On Data

& Stats

Krisna Gupta

krisna.or.id/slides/cipsdata/

Krisna Gupta who? 👀

Some publications 😎

About today

  • Some intro
  • Various ways data is presented
  • Some stats
  • Economic data
  • Q&A

I assume y’all have never worked with data before
I try to allocate more for Q&A

Intro

Why?

  • Efficient & powerful to support your story.

    • you don’t really need to be a math&stat-savy.
  • Shows how good you are in understanding an issue.

  • Objective, most of the time.

  • Everyone use ‘em these days. Tough luck for data haters.

Data is getting mainstream

via GIPHY

Some of my fave data source

Some jargons

  • Cross-sectional data: contains a snapshot of many subjects/individuals (people, countries, firms, etc) in a given time.

  • Time-series data: one subject observed for a long(-ish) periods of time.

    • Good if trend is important.
  • Panel data: combination of the two.

Various ways data is presented

Cartesian plane

  • A 2-dimensional plane.

    • pay attention to the x & y axis
  • Generally 2 forms: bar/column and line.

  • line chart is perfect for time series data.

    • usually x=time, y=value. ex
  • bar chart usually for cross-section data.

    • can be for time series too if you have one individual. ex
  • on values: check out the units!

Tables

  • Not the best visualization but very flexible. (ex)[https://comtrade.un.org/data/]

    • easy to costumize with templates & other viz.
    • can store many variables.
  • The most mainstream tools are microsoft excel & Google sheet.

    • Hence It’s best to have a .csv or .xlsx or something similar.
  • I certainly prefer working with these formats (among others)

Tables

  • First row is usually shows variables.
  • Note that a machine-readable tables are the best!
    • i.e., it is better to have one row for variables.
  • Example of bad machine-readability

Units in the X, Y and columns

  • Never lose sight of the units of your value.

    • e.g., thousand or millions, kg or ton, etc.
  • Especially important if you use various data source.

    • BI’s trade data is in 1000USD while Comtrade’s in USD.
  • Always read what’s X and Y.

    • if you make the graph, always write what’s X and Y.

Some statistics

The need to aggregate

  • How to process an information of the income of 1 million people?

  • When we have data of 1 million people, it’s impractical to look at 1 million values.

  • We look for one number that represent these 1 million values.

    • this is usually the average (or mean).
  • We also need to understand how the value is distributed.

    • called standard deviation.

Normal distribution

  • If we group values, take frequency, then sort them, we can make a distribution plot.

  • We can make a smooth approximation of the distribution plot with functions.

  • The most famous distribution is the normal distribution

  • Normal distribution’s characteristics:

    • frequency is highest around the mean
    • the tail is skinny (i.e., frequency is very small in the extremes)

Normal distribution

Normal distribution

When to use median

  • Median is the value lying in the middle of the whole group if we sort the value.

  • If we have 1 million people:

    • Sort their income from lowest to highest.
    • Median is the income of the 500.000th person.
  • Median is often use in the presence of non-trivial number of extreme values (i.e., fat tail).

  • income is often not distributed normally, so median is better.

  • example in excel.

Statistic: Percentage distribution of household income in the U.S. in 2020 | Statista
Find more statistics at Statista

Economic data

Real vs nominal

  • We use currency to express many economic variables.

  • We can’t aggregate car + food.

    • we can 200 million + 50 thousand.
  • But really what we want is the car and the food, not the money.

  • We need to take into account change in prices (i.e., inflation)

Real vs nominal

  • Say a firm can make 1 car and 100 food in 2020.

  • The firm’s GDP is $1 \times 200 + 100 \times 0.05 = 205$

  • in 2021, car’s price is increase to 210, hence GDP becomes 215.

  • Increased GDP?

    • not really, cuz the firm still just produced 1 car and 100 food.

Real vs nominal

  • It’s easy to imagine the complexity of this stuff in reality.

  • One thing is clear though: we want to exclude increase in GDP from price effect.

  • To avoid price effect, we use 2020 price so we can compare 2020 GDP with 2021 GDP.

  • Real GDP = When we use old prices.

Real vs nominal

  • Obviously to keep comparing, we still need to use 2020 prices when we calculate 2022 GDP.

  • also when we calculate GDP in 2023, etc.

  • Because we keep using 2020 prices, we say ‘constant price’.

    • 2020 is called ‘reference year’.
    • nominal GDP is calculate using ‘current price’.
    • nominal GDP = constant real GDP in the reference year.
  • The constant price changes from time to time.

GDP vs GDP per capita

  • GDP is an aggregate of the whole economy.

    • used to show how big & important the country is.
  • GDP per capita is the mean/average

    • used to reflect living standard & productivity.
  • Singapore vs Indonesia: rich vs powerful.

Fraction

  • Fraction is usually expressed with percent.

  • We use fraction to express how important an individual is to the group/population.

  • India imports 3.05 billion USD of CPO from Indonesia doesn’t say a lot.

    • India imports 61% of its CPO from Indonesia says how important supplier Indonesia is.
    • From Malaysia ~32%, FYI.

Percent change / growth

  • Growth is important to reflect how fast something is changing.

  • Percent change is nice cuz it’s unit-free.

  • It linearizes non-linear thing, which’s good and bad.

  • If your income drop by 50% today, will 50% increase tomorrow get you back to your old income?

Index

  • Index is prolly the most confusing thing.

  • Index can be in many forms with many different weight.

    • sometimes you might need to check its formula.
    • the purpose of index is still to give us one representative number to reflect the big picture.
  • For example, consumer price index (CPI) calculates a change in price level of many consumer goods.

  • Indeks Kedalaman Kemiskinan shows how deep the poverty of some area is.

A note on CPI and similar indices

  • CPI and many other indicies are shown in number near 100.

  • That’s because CPI is calculated as compared to 100.

    • 100 is the ‘base’, the year where CPI=100 is called reference year.
  • For example, if CPI in 2010=100 while 2020=154, that means prices in 2020 is 54% higher than 2010.

Q & A

via GIPHY