Slides.knit

class: center, inverse, middle

.pull-left-wide {
  float: left;
  width: 66%;
}
.pull-right-wide {
  float: right;
  width: 66%;
}
.pull-right-wide ~ p {
  clear: both;
}

.pull-left-narrow {
  float: left;
  width: 30%;
}
.pull-right-narrow {
  float: right;
  width: 30%;
}

.pull-right-extra-narrow {
  float: right;
  width: 20%;
}

.pull-center {
  margin-left: 28%;
  width: 44%;
}

.pull-center-wide {
  margin-left: 17%;
  width: 66%;
}

.tiny123 {
  font-size: 0.40em;
}

.small123 {
  font-size: 0.80em;
}

.large123 {
  font-size: 2em;
}

.red {
  color: red
}

.orange {
  color: orange
}

.green {
  color: green
}

.blue {
  color: blue
}
</style>

# From Geography to Prosperity via new methodological routes

### *Or*

## *What I am trying to achieve in my research*

### Christian Vedel,

### Email: [christian-vs@sam.sdu.dk](christian-vs@sam.sdu.dk)

### Updated 2025-07-03

---
class: middle

.pull-left-wide[
# What this is
> A personal of what I have *(hopefully not)* wasted my time with in the past ~6 years + Some thoughts ahead `$^1$`

.footnote[
`$^1$`*Hopefully you find this useful somehow - it's useful for me*
]

]

---
class: middle

## Why This Journey & Where It's heading

1. **Motivation**  
   - Can we improve life quality substantially? How?  
   – What Denmark’s micro-experiments teach about economic development

2. **Puzzles**  
   *Geography • Institutions • Culture • Persistence*

3. **Evidence toolbox**  
   Causal inference | GIS & archival data | Machine-learning tools

4. **Case studies**  
   - Coal & Peat
   - A Perfect Storm
   - OccCANINE
   - CHAOS

5. **What I'm spending my time on**  
   
---
class: middle

.pull-left-wide[
# From Rome to Economic History  
]

.pull-right-narrow[
![Rome inspiration](Figures/ChatGPT.png)
.small123[
*A much too flattering version of how I learned to love econ*  
*Generated by ChatGPT*
]
]

???
- **Spark moment – Rome, study trip, 2012**  
  *Realised a city can outlive generations; wanted to know the forces behind such endurance.*

- **Dual fascination**  
  *Blend of "hard" (quantitative) and "soft" (historical-institutional) sciences.*

- **Research lens**  
  *Causal inference × Geography × Institutions & Culture*  
  - How shocks (natural, social, technological) redirect long-run development.

---
class: middle

.pull-left-wide[
.pull-right-wide[
## How to improve the world?

### 'Easy' answer
- Go into politics

### 'Hard' answer
- Do research and learn something about it
- (And learn that you cannot generally improve things) 
]
]

.pull-right-narrow[
.pull-left[
  ![PhD defence](Figures/Defence.jpeg)
  .small123[
  **Current identity**
  ]
]

.pull-right[
  ![bdd](Figures/Naive_Christian.jpg)

.small123[
  **BTDT**
  ]

]

---

.pull-left[
# Projects overview
.small123[
#### Published
- .orange[Sharp et al (2023): A database of 1419 Danish cooperative creameries 1898-1945]
- .orange[McLaughlin et al (2023): A database of 531 Irish cooperative creameries 1897-1921]
- .green[Henriques et el (2024): Peat availability and energy shocks (in creameries)]
- .blue[McLaughlin et al (2025): Denmark vs Ireland: Producitvity (of creameries)]

#### Working papers
- .red[Religious identity in the US (Boberg et al, 2024), Denmark (Bentzen et al, 2023) and Ireland (McLaughlin et al, 2024)]
- .orange[Automatic Occupational Standardization with OccCANINE (Dahl et al, 2024)]
- .green[A perfect storm (Vedel, 2024)]
- .green[Track to Modernity (Görges et al, 2025)]
- .red[Milk and conflict in Ireland (McLaughlin et al, 2024)]
]

]

.pull-right[
.pull-righ-wide[
**.red[Culture];  
.green[Geography];  
.blue[Institutions];  
.orange[Data / methods]**
![Line](Figures/Line.png)
]

.small123[
#### Works in progress (too much)
- .green[Railways and the Happy Danes]
  + .blue[Tracks to Modernity (submitted)]
  + .orange[Railways and things to find in many historical pictures]
- .red[Facing success: Pictures and student biographies]
- .orange[CHAOS: Converting Historical Accounts Into Occupational Scores]
- .blue[Vikings and Nazis]
- .orange[Occunyms: Is Mr. Smith's son more likely to have the social status of a Smith?]
- .orange[PST version of OccCANINE ++]
- *Probbably a few things I forgot*
]

