Net Migration and Economic Growth Around the World
This post uses data from the World Bank to look at the effect of migration on countries around the world. I will begin by looking at all countries for which the World Bank has data, then drill down.
So to begin, the data used in this post:
1. Net migration, by country available here. The most recent data is from 2012. Net migration is defined as the “total number of immigrants less the annual number of emigrants, including both citizens and noncitizens. Data are five-year estimates.” As an example, the US reportedly had net migration of 5,007,887 (i.e., positive) in 2007 through 2012, while Bangladesh had a figure of -2,226,481 (i.e., negative) in the same years. That should fit with your intuition.
2. PPP GDP per capita. Data available here. The last year for which data is available is 2015.
3. Data on population through 2015.
I started by looking at immigration relative to the size of the population. I assumed that the net migration figure was the same in each of the five years. (I know – not correct, but reasonable.) I then divided the Net Migration from 2012 by Population from 2012. I then compared that to the annualized growth in PPP GDP per capita from 2012 to 2015. In other words, I looked at the Net Migration as a share of the Population in 2012 and the growth rate in the subsequent three years. I put both series up on a scatter plot.
Before I put up the graph, I would also note that I did leave some data out. It goes without saying that if a country did not report information, I did not include it. Additionally, countries reporting zero net migration were left out. After all, even North Korea has escapees, er, migrants, even if they won’t admit to it. Otherwise, everything went into the pot leaving a sample of 176 countries. Here’s what the relationship between Net Migration (from as a share of the Population in 2012 and the growth in PPP GDP per capita from 2012 to 2015 looks like for them:
The correlation is -0.32. That is, countries with higher Net Migration as a share of their Population tended to perform less well over the subsequent three years. In other words, it is better to give than to receive, at least when it comes to migrants.
Of course, if we want to understand the effect of Net Migration in the US and other Western Countries, perhaps it makes sense to narrow things down. The next graph uses only countries deemed to be “High Income” by the World Bank. I also restricted the sample to countries with populations exceeding 1 million people to avoid trying to learn life lessons based on recent happenings in Monaco or Andorra. Here’s what that looks like:
The population sample dropped from 176 countries to 44, and the correlation tightened up a bit to -0.48.
Frankly, I think the sample still needs cleaning up. Most of the points on the graph look bunched up because there are a few countries with very, very high Net Migration. For example, Oman is at 6.8%(!!!!), Qatar 3.6%, Kuwait 3.0% and Singapore 1.5%. These are mostly special cases, even for high income countries, and I would venture to say, provide very few lessons on immigration that are applicable to the US or most of the West. Limiting the sample to countries with Net Migrants to Population under 1.4%, the graph now looks like this:
This doesn’t change the outcome much, but it makes things easier to see. If desired, we can cut out one more outlier – this one on account of excessive economic growth. The point on the far right side of the graph is Ireland, bouncing back (in PPP GDP) from the monster collapse in 2007-8. Removing Ireland as well gives us this:
The absolute value of the correlation drops, but the fact remains: we are still left with a negative correlation between Net Migration as a percentage of the Population in 2012 and the growth in PPP GDP per capita between 2012 and 2015. We can do a bit more pruning, but frankly, the data simply refuses to support Holy Writ. Sure, these graphs don’t prove that immigration is bad for growth. However, they make it very, very hard to argue that immigration had a positive effect on growth during the past few years. Of course, that isn’t what we hear from our betters.
I will follow up this post with looks at other periods for which data is available from the World Bank. Meanwhile, I put together a spreadsheet that allows the user to make changes to the dates or downselect the data through income level, population, etc. It’s a bit large, but I will send it to anyone who contacts me for it within a month of the publication of this post. I can be reached at mike and a dot and my last name (note – just one “m” in my last name) and the whole thing is at gmail.com.
Updated about fifteen minutes after original posting. Figures 3 and 4 needed an additional significant digit on the Y-axis.
When you use / show the term “correlation” it’s not clear what correlation method / calculation is being used. I ask because you use a signed value and Rsquared, which is the most widely used correlation metric in regression analysis is necessarily and always a positive term. (the square of negative value is a positive value).
