Latvia is catching Estonia and falling behind Lithuania, and those are two different stories. Income per head is 70 % of the EU average in Latvia, 80 % in Estonia and 88 % in Lithuania. The lead is not in how many people work. It is in what an hour of work produces: about 20 % more in Estonia and 15 % more in Lithuania.
The model explains almost all of Estonia's lead with foreign firms and innovation. It explains none of Lithuania's. Lithuania's settings in the six areas the model can copy would leave Latvia poorer, not richer. Lithuania's lead sits where the model has no policy: in the productivity of its industry and in people leaving farming for other work.
Why it matters
Three countries started from one place. In 2000 Latvia's GDP per head was 35 % of the European Union average, Lithuania's 38 %, Estonia's 42 %. Up to 2008 all three grew fast, and Latvia faster than Lithuania. The crisis of 2008 hit Latvia hardest: from 57 % in 2008 to 52 % in 2010. Then the paths parted. Lithuania climbed from 60 % in 2010 to 84 % in 2019, Latvia from 52 % to 66 %. Estonia had its own downturn after 2019, which is why its lead over Latvia has shrunk.
The other indicators agree. Estonia's population has fallen by 2 % since 2000, Lithuania's by 18 %, Latvia's by 22 %. Lithuania's net migration, that is, arrivals minus departures, is +5.8 per thousand residents in 2025, Latvia's −0.7: after 2019 Lithuania became a country people move to, and Latvia has not. Output per hour of work in Latvia has grown by 142 % since 2000, faster than in Estonia but from a lower start, and Lithuania has gone further.
Why this matters beyond a contest. Because three countries with a similar history, a similar size and the same European Union rules are the best natural test of which policies actually work. If Estonia produces a fifth more per hour with the same employment, the question is what it does differently, and whether Latvia can repeat it.
What drives it
The lead can be split into four parts from the recorded data alone, without any assumptions. What one hour produces. How many hours each worker works. What share of working-age people work. And what share of the population is of working age.
Estonia: all of it is in the hour, and part is given back. Estonia's lead at purchasing power, that is, allowing for the price level, is 13 %. Output per hour gives 20 %, more than the whole lead, but Estonians work 9 % fewer hours per worker, and that takes part back. The employment rate gives almost nothing. In euro rather than at purchasing power Estonia's lead is much larger, and almost all of the difference is a higher price level: in Estonia the same money buys less.
Lithuania: the hour and the age of the people. Lithuania's lead at purchasing power is 22 %. Output per hour gives 15 %, the working-age share another 5 %, because people come to Lithuania to work since 2019. Hours and employment change almost nothing. Lithuania's price level is almost the same as Latvia's, so there euro and purchasing power tell the same story.
Then comes the question of what the model can say. The model has six areas in which Latvia can be given a neighbour's settings: migration, foreign firms, innovation and firm growth, education and skills, health and alcohol, credit and capital markets. In each, Latvia reaches the neighbour's 2025 level within five years. The chart shows by how much each area changes Latvia's GDP per head relative to the European Union average after fifteen years.
GDP per head relative to the EU average in 2041: if Latvia had Estonia's settings
Difference from the base path in 2041 if Latvia reached the neighbour's 2025 level in each driver group over five years from 2026. A model calculation; groups can reinforce each other, so the total is not the sum of the separate effects.
Show the data
Driver group
points, EU average = 100
Migration
−2.8
Foreign investment
+5.1
Innovation and firm growth
+7.6
Education and skills
+0.2
Health and alcohol
−1.1
Credit and capital markets
+0.3
All six together
+9.2
Estonia's settings give 9 points, and almost all of it comes from two areas: foreign firms (+5) and innovation (+8). Migration the Estonian way, that is, slightly more immigration, is a minus in the model (−3). The arrivals work, but the population grows faster than output, so income per head falls a little; the workforce page explains this. Together that is 96 % of Estonia's whole lead. The model explains Estonia almost in full, and two things should temper that. Its assumptions about how much productivity exporting and foreign firms deliver are wide. And Estonia has disadvantages such a calculation does not see: fewer hours worked, the downturn after 2022, a higher price level.
GDP per head relative to the EU average in 2041: if Latvia had Lithuania's settings
Difference from the base path in 2041 if Latvia reached the neighbour's 2025 level in each driver group over five years from 2026. A model calculation; groups can reinforce each other, so the total is not the sum of the separate effects.
Show the data
Driver group
points, EU average = 100
Migration
−5.5
Foreign investment
−1.6
Innovation and firm growth
−3.2
Education and skills
+0.3
Health and alcohol
−0.6
Credit and capital markets
−0.6
All six together
−11.2
Lithuania's settings give −11 points. Only in education is Lithuania ahead of Latvia, and that gives +0.3 points; it has fewer foreign firms, about as much credit, slightly fewer exporters. The largest minus is migration (−5). Lithuania's volume of immigration adds 260 thousand people in the model and 82 thousand jobs: most of the arrivals come to posts employers have offered, and the economy is 6.5 % larger. But the population grows faster than output, newcomers earn less in their first years, and unemployment is 1.9 points higher, so GDP per head falls. The cost to the budget is small: 1.9 points of government debt. An inflow of Lithuania's size makes Latvia larger, not richer per head, within fifteen years.
