantity).

Why important?

 Prevents mistakes (e.g., averaging nominal codes).

 Determines valid statistical summaries and visualizations.

C. Typical Questions by Data Type

 Numerical: What is the average? Is there a trend? Relationship between variables?

 Categorical: How many in each group? What are the percentages?

 Mixed: How do numerical values differ across categories (e.g., wages by education level)?

D. Choosing Visuals Based on Data Type

Data Situation Suitable Visuals

Single numerical variable Histogram, boxplot, density

Single categorical variable Bar chart, pie chart

Numerical vs numerical Scatterplot

Numerical vs categorical Grouped bar/boxplots

Data Situation Suitable Visuals

Categorical vs categorical Mosaic plot, stacked bars

3. From Data to Visualization

This process explains how raw data becomes a well-designed graph.

1. Key Steps

Step 1: Clarify the purpose

 What question are we answering?

 Who is the audience (student, policy-maker, general public)?

Step 2: Understand the data

 Identify variable types.

 Check missing values, outliers, errors.

Step 3: Choose the relationship to show

 Distribution (one variable)

 Comparison across categories

 Association between variables

 Change over time

 Composition (parts of a whole)

Step 4: Select appropriate graph

 Distribution: histogram, density, boxplot

 Comparison: bar chart, dot plot

 Relationship: scatterplot

 Time series: line graph

 Composition: stacked bars, area chart

2. Data → Visual Encodings

Visual elements used to represent data:

Position

 Most accurate representation of quantity (e.g., bar height, point position).

Length, Angle, Area

 Length is easy to compare (bar charts).

 Angles (pie charts) harder to interpret.

 Area/bubbles often misleading unless necessary.

Color

 Hue differentiates categories.

 Lightness/darkness shows order or magnitude.

  Use color-blind-friendly palettes.

Shape & Line Type

 Used to differentiate groups (e.g., dotted vs solid line).

3. Principles of Good Visual Design

1. Show data clearly

o Avoid 3D charts, heavy shading, misleading axes.

2. Declutter

o Remove unnecessary gridlines, borders, duplicate labels.

3. Emphasize what matters

o Use color or annotations for key points.

4. Self-explanatory graph

o Title with a message (e.g., “Inflation fell after 2017”).

o Proper axis labels, units, legend.

5. Consistency & readability

o Same color scheme, scale, fonts across graphs.

o Adequate white space.

4. Common Pitfalls

 Wrong chart for data type.

 Too many categories in one chart.

 Blind use of software defaults.

 Poor labeling or misleading scales.

 Ignoring audience needs.

4. Data Storage

Data storage refers to how datasets are saved, structured, and prepared for analysis and visualization.

A. Common Storage Formats

1. Text-Based Formats

 CSV, TSV

 Simple, lightweight, widely used.

 No metadata; everything imported as text.

2. Spreadsheet Formats

 Excel (.xlsx, .xls), Google Sheets

 Easy to use but prone to messy formatting (merged cells, inconsistent types).

 Excel can unintentionally change dates/codes.

3. Statistical Software Formats

 R: .rds, .RData

 Stata: .dta

 SPSS: .sav

 Preserve data types, labels, coding information.

4. Database Formats

 SQL (MySQL, PostgreSQL, SQLite)

 HDF5, Feather for large datasets

 Efficient for storing and querying millions of rows.

B. Data Structures for Visualization

1. Tidy (Long) Format

 Each variable = column

 Each observation = row

 Each table = one type of unit

 Best for R/ggplot2, Python/matplotlib, seaborn.

2. Wide Format

 Multiple variables spread across columns (e.g., income_2010, income_2011).

 Good for spreadsheets; must be reshaped for visualization.

C. Principles for Good Storage

 Rectangular format (no merged cells or totals inside data).

 Consistent column names (no spaces or special characters).

 Correct data types (numeric, date, categorical).

 Clear handling of missing values (e.g., NA).

 A separate data dictionary/README.

