3  Costa et al. (2026): Replication of “Does Money Buy Happiness”

In this chapter, we assume you have covered data wrangling tasks such as creating and modifying variables, and calculating summary statistics. There are quite a few visualisation opportunities here, so we assume you can create things like scatterplots, bar plots, and boxplots.

3.1 Task preparation

3.1.1 Introduction to the data set

For this task, we are using open data from Costa et al. (2026) who collated several replication attempts of Diener et al. (2010). The abstract of Costa et al. is:

This dataset results from a multi-lab replication of Diener et al. (2010) organized by the Collaborative Replication and Education Project (CREP). This project aimed to understand the relationships between different types of needs (basic needs, psychological needs, standard of living, etc.), people’s positive and negative feelings, and life evaluations. This dataset includes data collected across each site contributing to the study. It includes predictor variables associated with participants’ mood and current financial comfort. It also includes demographic data like education, gender, marital status, residential area and income. The data are stored in a. csv file to be used for additional analysis, as an educational tool, and further interdisciplinary collaboration.

You can also read the abstract of the original study by Diener et al:

The Gallup World Poll, the first representative sample of planet Earth, was used to explore the reasons why happiness is associated with higher income, including the meeting of basic needs, fulfillment of psychological needs, increasing satisfaction with one’s standard of living, and public goods. Across the globe, the association of log income with subjective well-being was linear but convex with raw income, indicating the declining marginal effects of income on subjective well-being. Income was a moderately strong predictor of life evaluation but a much weaker predictor of positive and negative feelings. Possessing luxury conveniences and satisfaction with standard of living were also strong predictors of life evaluation. Although the meeting of basic and psychological needs mediated the effects of income on life evaluation to some degree, the strongest mediation was provided by standard of living and ownership of conveniences. In contrast, feelings were most associated with the fulfillment of psychological needs: learning, autonomy, using one’s skills, respect, and the ability to count on others in an emergency. Thus, two separate types of prosperity—economic and social psychological—best predict different types of well-being.

To summarise, they were interested in the age old question of: Does money buy you happiness? The original study by Diener et al. collected different variables including income (in US Dollars), perception of well-being, household income/status, financial confidence, and daily experiences. Costa et al. represent the CREP (Collaborative Replication and Education Project) network who organise replication studies for student dissertation projects. Across several data collection sites, they replicated the methods of Diener et al., and this data set combines them all into one file.

For this activity, it was a little too awkward to work with the raw data, so we are focusing on your summarising and visualisation skills to work with their clean data.

3.1.2 Organising your files and project for the task

Before we can get started, you need to organise your files and project for the task, so your working directory is in order.

1. You can create a folder for the data analysis journeys book and a new sub-folder for this chapter, something like: analysis_journeys, create a new folder for this chapter called something like Costa_wrangle. Within Costa_wrangle, create two new folders called data and figures.

2. Create an R Project for Costa_wrangle as an existing directory for your chapter folder. This should now be your working directory.

3. Create a new R Markdown or Quarto document and give it a sensible title describing the chapter, such as Costa et al. (2026): Replication of does money buy happiness. Delete everything below line 10 so you have a blank file to work with and save the file in your Costa_wrangle folder.

4. We are working with a new data set, so please save the following data file: Diener_replications.csv. Right click the link and select “save link as”, or clicking the link will save the files to your Downloads. Make sure that you save the file as “.csv”. Save or copy the file to your data/ folder within Costa_wrangle.

You are now ready to start working on the task!

3.2 Overview

3.2.1 Load tidyverse and read the data files

Before we explore what wrangling we need to do, load tidyverse and read the data file. As a prompt, save the data file to this object name to be consistent with the activities below, but you can check the solution if you are stuck.

# Load the tidyverse package below
library(tidyverse)

# Load the data file
# This should be the Diener_replications.csv file 
costa_data <- ?

TipShow me the solution

You should have the following in a code chunk:

# Load the tidyverse package below
library(tidyverse)

# Load the data file
# This should be the Diener_replications.csv file 
costa_data  <- read_csv("data/Diener_replications.csv")

3.2.2 Explore costa_data

In costa_data, we have 36 variables that Costa et al. (2026) have already prepared for analysis. We are not going to work with all of the variables, so the key columns (variables) we want you to understand are:

Variable	Type	Description
dataSet	character	Data collection location/source.
ladder	double	The Cantril scale for people’s perception of well-being. Where they place themselves on a ladder from 1 worse life quality to 10 best life quality.
income	double	Annual income in US Dollars ($) or converted to US Dollars.
gender	character	gender category as man, woman, nonbinary, or NA.
age	double	Age in years.
residentialArea	character	Participant residential area of rural, village, urban/city, or suburban.
PositiveFeelings	double	Mean value for average daily smile/laugh and enjoyment.
NegativeFeelings	double	Mean value for average daily worry, sadness, depression, and anger.
BasicNeeds	double	Mean value for food and shelter needs met.
PsychNeeds	double	Mean value for psychological needs met.
study_site	character	Relabelling dataSet as study data collection site.

