How journalists tell stories with data

# How journalists tell stories with data
## 
### G. Elliott Morris , Data journalist and US correspondent, The Economist
### <class=‘date’>March 1, 2022 Prepared for a talk to the Data Analytics Symposium at Queens University, Charlotte, North Carolina</class>

---

---

## Background...

- From a small town on a barrier island in south Texas
  - Population 4,123 (2020)
  - 86% white, 7% Hispanic, 4% "two or more"
  - Median household income of $57,910
  - 11% poverty rate
- I went to college at the University of Texas at Austin
  - Studied government and history
  - With some statistics and computer science
  - Graduated in 2018, hired by _The Economist_ in the fall of my senior year
- I've always been interested in politics and policy, only really became fascinated with data in the second semester of college
  - Read a book called _Big Data_ by Kenn Cukier andViktor Mayer-Schönberger
  - In 2016, built a forecasting model for the US presidential primaries and general election
  - The rest, as they say, is history

---
class: middle, center, inverse

# [ Data ]  journalism

---
class: middle, center

# BIG DATA

---
class: middle, center

---
class: center

#### AT&T, Verizon, Comcast
#### FitBit, iRobot, Phillips, Sony
#### Tesla, Kia
#### Indeed

---
class: middle, center

---
class: middle, center

# 90%
### Of all data worldwide was created in the last 2 years
&nbsp; 
Google estimates there will be roughly 175 ZB of data 2025. That's about 175 billion times the size of the storage on your laptop.

---
class:  center

### Use #1. Actionable business insights

### Use #2. Telling stories?

---
class: center, middle, inverse

# [ Data journalism ]

---
class: center

# Data journalism today (perceptions):

## 1. Complex modeling

## 2. Large datasets "big(ger) data"

## 3. Resource-intensive learners and pipelines

---

---

.footnote[
_https://www.economist.com/graphic-detail/2021/04/10/our-house-price-forecast-expects-the-global-rally-to-lose-steam_
]

---

---
class: center, middle

# Data journalism today (for real):

## Most stories use small data

Datasets from academic studies

Single time-series from other companies or websites

Polls or social surveys

(Extracts of) Census data

---

## Data journalism today (for real):

&nbsp;

### The bread and butter of our daily work is improving stories with original data analysis and visualization.

---

---

---

---

---

---

---
class: center

## What do all of these examples have in common?

### 1. Examples of empirical journalism: or, "data in service of stories"1

### 2. Not "data science" or "computational journalism"

#### Lesson: Data journalism does not need to be complex to tell a story. In fact, most of the time, simpler is better.

---
class: middle, center

# " In fact, most of the time, simpler is better. "

&nbsp;

--
### Lower signal-to-noise ratio
### The "curse of dimensionality"

---
class: middle, center

---
class: middle, center

---
class: middle, center

# Data *analytics* is not data *science*

## Neither is data *journalism*

---

## Data *analytics* is not data *science*

**Data analysts** examine large data sets to identify trends, develop charts, and create visual presentations to help businesses make more strategic decisions.

**Data scientists** on the other hand, design and construct new processes for data modeling and production using prototypes, algorithms, predictive models, and custom analysis.

&nbsp;

Data (or "empirical") journalists do both of these things, but usually proceed from the former to the latter *only when necessary*
- and only when the data and modeling processes can supply enough precision.

---

## Data journalism in a nutshell

Most stories fall into one of 3 categories:

### 1. Small data that add context to a story
### 2. Visualizations explaining findings, meeting readers where they are
### 3. Complex statistical models or big(-ish) data that help us decipher relationships between variables and quantify uncertainty

---
class: middle, center

### 80% of stories fall in bucket 1-2

---

# Case studies

---

## Example 1: House redistricting

### Narrative:
<img src="figures/wasserman_redistricting.png" width="50%" />

---

## Example 1: House redistricting

### Data:

---

## Example 1: House redistricting

### Original analysis:

.footnote[
_https://www.economist.com/united-states/democrats-have-fared-surprisingly-well-in-congress-new-maps/21807593_
]

---

## Example 1: House redistricting

### Original analysis

.footnote[
_https://www.economist.com/united-states/democrats-have-fared-surprisingly-well-in-congress-new-maps/21807593_
]

---

## Example 2: News in 2021

.footnote[
_https://www.economist.com/graphic-detail/2021/12/18/2021s-biggest-stories-were-covid-19-and-americas-presidential-transition_
]

---
## Example 3: Covid-19 forecasts

Predictions for December 2021
<img src="figures/cdc_covid_nov21.png" width="100%" />
]

Actual values for Dec 2021
<img src="figures/cdc_covid_dec21.png" width="100%" />
]

---

## Example 3: Covid-19 forecasts

Predictions for January 2022
<img src="figures/cdc_covid_dec21.png" width="100%" />
]

Actual values for Jan 2022
<img src="figures/cdc_covid_jan22.png" width="100%" />
]

&nbsp;

Unpredictable = bad use of modeling for storytelling

---

## Example 3: Covid-19 forecasts

Think of alternative questions: not "what's going to happen next?" but "what's happening now?"

- Who isn't vaccinated?  Where are hospitals overrun? What counties are experiencing worse outbreaks?

---
class: center, middle, inverse

# Three principles

---

]

]

---

]

]

---

]

]

---

## Three principles

### 1. Explore, but don't blindly follow the data
### 2. Visualization *can* be the whole story!
### 3. Big data is not a silver bullet. Don't ask questions the data cannot answer.

&nbsp; 
&nbsp;

If you keep these three things in mind you will be a successful data analyst/journalist/scientist/whatever.

---

# Thank you!

&nbsp;

### Questions?

**Website: [gelliottmorris.com](https://www.gelliottmorris.com)**

**Twitter: [@gelliottmorris](http://www.twitter.com/gelliottmorris)**

_These slides were made using the `xaringan` package for R. They are available online at https://www.gelliottmorris.com/slides/_

]