D. Economics-Specific Examples

Source Format Problem Solution

NSSO/RBI CSV/Excel Multiple sheets, coded categories Read specific sheets; recode

World Bank CSV/API Metadata rows, wide data Skip rows; convert to long

Administrative data Stata/Excel Labels stored separately Merge metadata

Panel data CSV/Stata ID–year mismatch Reshape to tidy format

E. Tools for Handling Formats

R

 readr (CSV), readxl (Excel), haven (Stata), DBI (databases).

Python

 pandas (all formats), openpyxl (Excel).

Spreadsheet Tools

 Excel, Google Sheets, LibreOffice (for small tasks).

LESSON 2 — From Data to Visualization & Principles of Good Visual Design

1. The Power and Purpose of Data Visualization

Modern decision-making depends on evidence, but raw data (rows, tables, numbers) cannot be interpreted quickly or easily. Visualization
converts numerical information into visual structures that the brain processes rapidly.

Why Visualization Matters

 The human brain processes visuals 60,000× faster than text.

 Helps detect patterns, clusters, anomalies, and trends without reading thousands of numbers.

Four Core Purposes

1. Sense-making – finding structure in data.

2. Communication – conveying insights to non-technical audiences.

3. Memory retention – visuals are remembered more easily than numbers.

4. Persuasion through evidence – visuals strengthen arguments when honest.

Golden Rule

A visualization must reveal the truth, not decorate it.

2. Understanding Data: The Raw Material for Visualization

Correct visualization depends on understanding the structure and measurement scale of data.

2.1 Types of Data by Structure

1. Cross-sectional
Data from many units at one time.
Example: expenditure of 500 households.

2. Time-series
Data from one unit over time.
Example: monthly GDP, daily stock prices.

3. Panel/Longitudinal
Many units tracked over time.
Example: productivity of 200 firms from 2010–2025.

Used in STATA, SQL, Pandas, etc.

2.2 Data Types by Measurement Scale

Scale Type Nature Valid Comparisons Examples

Nominal Categories, no order Equality only Gender, region

Ordinal Categories with order Order, not magnitude Low < Medium < High
Scale Type Nature Valid Comparisons Examples

Interval Numeric, equal intervals, no true zero Differences only Temperature °C

Ratio Numeric, true zero Differences + ratios Sales, income

Important

 Bar charts must start at zero for ratio data.

 Do NOT use bar-length comparisons for interval or ordinal scales—they imply false ratios.

2.3 Responsibility of Data Cleaning

A visualization built on dirty data is a confident lie.

Cleaning must include:

 Removing duplicates

 Handling missing values

 Fixing entry errors

 Standardizing units (₹, %, kg)

 Calculating summaries (mean, totals, growth)

 Transformations (log scale, indexing)

Tools: Pandas, Excel Power Query, SQL, R.

3. The Visualization Pipeline — Turning Numbers into Narrative

Visualization is a 5-step reasoning process:

1. Problem definition
→ What question must the data answer?

2. Data preparation
→ Is data clean and structured?

3. Chart mapping
→ Which graph answers the question best?

4. Design encoding
→ Which visual principles ensure accuracy?

5. Refinement
→ Remove clutter, improve readability, ensure honesty.

Insight drives design—not software templates.

4. Matching Questions to Graph Types

Before picking a chart, identify the question type.

4.1 Comparison Questions

Used to compare categories.

Chart Best for Limits

Bar/Column Category comparison Must start at zero

Chart Best for Limits

Lollipop Cleaner ranked comparison Less volume clarity

4.2 Trend-over-Time Questions

Chart Best for Limits

Line Chart Time patterns, seasonality Too many lines cause overlap

Area Chart Trend + magnitude Exaggeration if transparency misused

4.3 Distribution Questions

Chart Best for

Histogram Frequency distribution

Box Plot Median, quartiles, outliers

Density Plot Smooth distribution

4.4 Relationship Questions

Chart Best for

Scatter Plot Correlation, clusters

Bubble Chart Extra variable (size)

Heatmap Intensity of relationship

4.5 Composition Questions (Parts of Whole)

Chart Best for Limitations

Stacked Bar Proportion within groups Hard to compare segments

Pie/Donut 2–3 large groups Useless beyond 5 categories

5. How We Read Visuals — Encoding Channels

Visualization uses encoding channels to represent data.