This makes me suspect you’re using correlation as R rather than Rsquared. (R^2). Therefore can you let me know what “correlation” on your chart is referring to in statistical terms?
if for example your 0.36 “correlation” is R rather than Rsquared, then it’s a normally referred to correlation of 0.13….. which is to say that only 13% of the variation of the data with respect to the regression fit is due to the change in the x-axis variable. This not considered of a statistically significant correlation by any statistician in regression analysis. — in other words it’s no correlation at all.
Which is why I ask the question.
Sorry to bother you again, but I forgot to ask about figure 4 chart in which you show “high income countries with net migration populatin < 1.4% of total population, but with Ireland removed".
In that chart I note that only two points make the correlation even as high as "-0.36" — … and both are the only points < 0% in "Annualized PPP GDP per Capita" for the 3 years in your data.
It is thus pretty important to know which countries those two points refer to since there may be clear and fundamental (and direct )and well documented reasons for their low "annualized PPP GDP per capita" than having any effect at all due to immigration population proportions.
Also I might point out that according to your chart's data (Figure 4) there are ~ 10 nations with immigrant populations between 0.5% and 1% with spanning the PPP per Capita range from -2.5% to +4.5% where the mean PPP per capita of the total data set appears to be approximately +2% – +2.5% (by eyeballing)… so the range for the 10 nations from ~ 0.5% to 1% in migration populatons covers the range of PPP per Capita from -2.5% to + 4.5% … which means the 2 neations with negative PPP per Capita are extremes with respect to the total data.
Also there can be no real correlation — no causation (which is what you're implying) — between a +0.5% to 1% migration population and the +0.5% to 0% or -0.5% migration population with respect to PPP GDP per Capita when both span the same range (excepting the 2 nations I've asked you to identify).
There are ~ 26 nations between ~ +0.5% and ~ -0.25% in migration population spanning the range of PPP GPD per Capita from 0% to +4.5% PPP GDP per Capita with no correlation at all… i.e. a pure scattergram in statistical parlance.
Thus your entire chart (Figure 4) and inferences you're making from it rests on just two nations of ~36 – 38 nations, and which nations are the only nations in that chart with negative annualized PPP GDP per Capita — which is why I ask if you will identify those nations and provide the data which you used to "annualize" their PPP GDP per Capita and the source you used for that data.
with ins e Dspnder imntions 9 data points betw
I assume by two points you mean the ones at the far left with negative growth rates. They are Norway and Cypress. Limiting the sample to countries with a positive growth rate – i.e., eliminating those two – gives a correlation of -0.21.
As I note in the post, I can keep pruning. And pruning. If the goal is to select a sample specifically designed to generate a positive correlation, I can provide you with that data set. But it cannot be arrived at without a heroic amount of cherry picking.
As to correlations, its just the correlation coefficient returned when you enter “correl” in Excel. More here, including the formula Excel uses: https://support.office.com/en-us/article/CORREL-function-995DCEF7-0C0A-4BED-A3FB-239D7B68CA92
Note that no regression is involved. I have found that both algorithms Excel uses for running regressions are inaccurate and I don’t know why they keep using it.
As I noted in the post, I am happy to provide you with the spreadsheet. Or if you want to wait, I will improve it and add functionality and ease of use before the next post.
You spend a lot of time in the minutia of the discussion without drawing a conclusion. I assume you have read the 2006 Card study, 2012 Hamilton study Rationalizing U.S. Immigration Policy: Reforms for Simplicity, Fairness, and Economic Growth, and also (Ottaviano and Peri 2008; Ottaviano and Peri 2010; Cortes 2008.
What concerns me is the constant look at the micro without looking at a solution. If your concern is to reform immigration policy, then stating it as such would be a good start. Hamilton study touches upon that in the end.
Mostly Card and the others find little impact upon earnings for US citizens. Immigrant children catch up with US citizen children and later assimilate into society. Immigration also lowers the median age of a population. The difference I believe you are pointing to is selectivity of immigration by the US to meet the needs of the market place not over burdening one sector as opposed to another.
What concerns me is the approach gives a resemblance to the hysteria created amongst those believing they are maligned by immigrants beyond the economy and other factors vocalized by Trump in his rise to power and the Alt-Right when immigrants had little to do with their downfall. People do rally around false gods and causes.
As with taxation, my primary wish is that people understand what the data actually shows and that what the data actually shows is very different from what most of us want to believe.
