Long-term policy questions for Latvia

Most of what will happen in Latvia in 2040 is already written into today. How many people will be of working age is set by those already born. What pensions will cost is set by contributions already made. What power plants will emit is set by those already under construction. Quantities like these have inertia. They change slowly, and a decision taken today becomes visible after five, ten or twenty years.

That is why the question on this site is always "what happens over fifteen years", not "what will happen next year". The model follows these slow quantities year by year, and follows how they affect one another. Some of these effects form a circle. Fewer people in work means a slower economy, a slower economy means more emigration, more emigration means fewer people in work still. That is a reinforcing loop. Other effects correct themselves. When there are suddenly more people who want to work, wage growth slows until firms take them on. That is a balancing loop. On every question page the section "What drives it" says which of these is the main one for that question, or that the main thing is not a circle at all but one assumption.

Two features of the base path run through every page, and both come from the labour market. The first is a slump in the 2030s. The workforce shrinks, retirements open posts faster than they can be filled, wages rise by 2030, and the economy cools in the middle of the decade before it recovers. How deep that slump is rests on three judgements with no Latvian record behind them: how easily firms replace workers with equipment, how long they take to open the posts they want, and how many of the posts that departures leave are advertised again. The second is absorption. When a policy adds people who want to work, the model employs about three-quarters of them within two decades and reports the rest as unemployment, because capital and demand take time to grow with the workforce. The European Commission's Ageing Report assumes, by convention, that all of them work. Every employment figure on the site is read with that in mind.

The result on every page is a base path, that is, what happens if nothing changes, and several packages that change one or more policies from 2028. The difference between a package and the base path is what the policy delivers and what it costs. Costs to the budget are shown in two ways, because Latvia's public finances are bound by the European Union's fiscal rules. First, a package's own effect on the budget. Second, what is left of it in government debt once the rules have responded, which is usually much less. A package that spends is mostly paid for by cuts the rule forces elsewhere, and a package that earns mostly lets the rule cut less. It is a calculation, not a forecast. It shows where today's assumptions lead, and every page names the assumptions its answer rests on most.

The model is one connected picture of Latvia's people, economy and public finances. It has fourteen modules, one per field, and they calculate together. At every step each module reads what the others produced and writes what it produced for them. In numbers: about nine thousand equations, of which about seven hundred describe quantities with inertia, the ones that carry the past into the future; 271 links through which the modules exchange quantities; and 113 named judgements, that is, parameters the record does not pin down, each with a plausible range. Two of those judgements are drawn on every page as its uncertainty band. The diagram shows the modules and thirty of the main links.

People. The population module follows every age group year by year: births, deaths, arrivals and departures, and how emigration responds to the wage gap with other countries. Arrivals come in three streams: Latvians who return, citizens of other EU countries, and people from outside the Union, who either settle or are invited by an employer to fill a post. The households module splits income across five income groups and calculates the poverty risk, inequality and benefits.

Work and the economy. The labour market module matches people who want to work with jobs by education and sector and calculates unemployment and wages. The economy module calculates output, investment, prices and how fast Latvia's productivity catches up with the European average. The firms module follows productivity, exporters, foreign-owned firms and the shadow economy. The banks and credit module calculates lending to households and firms and how cautious banks are after a crisis. The foreign trade module follows exports, imports and what the country owes abroad.

Public money. The state budget module calculates taxes from three bases, spending across eight functions, European Union funds and the fiscal rules, and from these the deficit and debt. Its base path starts from the 2026 budget and the European Commission's forecast for 2027. It carries the commitments already made, defence at five per cent of GDP and Rail Baltica among them, and lets interest rates return to about four per cent, which is the Commission's own convention. The pensions module calculates all three pension tiers by generation and the distribution of pensions, not only the average. The municipal finance module calculates municipal revenue, the equalisation fund and the network of schools and offices across six regions.

Services. The health module follows how many people at each age are healthy, chronically ill or dependent on care, how many of them the budget and the number of doctors can treat, and how many relatives provide care themselves. The education module follows pupils from kindergarten to university, teachers, the school network and learning outcomes.

Places and energy. The housing and regions module calculates where people live across six regions, how much is built and what housing costs. The energy module calculates electricity, heat, building renovation, emissions, the forest balance and what the European Union's climate targets cost Latvia.

This is a system dynamics simulation model. It is made of quantities with inertia (stocks), such as people, buildings, debt and capital, and of the flows that change them, and it is solved year by year from 2000. It is built so that each module also runs on its own, with the others replaced by recorded data, which is how it is checked.

A model of this size earns credibility only by reproducing what actually happened. So the parameters are set against the record in four layers, each of which can be inspected separately.

