Part 2— Estimated Living Expenses

Image source: http://horizonlightinginc.com

This is part 2 of a series of articles that explore the prospective salaries and living expenses across several countries and cities. If you want to read part 1:

Living expenses & salaries with data

In the previous part, we talked about salaries in the driven industry for California, UK and Germany. I also touched on the taxes of those locations and came up with a net annual salary estimate. Next I would like to talk about expenses in those places.

Basic Living Expenses

This is where is gets really messy and complicated. I have given up on the idea of achieving high validity, but that doesn’t have to be case with reliability. Put in a simpler terms, I expect the results of my “study” to not be accurate, but at least they will be consistent in their inaccuracy. The idea being that this methodology will help me compare the different countries and possibly cities.

So I propose a very simple model that I will use to compare the different locations.

Basic living expenses = Rent + Food + Utilities + Transportation

Here are some assumptions that I make:

• I will pay monthly rent and utilities
• Utilities will include: water, heat(gas), electricity, internet, mobile phone plan and whatever other mandatory fees there are
• Transportation is everything that deals with commute — price of gas, bus and train tickets, etc. (car payments are not included)
• All prices are monthly average
• I will be spending only for a single person (me)

Numbeo

Now that I know what kind of data I need, I can go and get it, right? Well its not so easy. The first thing I did was to google around (actually I use duckduckgo) for some quality data. The first credible looking resource was Numbeo.com, it has a lot of data on different cities and can also compare two locations. However their statistical model is not a very good fit my needs. For once their data is for an unknown subset of the population, and is probably very biased. But for sake of science, lets see what kind of data I get. Again I make several assumptions:

• 1 member household
• Public transport — monthly
• Rent — 1 bedroom, outside of city center
• All others — 0 or lowest value possible

Then I collected data on 38 different cities across the UK, Germany, California and Japan(my most recent place of residence):

Fig.1

I expected, rent to be the most “volatile” variable. San Francisco didn’t fail to disappoint and came out on top. I was also expecting London to be #2, but I was wrong.

The bar plots above are useful when comparing different cities, but I would like to look at whole regions and individual cities at the same time. There is a really cool way to do that - behold the mighty box plot. Each country and California are represented by a colored box. The line inside the box is the median value, which unlike the average, is much more useful when we have such outliers (for more info on median, click here ). The box plot is useful to point out outliers and to see how skewed the data is. For more information on box plots see wikipedia.

Fig.2 a,b

Fig.3 a,b

Fig.4

I personally think that the results are very interesting, but by no means are they final, nor complete. There is still more to be found, but I will try to make some conclusions for what I have found out so far. Feel free to correct me in the comments bellow or inline.

• California is indeed much more expensive than Germany, UK or Japan. I would need an extra

Again I want to emphasize that the above values are estimated by Numbeo, and are only used to compare different cities/states. That being said, the rent and food values for Japan are very close to what I currently pay and the estimated monthly basic is also pretty close. But the fact is that there are many more expenses that I make:

• Eating out - restaurants & bars
• Clothes
• Sports & fitness
• Domestic goods
• Books and online subscriptions
• Travel

There are some other, but these are the main ones. My believe is that with the exception of eating/drinking out, the rest of the expenses should not vary significantly from one country to another. That is, I expect to spend the same amount on books, clothes, etc., regardless of the country I am residing in. While this might not be case for everyone, I will assume it for the sake of simplicity. All that is left to figuring out, is how much one spends going out in different cities/countries.

Going out

In order to maintain some social life, I have found that you need to spend a proportional amount of money and time. I will only look at the former. Again, it depends very much on the person, and I will have to make some wild assumptions here as well. Here they go:

• Fast food and the likes do not count - judging from my experience those meals usually cost as much as cooking a meal yourself. Quality and other factors aside, I am only looking at the dollar sign per meal here.
• Eating out would be either lunch or dinner at a restaurant. I am interested at the cost for an average meal + the tip (if any).
• Coffee - I need to figure out an average value per beverage and average beverages per month
• Bars - again average beverages and their cost

So I come up the following formula:

Going Out = Lunch Price * Lunch Frequency + Dinner Price * Dinner Frequency + Coffee Price * Coffee Frequency + Alcohol Price * Alcohol Frequency

Lets start by estimating the frequency of each of the above. I would prefer to be a bit generous in my estimate, after all overestimating is much more desirable than the alternative.

I have been keeping a detailed personal accounting for nearly 8 years now. In 2012 I migrated to MoneyWiz (awesome app for OSX and iOS, highly recomended), and I didn’t bother importing my older data. Regardless I think that 4 years of observation should be more than enough. Exporting the data from my app was fairly easy and my amazing log keeping skills allow me to see individual transactions like lunch, dinner, coffee and beer. Take that, everyone that ever called me weird for keeping track of things.