Then it would be nice if we were honest about what purpose we want immigration to serve. If it is to generate economic growth we are doing it wrong. As per the data I showed for the US, if growth is our goal, we have been doing it wrong since the late 1960s. If charity is our goal there are no doubt cheaper and more effective ways to accomplish more desirable charitable outcomes.
As to Card’s study of the impact on wages, it is a very sad thing that the only ones who are willing to point out the joke around which the paper is based are those on the alt-right. The joke, of course, is pretending that the Mariel boat lift didn’t happen at precisely the same time and place, and get utterly swamped by the cocaine import boom.
Seriously. Why are the traditional right, the center, and the left willing to cede discussions of actual facts to people who look up to Richard Spencer? That just means that eventually folks like him will dictate policy. That is the last thing I want to see, but if his side is willing to look at the facts as they are and my side is not, in the long run, he wins. But if we face the way the world actually is, we can shape outcomes that more closely conform with our sensibilities than with Richard Spencer’s dreams.
You have no conclusions to the data and the constant review of the minutia is exactly what Spencer, Pollack, Breitbart, Bannon, and Trump would keep doing as it stirs the masses in their direction. When I see the former doing it; it is certainly not going to be a discussion on Economic impact and leads to other directions not akin to economics. Breitbart, Pollack, Bannon, Spencer, and Trump have other things in mind. With Bannon on staff, it is a serious blow to the integrity of The White House in support of what Breitbart emphasizes.
With every study, there are conclusions drawn. Statistical analysis points in a direction. Hamilton points to a conclusion which supports your suggestion in a change in how people are selected for immigration. Certainly if it is not based on ethnicity, race, or wealth. There are other factors there which you have not been strong in presenting.
I am not here to discuss whether Card (1987) is legitimate or not as it is a deflection on the issue and something which Trump is very good at doing when he does not want to discuss real issues. If you want to fix Labor issues, there are better ways to do so rather than concentrating on immigration. So far the attention has been on them rather than the issue. Other factors have had a far greater impact on Labor other than immigration. I do not believe the “feel good” people here understand your direction.
Perhaps I missed them, but I do not see the R-squared values.
Anyway, part of the problem is that the data sets are not orthogonal. Poor people are more likely to migrate to a place they expect will provide them with better opportunities. Being poor, their migration into a country lowers that country’s per-capita GDP. That does not mean that they lower the per-capita GDP of the citizens.
Note. My previous comment is not intended to imply that the alt-right, however defined, is always willing to look st data dispassionately. But a quick google search turns up a lot of discussion on immigration from the people of that bent that is more data-based than the feel good nonsense I see on ours.
I am no economist but it seems to me that using ‘growth’ as a stand in for economic well being is spurious. In a country with an aging population and a relatively low birth rate, isn’t it important to grow the base of people in the work force if we wish social insurance programs to be solvent?
Already said that, nice article “300 Million and Counting, Joel Garreau (senior fellow at George Mason). There is a better way to do it; however, it will not solve aliens slipping across the border the same as a wall will not solve the issue.
Nothing will solve the problem of illegal immigration. Nothing has solved the problem of burglary, either, but we still lock our doors to make burglary more difficult.
Mike, fyi and others. the “correlation” you show on the charts and which you have now provided the meaning thereof, is not the Rsquare (f^2) but only r. r^2 (Rsqauard( is the standard correlation coefficient used for all statistical analysis… which is why it isn’t a signed value (square of negative value positive result).
The means the correlations for the data in the charts you share are on the order of 0.12 – 0.14 (12% to 14%) which means there’s no correlation in fact.
Also you said “no regression was involved” but in real fact it was entirely involved since the correlation (r which you used) is the result of a regression of y on x— in this case a linear regression or y = mx. The fact is, Mike,when you show a “correlation” value you are in fact showing a metric describing the significance of a regression of x on y. A “correlation” means a regression analysis was done — just because you don’t show it or even see it in fact (in your case) doesn’t mean there was “no regression involved’. Regression was indeed central to the charts you show… you just omitted the regression result… assumedly because you didn’t know you were actually doing a regression by showing data on a chart and even publishing a “correlation” value.
Also if you’re using Excel RSQ() is a built in function as are the coefficients of a linear regression (slope() and intercept() functions in Excel). .