Encoding Channel Accuracy Example

Position Highest Scatter points

Length Very high Bar height

Angle Low Pie chart slice

Area/Size Medium Bubble chart

Color intensity Good for heat Heatmaps

Encoding Channel Accuracy Example

Shape Good for categories Scatter plot categories

Position and length = best.

Angles and 3D = worst and most misleading.

6. Principles of Good Visual Design

These principles ensure clarity, honesty, and efficiency.

6.1 Principle 1 — Clarity

A good chart is self-explanatory.

 Avoid 3D effects

 Remove non-data ink (decorations, heavy gridlines)

 Prefer direct labels over legends

 Avoid clutter

6.2 Principle 2 — Honesty in Scale

Two strict rules:

1. Bars for ratio data must start at zero

2. Time-series lines must use proportional scales

Misleading axes exaggerate differences dramatically.

6.3 Principle 3 — Pre-attentive Attributes

Instantly noticed visual cues:

 Color

 Contrast

 Size

 Length

 Position

 Direction

Use them to highlight insights—not distract.

6.4 Principle 4 — Visual Hierarchy

Guide the viewer's eye:

1. Insightful title

2. Main visual pattern

3. Axes & labels

 4. Notes & source

Weak title: Company Sales Chart

Strong title: Sales increased 35% after discounting.

6.5 Principle 5 — Color Must Serve Meaning

 Use ≤5 colors

 Maintain consistent mapping

 Avoid red/green combinations

 Use color-blind safe palettes (ColorBrewer)

6.6 Principle 6 — Typography

 Minimum font size: 8 pt

 Avoid ALL CAPS

 Avoid vertical text

 Stick to one font family

6.7 Principle 7 — Simplicity

Ask:

 Can this be understood in 5 seconds?

 Does every element serve meaning?

Remove anything decorative.

6.8 Principle 8 — Provide Context

Include:

 Mild annotations

 Benchmarks

 Captions when needed

 Data source

6.9 Principle 9 — Avoid Illusionary Correlations

Do not imply relationships not backed by statistics.

Visual correlation ≠ real correlation.

Use regression, tests (t, F, χ²), confidence intervals, etc., when needed.

7. Common Visualization Mistakes

Mistake Why it happens Harm

Pie chart with 10 categories Software defaults People cannot compare angles

Bar axis not starting at zero To zoom-in visually False exaggeration

Too many colors Aesthetic attempt Confusion and loss of meaning

Using 3D charts Looks “fancy” Distorts perception

No insight title Focus on tool, not message Audience confused

Overplotting many trend lines Trying to show all data Noise > signal

8. Visualization Tools Ecosystem

 Google Sheets – beginners

 R (ggplot2) – economists/statisticians

 Python (Matplotlib, Seaborn) – data scientists

 Power BI – business dashboards

 Tableau, Adobe Illustrator – publication-quality visuals

9. Visualization Ethics

A chart carries moral responsibility because viewers rarely verify the underlying data.

Ethical rules:

1. No exaggeration/minimization of differences

2. Visual proportions must match numeric reality

3. Axes must respect measurement scales

4. Colors should not manipulate emotion

5. Patterns shown must match real data

10. Practical Example (Line Chart)

Dataset: Monthly revenue (₹ lakh)

Jan 48 • Feb 52 • Mar 55 • Apr 49 • May 63 • Jun 72

Analysis question: When did revenue peak and what is the pattern?

Correct visualization choice

 Line chart (time-series)

 Title: Revenue peaked in June; steady rise before summer demand.

 Label axes clearly

 Highlight the peak

 Remove unnecessary gridlines

11. Data Storytelling + Statistical Validation

Use visuals for communication but verify claims using statistics:

 OLS regression

 Hypothesis tests (t, F, χ²)

 Correlation matrices

 Elasticities

 Growth rates

 Confidence intervals

Visuals + statistics = credible insight.

