Benchmarking Latvia’s economy: a new estimate of gross domestic product in the 1930s

Abstract

The interwar independent Republic of Latvia was among the first ten pioneering states, where a national statistical office published official estimates of total output (1934–1936). Paradoxically, however, Latvia is the Baltic country with the most disputed interwar economic growth performance. According to the authoritative account of Roses and Wolf in The Cambridge Economic History of Modern History (2010), Latvia‘s GDP per capita growth rate was the highest among European countries in 1929–1938. It impressively ranked number ten according to GDP per capita next to Sweden, France, and Norway. However, according to Norkus and Markevičiūtė (in Cliometrica 15:565–674, 2021. https://doi.org/10.1007/s11698-020-00216-z), it only surpassed Southern European countries, and its growth performance was mediocre. Both these contradictory estimates are derived by indirect methods. This paper contributes to the resolution of this controversy, directly estimating Latvia’s GDP in 1935 within the SNA 2008 framework, providing gross value-added calculation for 20 ISIC industries at basic and at market (purchasers’) prices. It provides a more fine-grained analysis of the composition of Latvia’s total output in comparison with interwar historical national accounts, where only 11 industries were distinguished. This estimate provides the benchmark for future research on Latvia’s interwar economic growth performance. Converting our estimates into monetary units, used in the Maddison Project Database, we assess Latvia’s position in the international GDPpc ranking, coming to conclusions that dovetail with Norkus and Markevičiūtė (2021) findings.

Appendices

Appendix 1: Statistical details of Latvian 1935 GDP calculation

Table 15 presents calculation of total agricultural land. This land consists of the land used for crops and gardening, animal husbandry (meadows and pastures), and private forests.

Table 15 Total agricultural land in Latvia in 1935 in hectares

Table 16 contains data used for calculation of intermediate consumption (actual expenses for crop fertilizers, seeds, animal fodder and other inputs per hectare). The aggregate amount of IC is then derived multiplying this expenditure by area used for crop production and animal husbandry. Finally, the GVA at basic prices of agriculture is obtained by subtracting IC from GVO. It looks like that agricultural accounts provide the information about the GVO at basic prices. In fact, Latvian farmers did not pay any product taxes, which otherwise should be subtracted from the GVO at producer’s prices to arrive to GVO at basic prices. Some products (food grain, milk and bacon pork) were subsidized. However, as far as farmers did sell these products directly to government-sponsored agencies, subsidies are already included into value of these products according to agricultural accounts, and there is no need to add them to GVO at producer’s prices to get GVA at basic prices (they should only be subtracted (adding taxes) estimating GVA at market prices).

Table 16 Calculations of GVA from agriculture (in Ls) 1935

Tables 17 and 18 present calculation of GVA of forestry and fishing correspondingly.

Table 17 Calculations of GVA (in Ls) from forestry 1935

Table 18 Calculations of GVA (in Ls) from fishing 1935

Next four Tables 19, 20, 21, 22 and 23 provide calculation of the GVA in five secondary ISIC industries, based on the formula GVA_i = GVO_i—IC_i. (see Sect. 3.1) and algorithm described in Sect. 3.2.2. Table 19 presents calculation for mining and quarrying (B); Tables 20 and 21 for manufacturing (C); Table 22 for electricity, gas, steam and air conditioning supply (D) and water supply, sewerage, waste management and remediation activities (E); Table 23 for construction (F). Rows in the table (C) present branches of manufacturing according to classification of Latvian national statistical office (see Sect. 3.1 in the main text).

Table 19 Calculations of GVA (in Ls) in the mining and quarrying (ISIC sector B)

Table 20 Calculations of GVA (in thousands Ls) in the manufacturing (ISIC sector C)

Table 21 Calculations of total GVA (in Ls) from manufacturing (ISIC sector C)

Table 22 Calculations of GVA (in Ls) of electricity, gas, water, steam and air conditioning supply (ISIC sector D) and water supply, sewerage, waste management and remediation activities (ISIC sector E)

Table 23 Calculations of GVA (in Ls) from construction

Table 24 provides the calculation of the GVA of wholesale and retail trade as part of “trade in the broad meaning”, surveyed by Latvian trade census on 26.06.1935.

Table 24 GVA (in Ls) from wholesale and retail trade (ISIC sector G)

The same data source is used also in the calculation of the GVA at basic of transportation of storage (ISIC sector H). However, this is only part of data. Another part is provided by Latvia’s central government budget for 1935 (Latvijas Republika. Valsts budžets 1935/36 saimniecības gadam 1935), which includes revenue and expenses of the state-owned enterprises in separate parts. We consider total revenue of post as its GVO at producer’s prices and identify intermediate consumption with its operating expenses, amounting to 823103 Ls. We divide these expenses by the post share in total revenue, which is 37.86%. Remaining shares belong to telegraph (4.74%) and telephone (57.40%).