It’s been pretty clear though that you don’t know very about what you’re doing when it comes to trying to draw conclusions from this and past data analysis you’ve shown. You might want to spend more quality time leaning more about drawing inferences and conclusions, and significance or lack thereof from data… especially with sparse data -sets you’re using in this and your other charts. I mean that to be a constructive criticism.
It is customary in charts and data correlations to show the Rsquared correlation coefficient as “Rsquare = xxx” or “R^2 = xxx” so that the audience knows what “correlation” means. If there’s only a correlation value shown (as you did on your charts) then the standard is that it mean the Rsquare. correlation coefficient.
I had asked if you could provide the source data linksfor:
1. The PPP GPD per Capita values you used to calculate the “delta” to the Annualized “delta” of PPP GDP per Capita for the years 2012 – 2015.
2. The data for immigration population values you used … presumably these were also annualized delta’s over the same time period or perhaps they were the average immigration population values for the period, or for some other time period.
I asked because quite frankly I want to know who’s source data you used (since there are several possible sources and they don’t always agree with one another), and because I don’t actually know what you mean by annualized “delta” , year 1 to year 3 For three years there are at the very least two (2) delta’s [year 2- year 1, and year 3 – year 2] and since you used the term “annualize” it implies that the delta’s were taken over shorter time period during the three years… perhaps quarterly or semi-annually.
I’d appreciate a description of more precisely what “annualized delta” means for a 3 year period.
Thank you for letting me know the two nations that had the only negative “annualized PPP GPD” values though. but you didn’t comment on the point I made that it’s only those two nations that make the chart look like there’s a relationship .. when in fact there isn’t based on the r^2 correlation of the regression of y on x.
The omission of those data points result in an r^2 = 4.4% Mike (= 0.21^2)… there was no correlation before omitting them either but omitting them which reduces the correlation coefficient even more only serves to show that that their inclusion doesn’t change anything… no correlation is no correlation whether it’s 12% or 4%. You’d have to study a little statistics (like a beginning lower division 1 semester class) to know this though.
Assume a population of 10 people with wealth 10 in year 1. They all have 10 percent growth and reach 11 in year 2.
A refugee has wealth 2 in year 1 and doubles to 4 in year 2 after becoming part of the group.
Aggregate wealth per person year 1: 10*10/10 = 10
Aggregate wealth per person year 2: (10*11+4)/11 = 10.36
Growth without immigration: 10 percent
Growth with immigration: 3.6 percent
I don’t think your aggregate measure is telling you anything at all.
Arne, thanks for the example of “the data sets are not orthogonal.” Well done.
I don’t want to alarm you but correlations appear in pretty much every introductory statistics textbook out there and are used by people who don’t know a regression if they saw one. i think you are reacting to what you are imagining what I have written and not what I wrote. I can provide my spreadsheet if it makes it clearer.
That only works if they generate more output than they consume.
Arne and Warren,
Before my time, Will Rogers used to talk about Okies leaving OK and arriving at CA and raising the IQ in both cases. But four options are possible. The one Rogers suggested, the opposite of that, OK benefits and CA is hurt and OK is hurt but CA benefits. I am of course leaving out neutral options. Even if it is the same group moving out of OK and into CA we would reach very different conclusions and policy choices under each scenario.
For the graphs I provided you simply cannot argue the Will Rogers “everyone wins” scenario.
Your analogy sucks. Moving a person from OK to CA will not change their IQ, but it will change their productivity.
Even if your thesis is correct, these aggregate data neither support nor contradict it. Using data to inform policy choices when the data do not bear on the policy is “not well done”.
Actually with respect to Mexican immigration the vast decline in the total fertility rate from 6 in the 1960s to 2.6 today is fixing the problem. as there will be less surplus population from which illegal immigrants are drawn. (Not so much in Central America however). The trends seem to indicate that soon Mexico will be at a total fertility rate that results in very small population growth.
It was the rapid increase in population of the 1960s and 1970s that lead to a large group of men who could go north.
Interestingy in large cities in east asia you see total fertility rates of less than 1 child per woman. (Korea and Shanghai).
Note that for native born americans the total fertility rate in 2014 was 1.8 which is less than replacing population over time.
