ECON 0150 | Economic Data Analysis

The economist’s data analysis pipeline.

Part 1.3 | Relationships Through Time

Data Structures

…three main relationships between data points.

The most effective summarization tool depends on the relationship between the data points.

	Cross-Sectional	Time-Series	Panel Data
Focus	Multiple units, one time point	One unit, many times	Multiple units, many time points
Shape	Wide format	Long format	Long format
Ex.	Household income, 2025	US GDP, 10 years	Household income, 10 years

> we’ve spent the first part of the class on cross-sectional data

> we’ll spend a bit of time on panel and geographic data later

Exercise 1.3 | Data Structures

Lets identify the variable type for each dataset.

• Dataset 1: household_incomes.csv
• Dataset 2: household_savings.csv
• Dataset 3: Monthly_Coffee_Prices.csv

Timeseries: Coffee Prices

What information should we use to set prices in January 2026?

Timeseries: Coffee Prices

What information should we use to set prices in January 2026?

> it’s difficult to know… do we choose the mode?

> lets just plot the price against time

Timeseries: Coffee Prices

What information should we use to set prices in January 2026?

> lets indicate with a line that these points are in squence

Timeseries: Line Graph

What information should we use to set prices in January 2026?

Timeseries: Trends

Do you notice a trend in price?

> there was a positive trend in 2021

> we can zoom out to get a bigger picture

Timeseries: Trends + Subtrends

Do you notice a trend in price?

> how have prices changed since 2000?

> prices have increased somewhat, with many periods of decrease

Timeseries: Background Shading

What information should we use to set prices in January 2026?

> with background shading its easier to see periods with a negative trend in price

Exercise 1.3: Timeseries

Lets use a linegraph to examine the trends in coffee prices.

• Data: Coffee_Prices.csv

Exercise 1.3: Timeseries

# Lineplot
sns.lineplot(prices, y='price', x='date')

Timeseries

What information should we use to set prices in January 2026?

> could there be seasonal trends within the larger trend?

Seasonality: January

What information should we use to set prices in January 2026?

> a boxplot gives us a picture of the prices just in January

> lets compare this to other months

Seasonality: Monthly Boxplots

In addition to the overall trend, are there monthly patterns?

> lets be more specific…

Seasonality: Monthly Boxplots

In which month was the record highest price set?

Seasonality: Monthly Boxplots

In which month was the record highest price set?

> look at the maximums

Seasonality: Monthly Boxplots

In which month was the record highest price set?

Seasonality: Monthly Boxplots

In which season are prices most spread out?

Seasonality: Monthly Boxplots

In which season are prices most spread out?

> look at the ranges

Seasonality: Monthly Boxplots

In which season are prices most spread out?

Seasonality: Multi-Boxplot

What is the trend in median price?

> look at the medians…

Seasonality: Multi-Boxplot

What is the trend in median price?

> look at the medians… pretty difficult to see

Seasonality: Quartile Lineplot

What is the trend in median price?

Seasonality: Quartile Lineplot

What is the trend in median price?

Seasonality: Quartile Lineplot

What is the difference between the largest and the smallest median price per pound?

> something like $1.30 - $1.21 = $0.09

Timeseries: Summary

Linegraphs show trends; multi-boxplots show between-period patterns.

• Use a linegraph to show a numerical variable through time.
• Highlight changes in a linegraph using shading.
• Use a multi-boxplot to show the distribution between multiple periods.

Exercise 1.3: Seasonality

Lets use a multi-boxplot to examine the seasonal patterns of coffee prices.

• Data: Coffee_Prices.csv

Exercise 1.3: Seasonality

# Multi-Boxplot
sns.boxplot(prices, y='month', x='price', whis=(0,100))