Fig.5 Lunch frequency and expenditures per week

Fig. 6 Dinning out frequency and expenditures per week

Fig. 7 Coffee shop frequency and expenditures per week

FIg. 8 Alcohol consumption frequency and expenditures per week

A little bit of clarification - in the title of each plot “bin” refers to the bin size of the histogram. For example, in the alcohol plot, I combined all the individual expenses between 500 and 1000 yen and counted them together. The reason I choose that number is because a beer in Japan usually costs about 500 yen. This way I can roughly estimate how many beers I usually have when I go out. My friends knows that I more often have single malt (which is a bit more expensive), but let’s keep the units in beer for simplicity sake. And ”frequency” simply refers to amount of weeks I had X times of lunch/dinner, etc. For example, over the course of 4 years (208 weeks), there were 85 weeks where I went out for a drink only 1 time per week.

In addition to the consumption frequencies, I have also added plots for the expenses. I hope you find them educational if you are interested in how much things cost in Japan. Maybe I can dedicate a post on that some time in the future.

Frequencies

As I mentioned above, I will try and stay a bit to the right side of the plot when I choose values for my model. Here is what I chose for the frequencies:

• Lunch - 5/week
• Dinner - 2/week
• Coffee - 2/week
• Beer - 3/week

Consumer prices

Alright, we got the easy part done. Now I have to devise a way to find out how much each one of those categories costs in each of the countries I am interested in. I have come up with 2 different approaches:

1. Use price estimate data from Numbeo
2. Use my price data and multiply by the Consumer Index / Restaurant Price Index (which I also collected from Numbeo)

Numbeo collects all kinds of data points, which you can freely access. I choose only three, to keep it simple. Bellow is the result of the first approach:

Fig.9 a,b,c

Unsurprisingly, beer is the cheapest in Germany (and probably better). And surprisingly meals in the UK are more expensive then those in California. I was expecting the life in sunny state to be consistently expensive, but I was proven wrong. And this is why I made this study - I can’t trust one persons biased ideas of what the world is like. Numbers and data are your friends, you need to learn to hear their voices… like I do… sometimes :P

Bellow you can see the computed values based on what I pay here and the individual Consumer Index of each city. I tried to account for tips, by increasing the lunch and dinner price by 10%.

Fig.10 a,b

Fig.11 a,b

I took the 75% quantile for each country, just to be on the save side, and named that the estimated value of the respective category. I also took the values for London and San Francisco as they were. Next I computed the monthly going out expenses based on the model above. Model 1 uses the Numbeo price estimates (Fig.9) where :

price for Lunch = price for Dinner = Meal Inexpensive

Model 2 uses my historical data for Nagoya and the Consumer Index / Restaurant Price Index.In addition I computed the mean between the two models. The results are shown bellow:

Fig. 12

As I mentioned before, I stayed on the side of caution and allowed for more exaggerated prices and frequencies. I can confirm this by comparing the model data with my actual expense data. My average going out expense per month for the last 4 years was about $250. It falls a bit short even from the lowest estimate, but there is no reason for concern. I would have been more worried if my estimates were lower than the actual numbers.

Looks like London is the city with the highest cost for social life. Even the UK as a whole is more expensive than San Francisco. I must admit I did not expect that. That is most likely due to the higher cost of restaurants.

I believe that by now the readers that have adverse effects to bar plots, have already closed the browser window. So it’s just us nerds and data voyageurs.

Putting it all together

Now that we have some idea what the most common expenses are, lets put this together with the salary data. I have followed the following model:

Disposable income = Net monthly salary — Basic living expenses — Going out(Model Average)

Disposable income is not a very appropriate name, but I am fresh out of ideas at the moment (leave me a comment if you have a better idea, and I will change it).

Fig. 13

Above (Fig.13), I added the total calculated cost, since it might be interesting to some people. Bellow (Fig. 14), I added 2 of the plots from part 1, since its easier to compare.

Fig. 14 Yearly Gross, Monthly Net and Monthly Disposable

Oh my, look at this! There are negative numbers for some data points. Before you start pointing out that this is impossible and trowing wild statements of the sort: “people there obviously live with those salaries”, I need to remind you, that the above is a MODEL. As the famous proverb says - “All models are wrong, but some are more useful than others”, so my clumsy model is without doubt wrong, but I find it very useful when it comes to comparing salaries between different regions. Also don’t forget that my estimation does not include many other categories like clothes, books, sports, etc.
Now! judgement time:

• From what I see, for the same position, living in London is much less profitable than anywhere else.
• Germany and the UK seem to be comparable.
• Silicon valley is beyond doubt the place where you can make it rain!

EDIT: I added an extra plots to show the expected expenses per city:

Fig. 15

Fig. 16 Estimated Combined Expenses (USD)

Join me in Part 3, where I take a different approach to estimate living costs in the UK.
In Part 4 I will try to do the same for Germany.
Or you can go back and check Part 1