Working off memory, about 13% or 14% of the population is foreign born. Assume a proportional percentage of the work force is foreign born (OK for ballpark discussion). Nothing that increases the work force by 13% can be dismissed as trivial. And it is very hard to believe that a 13% increase in the size of the workforce doesn’t have an effect on wages unless it increases the demand for labor by precisely the same amount.
As to what I am advocating… right now, nothing other than that we should all pay more attention to data. But if you want a policy prescription from me, well, I want an immigration policy that is at least intended to boost growth, and where conflicts arise, weighted toward benefiting those Americans with the fewest options in society. I believe they are getting hosed.
Let rephrase and repurpose Will Rogers. Assume two states, OK and CA, with people moving from OK to CA. You have four options:
1. Both states see an increase in growth.
2. OK sees an increase in growth, CA sees a decrease in growth
3. OK sees a decrease in growth, CA sees an increase in growth
4. Both states see a decline in growth
The first option is consistent with immigration and emigration both being good for the economy, and possibly with the the idea that second raters in OK can nevertheless do great things in CA due to differences in skill sets or some other factor. The second option implies that emigration is a positive, but not immigration – perhaps because the worst performers are those who move. The third option implies that immigration is a net boon, but not emigration. Perhaps more people are better, or there’s a brain drain from OK to CA. The fourth option implies both emigration and immigration are negatives.
The data set I looked at is a bit different, but it does seem to imply certain things. The fact that even sliced pretty fine, countries with more immigrants do not have faster economic growth seems to say to me that having more immigrants does not lead to more growth. I note that the fashionable belief is that more immigrants do lead to more growth. Thus, the fashionable belief doesn’t mesh with the data.
I would also suggest a thought experiment to go with your example. The country with a total population size most similar to Japan is Mexico. Japan has a lot more infrastructure, which implies that if you move a Mexican to Japan, you will increase that person’s productivity if productivity is due to infrastructure. Conversely, moving a Japanese person to Mexico, on average, would reduce that person’s productivit. But what happens if you move the entire population of Mexico to Japan, and the entire population of Japan to Mexico? Fast forward 20 years. Do you people that the Mexican population of Japan will be more or less productive than the Japanese population of Mexico? I would venture to say that for producing positive outcomes (economic growth, good standard of living, etc.), the culture the Japanese population carries with them is more important than the infrastructure they have built in their islands.
It takes a very long time for falling fertility rates to lead to less emigration.
As I see it Mike is looking at a very short term immigration rate issue. There is no difference in one human’s average intelligence or productivity from any other. The differences are only in luck of the draw (where born, parents economic status, educational opportunities etc).
The only other differences are in cultural affinities, appearance, race, ethic family name, & family ancestry.
Appearances, race, ethnicity, etc. have no relationship to productivity or intelligence, nor does cultural affinity. Productivity is a measure of output per hour not of whether somebody prefers tto not work as much as somebody else. Whether somebody works more or less is a matter of personal interest and personal economic status or need or desire.
An immigrant to a new nation and different culture than their own is still driven by their own personal interests in working or not working, or their desire for more or less wealth and personal family needs. 9 times out of 10 (or far more often imo) a person emigrates from their own own nation to immigrate into a another for either political reasons, family reasons (being with or nearer relatives who have immigrated earlier), or economic reasons… the latter being to find more work or more opportunities and live a better life than by staying where they grew up and where they have friends and family and close ties to their culture and comfort zones.
In the short term immigrants find and do the work they must just to pay the rent and put food on the table… which means for the most part they do the menial and low paying jobs. .. and in most cases they are competing with other immigrants for those jobs — which employers love because it reduces wage and wage growth, so improves or maintains their profits.
In the medium term they get better jobs with higher incomes, and begin to adapt to their new environment — cultural modifications ensue to make the adaptations — either due to laws or just peer pressure and wanting to be accepted.
In the longer term they have kids who go to schools in a culture that is entirely different than their parent’s culture. They learn the language and customs and culture of their home — which is the nation to which their parents immigrated. Of course the kids will remain tied to their parent’s culture — just look at the population of native born Americans in the former confederate states for example.