For railways, we use data on total revenue and interpret it as GVO at producer’s prices (Valsts dzelzceļu izdevniecība 1938). Like Latvian post, railways were state-owned. Their net profits were appropriated by the state. Railway and post employees, like other public sector workers, paid income tax. We include all ordinary expenses for running railways into IC, except for salaries. The taxes, computed for sector H (see Table 8 in the main text), are divided into 3 parts: 61.20% for ‘trade in broad meaning’, 7.29% for post and 31.51% for railways. These ratios/percentages are obtained from their shares in the GVA excluded taxes (Table 25).

Table 25 GVA (in Ls) from transportation and storage (ISIC sector H)

The calculation of the GVA of accommodation and food service activities (see Table 26) is based on relevant part of data of the 1935.06.26 trade census. However, we noticed that according to the 1935 Population Census data (Salnītis 1939) there were less people working in sector I than according to the Trade census (Salnais and Jurēvics 1938). Thus, in order to avoid double counting, when some persons received wages from other sectors and worked unsalaried in hotels and restaurants, we computed the compensation for employees with the 1935 Population Census data.

Table 26 GVA (in Ls) from sector I

For the information and communication ISIC sector (see Table 27), we use data from the Latvian government budget for 1935 (Latvijas Republika. Valsts budžets 1935/36 saimniecības gadam 1935). We consider as the GVO at producer’s prices the revenue of the telegraph, telephone and radiophone. According to our interpretation, the intermediate consumption of telegraph and telephone consists of operating expenses and part of ordinary expenses. However, our source provides only total ordinary expenses (including administrative expenses, household expenses, publishing expenses, and service clothing) for post, telegraph and telephone taken together (823,103 Ls). We divide them between post, telegraph and telephone by their shares of revenue (GVO), obtaining 37.86% for post, 4.74% for telegraph and 57.40% for telephone. We include into intermediate consumption of radiophone administrative, household, and representation expenses, membership fees and operating expenses. Table 27 sums up our calculation for the ISIC sector J.

Table 27 GVA (in Ls) from sector J

Next ISIC 4 sector (K; financial and insurance activities) is nearly completely covered by the Latvian 1935.06.26 trade census data, which includes data on banking, credits, insurance, pawnshops, and other financial activities (Table

Table 28 GVA (in Ls) from financial and insurance activities (ISIC sector K)

28).

Next, Table

Table 29 GVA from real estate activities (ISIC sector L)

29 provides calculation of the GVA of real estate activities (ISIC 4 industry L), explained in detail in the main text (see Sect. 3.2.4).

Table

Table 30 GVA (in Ls) from professional, scientific and technical activities (ISIC sector M)

30 provides GVA calculation of the ISIC industry M (professional, scientific and technical activities), comprising both market (for profit) and non-market (not-for-profit activities).

Table 31, presenting calculation of the GVA of administrative and support service activities (ISIC sector N), consists of two parts. One of them comprises activities covered by the Manufacturing and crafts census data, which allow to compute intermediate consumption. Another is based on data of Trade census, where only component method could be applied.

Table 31 GVA (in Ls) from administrative and support service activities (ISIC sector N)

Table 32 presents calculation of the GVA of public administration, defence, and compulsory social security (ISIC sector O). Section "Administration of the state" in this table comprises the compensation of employees of the central offices (Presidency, the Minister Cabinet), the State Controller’s office, Ministry of Foreign Affairs, Ministry of Finance, Ministry of Agriculture. We exclude salaries of teachers at the marine schools from the expenditure by the Ministry of Finance, including them into education. We do the same with expenses for agricultural schools, which were maintained by the Ministry of Agriculture.

Table 32 GVA (in Ls) from public administration, defence, and compulsory social security (ISIC sector O)

Estimating the GVA of the social security administrated by central government, we use data on the expenditure of the Ministry of the National Welfare (Tautas labklajibas ministrija), and that on the expenses of the state co-financed medical insurance funds. However, we include the administration expenses of these offices only, assuming that part of their ordinary expenses is already included into expenses of the health care sector, encompassing also the spending on social work. We proceed the same way with expenditures of urban and rural municipalities (sector O). Remaining parts are included into sectors P and Q.