12. Visualization Quality Rubric (100 Marks)

Criterion Marks

Clarity 20

Accuracy 25

Design discipline 20

Accessibility 15

Insightful title 20

Total 100

13. Final Guiding Principles

An effective visualization is:

 Insight-driven, not template-driven

 Cognitively precise

 Ethically honest

 Minimal but contextual

 Audience-friendly

 Readable without explanation

 Statistically validated when needed

Introduction

Economics is fundamentally a discipline of inference, explanation, and influence. Economists observe real-world phenomena through data
—prices, wages, production levels, consumption patterns, employment cycles, trade flows, firm growth, and policy impacts. Yet data alone
does not create knowledge. For data to become evidence, and for evidence to shape decisions, two essential transformations are required:

1. Technical transformation
Raw data must be cleaned, standardized, merged, reshaped, and validated.

2. Communicative transformation
Analytical insights must be communicated clearly—through visual design, narrative structure, cognitive accessibility, and honest
representation.
The first process is data wrangling.
The second is data persuasion—not emotional persuasion, but logical influence grounded in clarity, ethical visual design, and statistical
validation.

Economists do more than analyze datasets; they communicate policy implications, decipher market behavior, and translate numbers into
public understanding. A rich dataset poorly communicated may fail to influence; a modest dataset, properly wrangled and ethically
visualized, can reshape economic thinking, corporate strategy, or policy action.

2. Data Wrangling: The Backbone of Economic Communication

In academic exercises, students often receive clean, well-structured datasets. In real-world economic systems, data is messy and
inconsistent. Before analysis or visualization, data must undergo systematic transformation to eliminate distortions and reveal economic
truth.

2.1 The Real Characteristics of Raw Economic Data

Because economic data is human-recorded or institution-recorded, it often contains:

 Multiple names for the same entity (e.g., “U.P.”, “UP”, “Uttar Pradesh”)

 Separate files from different enumerators or agencies

 Different units for income, expenditure, output, tax, or trade flows

 Incorrect or inconsistent date formats

 Missing values due to non-response or recording failures

 Duplicates from multi-stage data exports

 Outliers—sometimes real economic shocks; sometimes errors

 Non-relevant variables that dilute analytical focus

Raw data may contain truth, but not meaning until wrangling provides structure.

2.2 Data Wrangling Toolkit for Economics Students

Conceptual understanding is essential, but practical execution depends on tools such as:

 Python (Pandas) for cleaning, transformation, reshaping

 Apache Spark for large-scale processing of national datasets

 SQL (PostgreSQL) for relational storage and merging

 Tableau / Power BI for visualization and dashboards

 Adobe Illustrator for print-quality academic charts

The tool enables automation—but the economist ensures correctness, consistency, and ethical discipline.

2.3 Stages of Data Wrangling in Economic Analysis

Stage 1: Data Acquisition

Common sources include:

 Household consumption surveys

 Public datasets from Kaggle or national repositories

 Primary surveys conducted through digital forms

 Official macroeconomic indicators from RBI, Census, or Planning Commission

Data must be credible, time-stamped, and comparable before being used.

Stage 2: Parsing and Format Standardization

Datasets appear in formats like:

 CSV

 Excel

 JSON

 API outputs

 HTML-scraped tables

 SQL exports

Incorrect parsing can produce misleading aggregates even when the data “looks correct.”

Stage 3: Cleaning the Data

Key tasks include:

1. Standardizing identifiers (e.g., names, regions)

2. Unit consistency (₹, lakh, crore; kg, tonne, etc.)

3. Standardizing date formats for time-series

4. Duplicate removal to avoid bias

5. Handling missing values using appropriate strategies

6. Correcting impossible entries (e.g., negative age)

7. Verifying outliers rather than blindly removing them

Goal: Fix noise, not behavior.

Stage 4: Reshaping the Data

Different analyses require different structures:

 Long format: suitable for time trends

 Wide format: ideal for category-level comparisons

 Regression-ready matrices: numeric and complete

 Policy segmentation: grouping by region, gender, income class

 Derived variables: elasticity, growth rate, ratio measures, index values

Stage 5: Data Merging and Joins

Economics often involves combining datasets, such as:

 Household income + consumption

 State GDP + employment

 Firm input + output data

  Trade flows + exchange rates

Merging requires correct keys (household ID, state-year, firm-year).

