Day 4 · 09:30 · 90 min

Visualising the headlines

Correlation, causation, and the press release

Day 4 · 09:30

A headline is a chart with the axes removed. Your job, as the person who still has the axes, is to put practical significance back in the room before a press release becomes a strategy.

You will leave able to

  • Separate statistical significance from practical significance.
  • Read a correlation heatmap without over-claiming.
  • Rewrite a media claim as a caption you would sign.

The original lesson used social-media and adolescent-sleep headlines. The mechanism is identical in a Nigerian financial-services EXCO. A typical morning pack:

  • *CBN hike tanks bank profits.*
  • *Digital share soars as branches die.*
  • *Naira slide drives NPLs.*

Each of those sentences asserts causation. The underlying tables are almost always correlations, often small ones, often estimated on a short window, often without a competing explanation (mix shift, one-off fee income, a single name in the NPL book).

Visualise the strength, not the stars

Download macro-correlations.csv· pairwise r for a fictional bank book
R
library(tidyverse)
corr <- read_csv("macro-correlations.csv")
corr %>%
ggplot(aes(var_x, var_y, fill = r)) +
geom_tile() +
geom_text(aes(label = sprintf("%.2f", r)), size = 3) +
scale_fill_gradient2(low = "#B8963E", mid = "#F4F0E8",
high = "#0B1F3A", midpoint = 0) +
coord_fixed() +
labs(
title = "What actually moves with what",
subtitle = "Pairwise correlations — not a causal map",
x = NULL, y = NULL
) +
theme(axis.text.x = element_text(angle = 45, hjust = 1))
policy rate
fx ngn usd
npl ratio
nim
fee income
cost income
loan growth
digital share
1.0
0.4
0.2
0.5
-0.1
0.1
-0.2
0.1
0.4
1.0
0.3
0.1
0.1
0.3
-0.2
0.2
0.2
0.3
1.0
-0.2
-0.0
0.4
-0.3
-0.1
0.5
0.1
-0.2
1.0
-0.2
-0.1
0.1
0.1
-0.1
0.1
-0.0
-0.2
1.0
-0.3
0.1
0.4
0.1
0.3
0.4
-0.1
-0.3
1.0
-0.1
-0.3
-0.2
-0.2
-0.3
0.1
0.1
-0.1
1.0
0.2
0.1
0.2
-0.1
0.1
0.4
-0.3
0.2
1.0
policy
fx
npl
nim
fee
cost
loan
digital
Correlation heatmap: CBN policy rate, FX, NPL, NIM and digital share of transactions. A statistically significant r of 0.08 is a pale tile — not a headline.

Read the tiles, not the folklore. Policy rate versus NIM is a real relationship in this book (r ≈ 0.48). Policy rate versus digital share is a pale tile (r ≈ 0.05). A slide that treats them as equal ‘impacts’ is how a digital budget gets cut after a hike.

Rewrite the headline

HeadlineWhat the data can supportWhat to put on the slide
CBN hike tanks profitsNIM co-moves with policy rate; fee income does notNIM vs policy rate, last 16 quarters, r = 0.48. Fee income uncorrelated.
Digital kills branchesDigital share vs cost-income is negative and modestDigital share and cost-income, r = −0.29. Mix, not a funeral.
FX drives NPLsFX vs NPL r ≈ 0.33 — real, not destinyName concentration and sector exposure still do more work.

Exercise 4.1

A press release you would sign

Pick one headline from this morning’s business pages (or invent a plausible one in your sector). Write: (1) the implied causal claim, (2) the association you could actually estimate, (3) a ggplot title and subtitle that refuse to over-claim.

  • If the verb is ‘drives’, ‘tanks’, ‘soars because’, it is causal.
  • A good subtitle names the window and the unit: *quarterly, 2022–2026, one bank book*.