But within a generation or two the kids are as American as any native born American… though Italians and Irish and Swedes and Norwegians and Japanese and Chinese, and Koreans still to this day after multiple generations in the US have and retain strong cultural ties to their family ancestral heritage. And by the way I and my wife are both just 2nd generation immigrants…. and about 1/3rd of everybody else I know is a 2nd generation immigrant as well…. all of whom are as American as apple pie. Funny though how everybody still celebrates their national origin and ethnic heritages and many of the customs… though the 3rd generation.gives them far less attention. .I know several “illegal immigrants” who’s kids have now graduated from high school and colleges and are working in very well paying and even very high paying jobs, with girlfriends or boyfriends who aren’t of their same heritage or ancestry. Their kids are as “American” as my own.. in everyway but who still speak two languages fluently, and who still celebrate holidays of their parent’s nation.
So there is no medium term or long term issue with immigrants unless one is a racists or white supremacists who think that because the nation was founded by WASP’s the population should be dominated by domineering WASPs who should remain in the vast majority to insure their Waspish ways and traditions aren’t modified or adjusted over time by immigrants.
In other words you can’t make an economic case for selective immigration over the medium and long term and at best only in the very near term can one try to make an economic case based on immigration sources.. though I don’t find that or any data to show any significance to any adverse economic conditions… or restricting economic growth.
My wife’s grandfather and grandmother on her father’s side were immigrants — both labored in the migrant agricultural camps from Colorado to Washington, Oregon, CA, Arizona and back to Colorado annually. My wife’s grandfather and father and uncle lived in places like cleaned out chicken shacks with dirt floors and went hungry a lot.
When the war started, her father got a job in the shipyards and learned carpentry and how to read plans. After the war he saved some moey and bought a small lot in a nearby put low population area and built a house with his wife’s father as his helper … a house for speculation sale. From there it was history as he became a very, very wealthy major contractor in Northern California for major corporation construction projects. Needless to say my wife and her sisters grew up not wanting for anything and expecting to get whatever they wanted — and did. Her father’s brother on the other hand remained an agricultural picker and finally a low end foreman but was always dirt poor, drunk half the time and died by the time he was 55… and he looked like he was 75 or 80 at that time. He was a nice guy, very friendly, who wouldn’t harm a flea, but he had zero interest in money or wealth or living a better life… though he could have since he was smart as a whip and his wealthy brother offered him multiple opportunities with a big stake in money if he wanted to pursue those opportunities.
Oh, in case you’re wondering… my wife isn’t Mexican or South American or even of European heritage but she looks just like every other American WASP.as did her father and uncle. I have to say though her father’s mother might have had some European blood…. maybe.. but even that is mostly speculation.
So I am trying simply point out that immigration to the US of anybody from anywhere and any culture is perhaps and only perhaps a minor and insignificant economic issue in the very short term only and none at all in the medium or long term. The only issue I can see people have with immigration is that they’re fundamentally scared to death of becoming a racial minority or being unable to exploit racial minorities and make sure they remain the “superior” race…. as if their race (my own as well) is any more superior than any other.
This nation has always and continues to thrive and advance on the backs of immigrants… how many of our best and brightest scientists and academics were immigrants or the offspring of immigrants? The only thing I can see holding back immigrants is our nations’ racist attitudes and thus economic and educational effects these have on immigrants who aren’t of the Northern European variety.
What do you think happened with immigrants? Did they suddenly have some major mutation and thus some recent (last 100 years) evolutionary trait that make them less capable or less adaptable than any in the past?
You’re out to lunch on your immigration kick, Mike. . . .
By the way, I just did an analysis of US corporate profit growth per capita since 1950 and a relation to per capita personal income growth after taxes in both cases… for another thing I was interested in… unrelated to your immigration gig.
There’s no way in hell you can show or imply a relationship to corporate profit growth or it’s ups and downs and immigration, nor can you show any relationship what-so-ever to per capita personal income growth over that same time period. And if you can’t show a relationship of immigration to business profits or to personal income after taxes (or before taxes for that matter— I’ve look at both) then you ‘re simply making things up. You’ve yet show a single piece of statistically significant data to support your hypothesis.
What I find more alarming is that you don’t know the difference. If you don’t know the difference then you’re not being rational, and if you’re not rational then you’re irrational, and if you’re irrational then you’re operating on beliefs that have no merit in realities — hence living in a fantasy world of your own imagination;.
That’s pretty much they reason Trump won in the election. People living in a fantasy world.