]

.footnote[
.tiny123[
*If anyone wonders where my name is in many of these - then don't get a last name starting with 'V' yourself.*
]

]

---
class: inverse, middle, center

## Some projects, published, working paper, early stage

---
class: middle
# Peat and coal
.small123[
*Published in Energy Economics in 2024*
]

.pull-left[
.small123[
"Adaptability, diversification, and energy shocks: A firm level productivity analysis"  
w. Sofie Henriques, Paul Sharp and Xanthi Tsoukli  
[doi.org/10.1016/j.eneco.2024.107887](https://doi.org/10.1016/j.eneco.2024.107887)
]

- Because of WWI and the Spanish Flu, Coal imports from England were strained
- Some creameries were located closer to an (inferior but available) alternative: Peat
- Short run and long run benefits

### Process reflection
- Started off in another PhD students project - needed help with the econometrics
- Working paper out in 2021
- Was rejected approx 4-5 different journals
- Finally decided to fix up weird econometrics

]

.pull-right[
![Plot](Figures/energy_economics.jpg)
.small123[*Figure 6:* Descriptive plot of efficiency and peat avilability]

]

---
class: middle
## A perfect storm: First-Nature Geography and Economic Development

.pull-left[
.small123[
Submitted to the AER 44 days ago (fingers cross for reports with the rejection)  
https://arxiv.org/abs/2408.00885
]

- How much of the location of prosperity is determined by first-nature geography?
- In 1825 an unexpected storm completely altered the first-nature characteristics of North Western Denmark
- 25% population growth within a generation + fertility + manufacturing

### Process reflection
- Stumbled on the idea on vacation
- The struggle was with framing it in the literature
- When should you prioritize your own stuff?

]

.pull-right[
![Figure1](Figures/Map_nobg.png)
*Figure 1: A map of how the West Limfjord region changed in 1825*
]

---
class: middle
# OccCANINE

.pull-left[
.small123[
w. Christian Møller Dahl & Torben Johansen   
Final review of new version TBD before submitting to JEH
https://arxiv.org/abs/2408.00885
]

- An important economic activity is work
- Historical sources are ripe with information of peoples' occupation but it needs to be standardized
- We train a language model on ~20 million observations spanning 13 different language from 29 different sources
- Open source, user friendly, fast and highly accurate

### Process reflection
- Was an appendix
- Networking effort to get data
- Last mile problem

]

.pull-right[
![Archirecture](Figures/Architecture_small.png) 
*Figure 1: Conceptual architecture*

.small123[
#### Some numbers:
- 10k occupational descriptions: 27 seconds
- 100k occupational descriptions: 5 min.
- 1 million occupational descriptions: 45 min.
`$\rightarrow$`  .red[*Compared to 3.5, 35 and 350 RA working days*]
]

]

---
## CHAOS: Converting Historical Accounts into Occupational Scores
.small123[
w. Matt Curtis, Torben Johansen, Julius Koschnick
]

.pull-left[
.small123[
- Your job is a useful proxy of your (expected) income
- Common setup in history: 
  + Source: Some agency noted average wages for certain occupations
  + In the censuses we now what job people had
- Can we infer their level of income (already done) flexibly and automatically using the most relevant *local* sources + skill, education, etc. (contribution)
- Trick: It's all based on probabilities of *OccCANINE*

### Process reflection
- Presented this summer 
- Let's see how it goes
- 'Residual' time project

]
]

.pull-right[
![Archirecture](Figures/CHAOS.png) 
]

---
# CHAOS (if time)

.pull-left-wide[
### Intro 1
- Given:
  1. classifier `$d \mapsto Pr(h|d)$`, and
  2. source `$\{d_i, \mathbb{E}(y_i|d_i)\} \forall i \in[0,n]$`,
  
where `$d$` is a text string, `$h$` is some classiciation of that string and `$y$` is some outcome of interest, can we:

> Estimate `$\mathbb{E}(y_i|d_i)$` for new observations of text strings, `$d$`

### Intro 2
- Can we combine relevant sources to get local estimate of expected occupational wages?

]

---
# Background

.pull-left[
- A lot of papers out there are based on 'occscore' from IPUMS
  + 1950s US occupational wages
  + Less ideal the further you get away from 1950s US
- What if you could construct something similar, automatically using the most relevant local sources?