Incorrect joins produce visual lies, even if charts appear valid.

3. The Art of Persuading with Data: An Economist’s Communication Language

Persuasion in economics comes from logical clarity, not emotional appeal. Effective communication combines:

 Clear narratives

 Cognitive-friendly design

 Truthful visualization

 Statistical validation when needed

Economics persuasion appears in three forms:

stories, charts, and validated inference.

3.1 The Four Pillars of Data Persuasion

Pillar 1: Credibility

 Transparent sources

 Verified outliers

 Honest preprocessing

 No hidden imputations

 No fabricated or exaggerated visuals

Pillar 2: Cognitive Usability

The audience should instantly detect:

 Who is compared

 What increases or decreases

 How large differences are

 What pattern emerges

If comprehension takes > 30 seconds, persuasion is lost.

Pillar 3: Communication Aesthetics

Effective charts follow:

 Minimal clutter

 Honest axis scales

 Clear visual hierarchy

 Purposeful colors

 Horizontal labels

 Simple annotations
Elegance emerges from discipline, not decoration.

Pillar 4: Ethical Interpretation

 Bars start at zero for ratio-scale data

 Time-series slopes preserved

 No 3D distortions

 No false correlation implied without statistical proof

 No emotionally manipulative colors

3.2 The Persuasion Pipeline for Economists

1. Define the analytical question

2. Wrangle the data to answer the question

3. Choose the appropriate visualization

4. Encode the conclusion in the title

5. Label axes and units clearly

6. Add light annotations

7. Provide data source

8. Validate with statistical methods (e.g., OLS, t-tests) if relationships are claimed

4. Visualization Psychology in Economic Influence

Visual decoding is governed by cognitive accuracy.

Perception accuracy ranks as:

1. Position on scale (strongest)

2. Length

3. Area

4. Color grouping

5. Angle (weak—pie charts)

6. 3D depth (misleading)

Thus, economists rely on:

 Line charts → trends

 Bar/column charts → category comparison

 Scatter plots → relationships

 Stacked bars → composition

 Histogram/boxplot → distribution

5. Chart Library with Economics Use-Cases

1. Bar/Column Charts (Comparison)

Examples:

 Wheat output across states

 Sectoral employment shares

 Firm-level cost comparison

2. Line Charts (Trends)

Examples:

 GDP growth over quarters

 Price inflation cycles

 Seasonal fluctuations in demand

3. Scatter Plots (Relationships)

Examples:

 Income vs discretionary spending

 Price vs quantity (elasticity patterns)

4. Histograms/Boxplots (Distribution)

Examples:

 Wage distribution

 Consumption inequality

5. Stacked Bars (Composition)

Examples:

 Tax revenue sources

 Sector-wise GDP contribution

6. Economics Case Studies (Expanded)

Case Study A: Development Economics

Topic: Consumption inequality across Indian states

Issues:

 Mixed naming

 Missing income values

 Unit inconsistencies

 Duplicates

 Need for monthly aggregation

Wrangling:

 Standardized state names

 Converted income to ₹ lakh

 Median imputation

 Duplicate removal

 Reshaping to long format

Message encoded:
“High-income households spend 3× more on discretionary festival goods; demand rises sharply post-monsoon.”

Visuals:

 Line charts for trends

 Bars for magnitude

 Scatter for validating relationships

Case Study B: Behavioral Economics

Issue:
Ordinal preference data was plotted using bar heights → creates false magnitude comparisons.

Correction:
Use ordered categories or scatter-like positions, not ratio-implying bar lengths.

Lesson:
Truthful design strengthens economic inference.

Case Study C: Firm Theory

Topic: Input-output efficiency over 10 years

Wrangling:

 Merged by firm-year

 Indexed outputs

 Filtered cost variables

 Created regression-ready frame

Statistical validation:
OLS regression confirmed correlation.

Encoded message:
“Digital adoption explains 50% of productivity divergence post-2019 (95% confidence).”