The conclusion is uncomfortable but clear: Lithuania's lead, about 17 points, sits where the model has no policy at all. The recorded data show where. Productivity in Lithuanian industry has grown about twice as fast as in Latvian industry since 2000. Lithuania gained twice as much from people moving out of farming into other sectors, while a larger share of Latvians still work in farming than in either neighbour. And Latvia's credit boom before 2008 put its money into construction and property, Lithuania's smaller boom more into production. None of that is the result of one decision, and none of it is in the model.
What Estonia and Lithuania do differently
Three things in which the neighbours differ and which Latvia can change, and two in which they differ but which are harder to change. For each: what differs on the recorded data, what the model says about it, and where this site has more.
Foreign firms and innovation (Estonia)
What differs
Estonian businesses spend 1.18 % of GDP on research and development, Latvian ones 0.36 %, and almost twice as large a share of people in Estonia work in foreign-owned firms as in Latvia. Estonia's exports are 78 % of GDP, Latvia's 63 %.
What the model says
these are the two areas that give almost the whole explanation of Estonia's lead, +5 and +8 points, but the model most likely overrates them.
Where there is more
on the page about catching up with Europe, where attracting foreign firms and research support are two of the packages.
Immigration for work (Lithuania)
What differs
Lithuania issued 54,000 first residence permits in 2025, Latvia 9,100. Lithuania's net migration is +5.8 per thousand residents, Latvia's −0.7. Lithuania's arrivals are workers invited by employers.
What the model says
an inflow of Lithuania's size is hired, 82 thousand more people in work after fifteen years, and the economy is 6.5 % larger. GDP per head is nevertheless 5 points lower, because the population grows faster than output and newcomers earn less in their first years. Lithuania's lead per head is not its immigration; its immigration is what a larger economy looks like.
Where there is more
on the workforce page, where opening immigration is the cheapest policy per worker.
Alcohol and health (both)
What differs
in Latvia 14.2 litres of pure alcohol per adult are drunk a year, in Estonia 10.4, in Lithuania 11.2. Life expectancy in Latvia is 76.4 years, in Estonia 79.4; healthy years after 65 are 4.2 in Latvia and 8.9 in Lithuania.
What the model says
Estonia's health and alcohol settings would prevent 54 deaths per 100,000 residents a year in Latvia, but they do not change GDP per head, and a health budget at Estonia's level raises debt by 20 points. It is an area of lives, not of growth.
Where there is more
on the health and care page, where excise is the only policy that brings the budget money.
Credit and the price level (Estonia)
What differs
Estonia's private sector debt is 95 % of GDP, Latvia's 58 %; household debt is twice as large. Estonia's price level is much higher, which is why its lead in euro is larger than its lead at purchasing power.
What the model says
Estonia's depth of credit cannot be reached in Latvia by any policy within ten years; the caution of banks after 2008 holds credit back in the model, and the credit area gives only 0.3 points.
Where there is more
on the page about catching up with Europe, where the credit package is one of the smallest.
Industry and farming (Lithuania)
What differs
Lithuanian industry produces a much larger share of output than Latvian industry, and a smaller share of Lithuanians work in farming. Lithuania's share of people with tertiary education is 50 % against Latvia's 41 %.
What the model says
nothing, and that is honest: the model has no policy that moves people between sectors or investment between production and property. That is the largest unanswered part of this page.
Where there is more
nowhere on this site yet; it is the next thing the model would have to learn.
What would bring Latvia closer to its neighbours
This is the only page on the site without packages calculated to 2040, and there is a reason. A neighbour's settings are not a policy that parliament can pass: they are the sum of fifteen years of decisions, accidents and history. What Latvia can decide is examined on the page about catching up with Europe, and the answer to this question is there too. Attracting foreign firms, research support, bringing envelope wages onto payslips and education together raise Latvia's GDP per head relative to the European Union average by about a fifth of Lithuania's current lead by 2040. That is not nothing, but it is not Lithuania either.
Two things the neighbours show that are not examined for Latvia on any page. The first is exports: Estonia's exports are 78 % of GDP, Lithuania's 71 %, Latvia's 63 %, and all three countries have grown with their exports since 2000.
The second is research. Estonia has more than tripled its spending on research and development since 2000, to 2.0 % of GDP; Latvia is at 0.9 %, and the largest difference is in business spending, not public spending. The model says innovation is one of the two areas that explain Estonia's lead. But that explanation rests on an assumption about how much productivity research really delivers, and that assumption is one of the least certain in the whole model.