Estimating the GVA of education (see Table 33), we reclassify as parts of M the State archive, which employed 19 persons in 1935 (Baltais and Salnais 1936: 102), and the Monument board (Pieminekļu pārvaldē), employing 31 persons. We include into R the State library (38), the State Art Museum (5), and the National opera, with a staff of 344 in 1936 (Baltais and Salnais 1936, 102). The sum of these exclusions (437) reduces total employment at the Ministry of Education from 3821 to 3384 employees.

Table 33 GVA (in Ls) from education (ISIC sector P)

The exclusion of the expenditure on these institutions from the total ordinary expenses of the Ministry of Education reduces them to 20,179,998 Ls, according to state budget for 1935–1936. The report of the State control office on the execution of this budget provides no details. Hence, we can base the analysis of the contribution of public education to GDP only on this source. Thus, total planned compensation of all employees was 18,996,584 Ls, i.e. 94.13% of total ordinary expenditure.

Rural municipalities had to pay the costs of maintenance and reparation of school buildings, making out 2,027,000 Ls in 1935–1936 (Valsts statistiskā pārvalde 1936b, 27). This was also the responsibility of urban municipalities, which had to maintain the infrastructure of secondary education. However, in addition, urban municipalities had to contribute to compensation for teachers of secondary schools. In the absence of the information about the exact use of the 11,843,000 Ls, spent by Latvian urban municipalities on education in 1935, we are following Klimantas and Zirgulis (2020). They had the same difficulty interpreting education expenditures of Lithuanian urban municipalities. These, differently from rural municipalities, had to contribute to paying salaries to teachers of secondary schools. According to their judgment, salary expenditures made up 37.5% of the education expenditure of Lithuanian urban municipalities. Applying the same ratio to education expenditures of Latvian urban municipalities, we arrive at 4,441,125 Ls of their contribution to the CE in the P sector.

As for CE and NOS (see Table 31), we add the income tax paid by the workers of P, which is estimated from data by the Latvian tax authorities, and employment in education according to the 1935 census. We decrease its size (from 268,740 to 260,717 Ls) by the putative share of 437 persons, which were reclassified as workers in M an R (8,023 Ls), and CFC, calculated by assuming that the CFC norm is equal to 1/12 GVA. CE by the central government, municipalities and private education enterprises, NOS of private enterprises, taxes on production and CFC amount to 26,866,812 Ls. Converting this sum into Lithuanian currency according to the currency exchange rate in 1935 (1 Lt = 0.518 Ls, or 1 Ls = 1.93 Ls), the added value of sector P in Latvia was 51,852,947 Lt, which by 40.48% exceeds the added value of P in Lithuania (36,910,226 Lt according to Klimantas and Zirgulis (2020) despite larger population of Lithuania.

Next, Table 34 presents the calculation of the GVA of the human health and social work activities (ISIC sector Q). Compensation of employees (10,402,444 Ls) constitutes 63.5% of ordinary expenses, which is rather close to 59%, derived for neighbour Lithuania in 1937 by Klimantas and Zirgulis (2020, p. 276). Our source reports 393,056 Ls in expenses on reparation work, which belong to CFC. CFC also includes expenses on the replacement of the long serving medical inventory. In the absence of information, we make the guess that these expenses were no less than 20% (198,324.2 Ls) of total inventory expenditures (991,621 Ls). Adding total CFC (591, 380, 2 Ls) to CE and subtracting 33,672 Ls in taxes, one arrives at 10,960,152.2 Ls of health care contribution to the Latvian GDP in 1935, where CFC makes out 5.4%.

Table 34 GVA (in Ls) from human health and social work activities (ISIC sector Q)

Gold parity Latvian and Lithuanian currency exchange rates in 1935 were 1 Lt = 0.518 Ls, or 1 Ls = 1.93 Ls. The added value of sector Q in Lithuania in 1937 amounted to 2,643,059.3 Ls or 5,102,431.1 Lt (Klimantas and Zirgulis 2020: 276). It was 22.9% of its Latvian counterpart, although the population of Lithuania (2,549,668 on 31.12.1937; see Centralinis Statistikos Biuras 1938, 13) was markedly larger than the population of Latvia (1,950,502 according to 15.02.1935 census). This disparity stands in neat correspondence to findings of recent research on the uneven advancement of the welfare state in the interwar Baltic States (Norkus et al. 2021).

Tables 35, 36 and 37 provide calculation of the GVA for remaining ISIC 4 industries: R (arts, entertainment and recreation), S (other service activities), and T (activities of households as employers; undifferentiated goods- and services-producing activities of households for own use). Calculation of the GVA of arts, entertainment and recreation in Table 35 consists of two parts. One of them is based on the data of the 26.06.1935 Trade census and comprises commercial activities. Another comprises activities of public sector. Information on CE for museums is given by Salnītis (1938). These data are for 1936 and 1937. We use them as they are the best available data. For the theaters, we compile CE from Salnītis (1938, 143) and we assume no profit or taxes. Similarly, we compute GVA for the libraries.