Each module reproduces its own field, 2000 to 2025. The parameters that describe how things are, such as death rates by age, the share of pupils in each kind of school, the energy content of buildings, are read from the data directly. The parameters that describe how people and firms respond, such as how fast wages rise when workers are scarce or how quickly a boiler is replaced, are estimated by running the module against the record. The other modules are replaced by recorded data, and the module is asked to match several series at once. A part of the record is held back from the estimation and used as a test. A module that fits every series it was tuned on and misses the one it was not has learned the data, not the mechanism.

The modules are also calibrated together. A module that is right on its own receives modelled inputs instead of recorded ones once it is connected to the others, and fourteen small errors can compound. So the connected system is calibrated as a whole against six series: real output, prices, the consumption share, the wage share, government revenue and government spending. Government debt and the pension replacement rate are deliberately held out as the test. Debt is the sum of twenty-five years of revenue and spending errors, and a system that gets debt right by accident would not pass it.

What Latvia's own record cannot tell, the neighbours' records can. Some responses cannot be identified from one country's twenty-five years. How fast a country's technology catches up with the frontier as it gets closer. How strongly emigration responds to the income gap. How much more productive foreign-owned firms are. How the shadow economy responds to the tax wedge on labour. How alcohol consumption responds to its price. These are estimated on a panel of eight countries, Estonia, Lithuania, Poland, Czechia, Slovakia, Slovenia and Hungary beside Latvia, with the European Union average as the reference, from the same harmonised sources for 2000 to 2025. The Baltic-only estimate is reported beside the eight-country one. A number that cannot be told from Latvia's data alone is still a number with evidence behind it, not a guess.

What remains is declared, and shown. After the three layers, 113 parameters are still judgements: values from the economic literature, from the European Commission's conventions or from a decision taken openly, each with a plausible range. All of them are screened for how much they move each page's headline. The two that matter most for each question are varied across their ranges and drawn as the uncertainty band on the page's headline chart; the rest are listed in the model's documentation. Every number on the site is then checked once more against 73 recorded values: 65 are within tolerance, and the 8 that are not are known and named on the pages they concern. The external benchmarks the model did not use, the Commission's 2024 Ageing Report, its spring 2026 forecast and Latvia's 2026 Draft Budgetary Plan, are checked the same way. Where the model misses, the page says so in "What to keep in mind".

On every page the dots in the charts are recorded data. They come from public sources, and every data series is recorded in the site's internal documentation with its source, the date it was retrieved and a quality mark.

Population and migration. The Central Statistical Bureau (CSB) and Eurostat: population by age group, births, deaths, arrivals and departures, 2000 to 2025; Eurostat's population projections as a comparison; the Office of Citizenship and Migration Affairs and the CSB: residence permits by reason, and residents by citizenship.

Labour market and wages. The CSB labour force survey and Eurostat: people in work, unemployment, employment by education and sector, vacancies, the average wage and productivity, 2000 to 2025.

The economy and public finances. Eurostat national accounts and government finance statistics: gross domestic product and its components, government revenue, spending by function, the deficit and debt, 2000 to 2025; the Ministry of Finance's 2026 Draft Budgetary Plan and the European Commission's spring 2026 forecast for the near term; the Commission's 2024 Ageing Report as an external benchmark; the public record of Rail Baltica's remaining cost and its European co-financing.

Pensions. The CSB and the State Social Insurance Agency (VSAA): the number of pensioners, the average pension, the distribution of pensions by amount and age group, contributors; the Bank of Latvia: pension fund assets.

Health and care. Eurostat, the CSB, the Organisation for Economic Co-operation and Development (OECD) and the World Health Organization: life expectancy, healthy years, avoidable deaths, waiting lists, the number of doctors and nurses, health spending, 2004 to 2025; long-term care data are sparse and mostly from the European Commission's 2016 review.

Education. The Ministry of Education and Science and the CSB: pupils, teachers, the school network; the OECD: learning outcomes (PISA); Eurostat: education spending.

Housing and regions. The CSB: population, employment and services across six regions, the housing stock and construction, 2000 to 2025; the Bank of Latvia and Eurostat: mortgages and housing costs; property market reviews: prices by region.

Energy and climate. Eurostat, the International Energy Agency and the European Commission's climate reports: the energy balance, prices, emissions by sector and the forest balance; the Ministry of Agriculture: forests and felling; the State Audit Office: renovation costs; the CSB household budget survey: energy spending by income group.

Firms and banks. The SSE Riga shadow economy index, European Commission business statistics, Eurostat, the Bank of Latvia and the World Bank: the number of firms, investment, credit, household and corporate debt.

Other countries. The same Eurostat series for Estonia and Lithuania, and for Poland, Czechia, Slovakia, Slovenia and Hungary where they are used to estimate parameters, 2000 to 2025; the World Health Organization's and the World Bank's country panels, including the governance indicators.

Known gaps: some series before 2005 are reconstructed rather than measured; data on long-term care, regional services and the economics of forestry are sparse; 2025 values for part of the public finance series are not yet published. Where the model does not match the record, the page says so in "What to keep in mind".