TipTry this

Now we have introduced the data set, explore them using different methods we introduced. For example, opening the data object as a tab to scroll around, explore with glimpse(), or even try plotting some of the variables to see what they look like..

3.3 Summarising/visualising your data

Compared to other chapters, we are using the clean version of their data set (you can see their data cleaning script here if you are interested). So, the majority of this chapter we are going to spend on summarising and visualising the data to practice answering questions you might have of data.

3.3.1 Demographics

1. How many data collection sites are included in the data? There were data collection sites.

CautionShow me the solution

costa_data %>% 
  count(study_site) %>% 
  nrow()

[1] 9

2. Looking at the sample size per data collection site, what was the largest and smallest number of participants? The largest site had and the smallest site had .

CautionShow me the solution

costa_data %>% 
  count(study_site) %>% 
  arrange(desc(n))

study_site	n
Study_Site_3	213
Study_Site_6	200
Study_Site_4	183
Study_Site_2	163
Study_Site_5	127
Study_Site_7	100
Study_Site_9	99
Study_Site_1	57
Study_Site_8	29

3. To 2 decimal places, the mean age of all the participants was (SD = ).

CautionShow me the solution

costa_data %>% 
  summarise(mean_age = round(mean(age, na.rm = TRUE), 2),
            sd_age = round(sd(age, na.rm = TRUE), 2))

mean_age	sd_age
22.58	6.52

4. Is there much variation in the participant ages across data collection sites? Identify an appropriate visualisation type to visualise age across study_site. There is creative freedom here, but think about what information you have available in these two variables.

   • Did any study sites opt out of collecting age data? No Yes

   • Which study site had the largest variation in participant ages? Study site 3 Study site 8 Study site 9 Study site 4

CautionShow me the solution

The easiest option is a boxplot (or equivalent) with study site along the x-axis.

costa_data %>% 
  ggplot(aes(x = study_site, y = age)) + 
  geom_boxplot() + 
  labs(x = "Data Collection Site", 
       y = "Age") + 
  theme_classic()

5. In psychology research, one potential issue along the WEIRD continuum is whether participants are concentrated in urban or city environments. How many participants reported living in rurally or in a village?

   • Rural:

   • Village:

CautionShow me the solution

costa_data %>% 
  count(residentialArea)

residentialArea	n
rural	34
suburban	172
urban_city	485
village	20
NA	460

3.3.2 Income and well-being

6. The Cantril scale of well-being lets people select which rung of a ladder they would place themselves for life quality, where responses range from 1 to 10. Try and reproduce the following histogram to visualise the responses to ladder. Look closely at each feature to reproduce, like the bins and axis breaks.

CautionShow me the solution

costa_data %>% 
  ggplot(aes(x = ladder)) + 
  geom_histogram(binwidth = 1) + 
  theme_classic() + 
  scale_x_continuous(breaks = 1:10, 
                     limits = c(1,10)) + 
  scale_y_continuous(breaks = seq(0, 300, 50), 
                     limits = c(0, 300)) + 
  labs(x = "Cantril Well-being Scale", 
       y = "Frequency")

7. Now, try and reproduce the following histogram to visualise the responses to income. Look closely at each feature to reproduce, like the bins and axis breaks. Hint 1: There is a huge - almost certainly an error - value to exclude. Hint 2: scientific notation can be annoying here, so add the code options(scipen = 999) before your plot code.

CautionShow me the solution

# increase scientific notation threshold 
options(scipen = 999)

costa_data %>% 
  filter(income <= 1000000) %>% 
  ggplot(aes(x = income)) + 
  geom_histogram(binwidth = 50000) + 
  theme_classic() + 
  scale_x_continuous(breaks = seq(0, 1000000, 100000),
                     limits = c(0, 1000000)) +
  scale_y_continuous(breaks = seq(0, 300, 50),
                     limits = c(0, 300)) +
  labs(x = "Annual income ($)", 
       y = "Frequency")

8. After visualising the Cantril scale (ladder) and annual income (income) in isolation, one potential question is whether there is a relationship between the two. This task has a few steps for you to consider:

   • Create a scatterplot between ladder and income. Add a regression line for a linear relationship. Remember to exclude the huge unrealistic extreme value. If you did not adjust scientific notation before, you will need to add it here.