What Latvia does not need to copy: Estonia's depth of credit, which also gave it a deeper crisis, and Lithuania's volume of immigration without employer demand. Neither country has found how to give pensioners a larger share of growth. With Estonia's settings Latvia's pensioner poverty risk would be 1.8 percentage points higher in the model, because wages grow faster than pensions.
Assessment
The table holds the latest recorded data on which the three countries can be compared. These are not model calculations.
Indicator
Latvia
Estonia
Lithuania
Year
GDP per head, EU average = 100
70.2
79.8
87.6
2025
Employment rate, 20 to 64, %
78.2
81.7
80.0
2025
Employment rate, 55 to 64, %
73.6
77.0
69.9
2025
Unemployment rate, %
6.9
7.5
6.9
2025
Tertiary education, 25 to 64, %
40.7
42.1
49.9
2025
Research and development, % of GDP
0.94
1.98
1.12
2024
Business research and development, % of GDP
0.36
1.18
0.50
2024
Investment, % of GDP
23.5
23.6
21.9
2025
Exports, % of GDP
62.6
78.2
71.3
2025
Private sector debt, % of GDP
57.5
95.4
55.1
2025
Household debt, % of GDP
19.8
39.3
23.8
2025
Taxes and contributions, % of GDP
35.5
36.8
33.3
2024
Tax wedge at the average wage, %
38.9
39.3
36.1
2025
Shadow economy, % of GDP
21.8
20.8
23.6
2025
Net migration per 1,000 residents
−0.7
−2.1
5.8
2025
First residence permits, number
9,073
5,228
53,566
2025
Children per woman
1.24
1.18
1.11
2024
Life expectancy at birth, years
76.4
79.4
77.5
2024
Healthy life years at 65
4.2
7.6
8.9
2024
Recorded alcohol sales, litres per adult
12.0
12.4
10.3
2024
Income inequality (Gini)
35.6
30.9
35.5
2025
Government effectiveness (World Bank index)
0.83
1.50
1.24
2025
The table says three things. First, Latvia is not "the Baltic laggard". Employment, investment, the tax burden and the shadow economy are close in all three countries, and the idea that Estonia's shadow economy is much smaller no longer holds on the latest data. Second, the two rows in which Latvia is furthest behind are business research and health, and they call for different answers. The first is a growth question, the second a question of lives. The gap in healthy years, 4.2 against 8.9, is the largest relative gap in the whole table. Third, the row on government effectiveness, in which Latvia trails both, gives nothing in the model. The model has no path by which better administration becomes higher productivity, and that is a gap in the model rather than in the administration.
For a decision-maker this gives: Estonia is a clear example Latvia can copy with policies already calculated on this site: foreign firms, exporters, research. Lithuania is an example most of which lies outside any policy calculated here. Before copying it one has to understand what exactly happened in its industry and its countryside after 2010.
What to keep in mind
The model part of this page is a test, not a plan. "If Latvia had Estonia's settings" is not a policy, and the result is shown after fifteen years, in 2041, not in 2040 as on the other pages, because it is a different calculation with a different start. It is best read as the answer to the question "can the model's drivers produce a gap as large as the real one", and the answer is: for Estonia almost in full, for Lithuania not at all.
Explaining Estonia in full rests on two assumptions. How much productivity exporting firms deliver and how much foreign firms deliver are assumptions with a wide range in the model. At the lower end of the range Estonia's settings explain clearly less than the whole lead, and it is the lower end that the Baltic data support: Estonia's foreign-owned firms are only slightly more productive than its domestic ones in Estonia itself. That means the foreign firms and exports packages on the page about catching up with Europe may be overrated too.
The sign on migration for Lithuania is about income per head, not about the economy. The model's arrivals are mostly hired on arrival, as Lithuania's are, and the economy grows with them. What falls is income per head, because the population grows faster than output and newcomers earn less in their first years. The model employs about three-quarters of an extra labour inflow within fifteen years; the workforce page says what that rests on. If integration is faster than the model assumes, the minus is smaller.
Lithuania's residual is not a question of assumptions. However the model's assumptions are varied within their range, Lithuania's settings give Latvia a minus. The residual is productivity in industry, the move out of farming and the direction of investment before 2008, and none of them has a policy in the model. That is the next thing the model would have to learn before this page can say more.
Euro and purchasing power tell different stories about Estonia. Estonia's lead in euro is almost twice its lead at purchasing power, because the price level in Estonia is higher. This page uses purchasing power throughout, because it shows what people can actually buy with their money; a Latvian who compares wages in euro sees a larger gap.
There are two alcohol rows. The World Health Organization's estimate of what is drunk puts Latvia far ahead of its neighbours; recorded sales in the table put Estonia close to Latvia, because Estonians buy part of their alcohol in Latvia. Both are correct, they measure different things, and the difference between them is the cross-border trade that is one of the assumptions on the health page.
Some of the data are older than 2025. The share of foreign-owned firms is from 2020, the skills comparison from 2022, research spending from 2024. The table gives the year beside each row.