#### Uses
- .red[*Women's return to education in the US*]
- What happens to income / occupational structure when new technologies emerge?
- Can economies adapt to climate change?
- What is the social mobility across different types of societies?
- What was the effect of abolishing slavery?
- Do certain institutions promote economic development?

]

---
## How to build an occscore index
*How e.g. IPUMS occscore is constructed*
.pull-left[
1. Find *Source table*:

a. Occupational descriptions: `$d_i$` ("carpenter", "best carpenter", "fishmonger", "butcher", "farmer", "has some land")  
    b. Average incomes for each of these: `$y_i$` (125, 130, 198, 201, 107)

2. Turn descriptions into some system (OCC1950, HISCO, PST, etc.)

3. Make judgement calls to collapse and distribute information into an index with *one* average income for each code

4. (Hidden): Application: We have some new data where we only have descriptions. We want to know what they *earned* on average.

]

.pull-right[
.red[
- *All of this consists of various difficult, important and non-replicable judgement calls.* 
- *We want to automate this.*
]
]

---
# What we do (1/2)

### *"It could easily accomplished with a computer"* - Dr. Strangelove
.pull-left-wide[

Imagine having some data:

| Occupation  | Income |
|-------------|--------|
| Carpenter   | $2.00  |
| Butcher     | $2.00  |
| Farmer      | $2.30  |

Form this imagine constructing a function which takes a text string and outputs an estimated income, years of education, social status, etc. for a given place `$i$` and time `$t$`:

$$
f_{it}(\textit{"He has some land"}) = 1.71844
$$

]

---
# What we do (2/2)

An occupational index is a weighted mean:
`$$\mathbb{E}(y_i|h_j)=\sum_{i \in \mathbf{S}} y_i w_i.$$`

E.g. the income of "61110 Farmer" could be 
`$$0.47\times y_\text{"farmer"} + 0.45 \times y_\text{"has some land"} + 0.08\times y_\text{"butcher"}$$`
`$$0.47\times 123 + 0.45 \times 107 + 0.08 \times 198 = 121.8$$`

#### OccCANINE can help with deriving these weights:
`$$w_i= \overset{OccCANINE}{\overbrace{Pr(h_j|d_i)}} \frac{\overset{Prior}{\overbrace{Pr(d_i)}}}{\underset{Norm.}{\underbrace{Pr(h_j)}}}$$`

...And we can also use OccCANINE to apply HISCO wage index to descriptions. E.g. if a census entry is 75 pct "61110", then it should be given this 75 pct of this income

---
# Application

.pull-left-wide[
- Commission of labor reports spanning 1750-1900 for various occupations and countries/states
- London Trademan - an 18th description of what historical occupations required (which we turn into skill measures using NLP)

#### Product:
A measure of long run changes in skill premium

]

---
# Examples

.pull-left-narrow[
- What we do with it
]

.pull-right-wide[

.panelset[
.panel[.panel-name[A]
.pull-left-wide[
.pull-right-wide[
![Wright](Figures/Carroll_D._Wright2.jpg)
*Caroll D. Wright*
]
]

]
.panel[.panel-name[B]
.pull-left-wide[
.pull-right-wide[

![Worktime2](Figures/col-cover.png)
]
]

]
.panel[.panel-name[C]
![Worktime2](Figures/col-first-tab.png)
]
.panel[.panel-name[D]

.pull-left-wide[
.pull-right-wide[
![Worktime2](Figures/ltm.png)
]
]

]
.panel[.panel-name[E]
![Worktime2](Figures/avg_wage_by_place.png)
]

.panel[.panel-name[F]
![Worktime2](Figures/Blacksmith.png)
]

]

---
# How I spend my time

.pull-left-wide[
.panelset[
.panel[.panel-name[Weekly work time]
![Worktime1](Figures/Time_worked.png)
]
.panel[.panel-name[Share spend]
![Worktime2](Figures/Timeshare_spend.png)
]
.panel[
.panel-name[Table]

|Metacategory         |Before PhD |After PhD |Pct change |
|:--------------------|:----------|:---------|:----------|
|Admin                |21.4%      |18.5%     |-13.8%     |
|Dissemination        |5.92%      |5.67%     |-4.26%     |
|Learning             |22.7%      |5.76%     |-74.7%     |
|Public dissemination |0.491%     |0.698%    |42.2%      |
|Research             |42%        |42.3%     |0.741%     |
|Teaching             |7.39%      |27.1%     |266%       |

]

---
class: middle
# An agenda
.pull-left-wide[
#### Two legs: 
Doing applied work in EH and developing the methods for it
  1. The mystery of prosperity and how to create it
  2. How to get reliable data and inference
]