   • Does it look like there is a positive or negative relationship? It looks like there is a positive negative relationship, meaning as annual income increases, values of well-being tend to decrease increase.

   • Working with data like annual income can be tricky as it tends to be quite skewed, with lots of people earlier lower values and a few people earning higher values. There are different ways of handling it, but if you replace income with logIncome (applying a common logarithm to income), is there relationship similar to the raw income values? Yes, it is the same. No, it is different.

CautionShow me the solution

costa_data %>% 
  filter(income <= 1000000) %>% # ignore the huge outlier
  ggplot(aes(x = ladder, y = income)) +
  geom_point() + 
  geom_smooth(method = "lm") + 
  scale_x_continuous(breaks = 1:10, 
                     limits = c(1, 10)) + 
  theme_classic()

If you replace the income variable with LogIncome, there is a still a slight positive relationship.

costa_data %>% 
  filter(income < 1000000) %>% # ignore the huge outlier
  ggplot(aes(x = ladder, y = logIncome)) +
  geom_point() + 
  geom_smooth(method = "lm") +   
  scale_x_continuous(breaks = 1:10, 
                     limits = c(1, 10)) + 
  theme_classic()

To make more specific conclusions, you would need to apply your statistical models we do not assume you have covered in this chapter.

9. For a final task, we would like you to recreate this scatterplot to visualise the relationship between psychological needs (PsychNeeds) and positive feelings (PositiveFeelings). However, one potentially new feature is splitting a plot into different panes or facets. Use your search skills to see how you can spread the plot into each study_site, where we can explore whether the relationship is roughly consistent across the study sites.

   • How would you describe the relationship across study sites? Across the study sites, there is a positive negative relationship, meaning as meeting psychological needs increases, values of positive feelings tend to increase decrease.

   • If you use geom_point() to visualise the relationship, what is a potential problem with the data to fit a linear relationship to?

     There are no obvious problems.The data look skewed, where there are few high responses and lots of low responses.The data are closer to ordinal, with few response options per combination of variables.

CautionShow me the solution

To recreate the plot, we have a scatterplot between psychological needs and positive feelings. We add jitter to the points to avoid some overlap and we use facet wrap to spread study site across panes.

costa_data %>% 
  ggplot(aes(x = PsychNeeds, y = PositiveFeelings)) + 
  geom_jitter() + 
  geom_smooth(method = "lm") + 
  facet_wrap(~study_site) + 
  labs(x = "Psychological Needs", 
       y = "Positive Feelings") + 
  theme_classic()

If we replace geom jitter with geom point, we can see the data are closer to ordinal. Both variables are the mean of few items, so there are only certain response options available, meaning a standard parametric analysis technique would be potentially inappropriate.

costa_data %>% 
  ggplot(aes(x = PsychNeeds, y = PositiveFeelings)) + 
  geom_point() + 
  geom_smooth(method = "lm") + 
  facet_wrap(~study_site) + 
  labs(x = "Psychological Needs", 
       y = "Positive Feelings") + 
  theme_classic()

3.4 Conclusion

Well done! Hopefully you recognised how far your skills have come to be able to do this independently, regardless of how many hints you needed.

For this chapter, we focused on summarising and visualising the data to be accessible early in your learning journey. Once you are further along in your studies, why not try and apply some inferential statistics to address their research questions? For example, does meeting basic and psychological needs predict perceived well-being on the Cantril scale? Do basic and psychological needs predict positive (or negative) feelings?

Costa, O., Feragen, Q., Wagge, J., & Woodward, A. M. (2026). Data from CREP Replication of Does Money Buy Happiness, Diener et al. (2010). Journal of Open Psychology Data, 14(1). https://doi.org/10.5334/jopd.156

Diener, E., Ng, W., Harter, J., & Arora, R. (2010). Wealth and happiness across the world: Material prosperity predicts life evaluation, whereas psychosocial prosperity predicts positive feeling. Journal of Personality and Social Psychology, 99(1), 52–61. https://doi.org/10.1037/a0018066