Table 35 GVA (in Ls) from arts, entertainment and recreation (ISIC sector R)

Table 36 GVA (in Ls) from other service activities (ISIC sector S)

Table 37 Calculations of GVA (in Ls) from activities of households as employers; undifferentiated goods- and services-producing activities of households for own use (ISIC sector T)

Calculation of the GVA of other services (ISIC industry S) consists of two parts too. One of them comprises activities covered by the data collected during the 26.06.1935 Trade census, included into “trade in the broad meaning”. Another refers to services provided by the professional workers of religious communities. The Latvian government did not pay salaries for employees in churches. All established confessions received state subsidies, but they were used exclusively for reparation of old or construction of new church buildings. Using data provided by Balevics (1978, 220) on fees collected by Catholic priests and those in Baltais and Salnais (1936, 23–24) on prices of various clerical services in Lutheran churches, we estimate that fees in Lutheran and other Protestant churches were 3 Ls for baptisms, 18 Ls for marriages and 60 Ls for funeral. In Orthodox and Catholic churches, they were, respectively, 2 Ls, 12 Ls and 40 Ls. According to Salnītis (1938, 106), there were 56.13% Lutherans and other Protestants, 8.94% of Orthodox, and 24.45% Roman Catholics. Thus, the total number of marriages (16,474), births (34,419) and deaths (27,660) were divided by these proportions and used to compute compensation for employees in churches.

Last table in this appendix (Table 37) refers to activities of households as employers; undifferentiated goods- and services-producing activities of households for own use. There are two parts. One of them comprises work of domestic servants (3846 males and 28,788 females according to 12.02.1935 Population census). Taking the average salary of unskilled workers [2.35 Ls for males and 1.55 Ls for females per day (8 h)], we compute annual compensation of employees, assuming that workers worked 25 days per month during 12 months. Another part includes farmers’ work outside the farm, which was accounted for in the agricultural accounts. We assume no taxes or subsidies for these activities. Since we also do not have the IC, we use the ratio of GVA to GVO from agriculture.

Appendix 2: Update of Lithuanian interwar purchasing power parity estimates and estimation of purchasing power parity of Latvian lats

Purchasing power is the power of money expressed by the number of goods or services that one unit can buy. Purchasing power parity (PPP) is based on the idea that goods in one country will cost the same in another country. According to this concept, two currencies are in equilibrium or at a par, when the following equality holds: S = P₁/P₂, where S = exchange rate of currency 1 to currency 2; P₁ = cost of good X in currency 1; P₂ = cost of good X in currency 2. If P₁/P₂ is less than S, then this good is relatively cheaper in the country using currency 1; if P₁/P₂ is larger than S, then it is relatively more expensive in the country using currency 1. To make a meaningful comparison of price levels across countries, a wide range of goods and services should be used when composing product baskets.

In this appendix, we describe the methodology and price baskets used for the calculation of purchasing power parities (PPP) of Latvian, Swedish and Lithuanian currencies. The main aim is to compare a weighted average of relative prices across these countries, drawing on index number theory (Bolt et al. 2018). Denote p_j—the vector of prices in country $j$ and let q_j be the vector of products. Nominal GDP in country $j$ is then ${Y}_{j}={{\varvec{p}}}_{{\varvec{j}}}{{^{\prime}}}\cdot {{\varvec{q}}}_{{\varvec{j}}}$, the sum of spending on (domestic) products. Given these vectors for two countries, we come to the idea “what would a person in country $k$ have to spend to purchase the same bundle of products as a person in country $j$”. This question might be answered with the Laspeyres price index:

$${P}_{jk}^{L}=\frac{{{\varvec{p}}}_{{\varvec{k}}}^{{\prime}}\cdot {{\varvec{q}}}_{{\varvec{j}}}}{{{\varvec{p}}}_{{\varvec{j}}}^{{\prime}}\cdot {{\varvec{q}}}_{{\varvec{j}}}}=\frac{\sum_{i=1}^{n}{p}_{ki}{q}_{ji}}{\sum_{l=1}^{n}{p}_{jl}{q}_{jl}}=\frac{\sum_{i=1}^{n}\left(\frac{{p}_{ki}}{{p}_{ji}}\right){p}_{ji}{q}_{ji}}{\sum_{l=1}^{n}{p}_{jl}{q}_{jl}}=\sum_{i=1}^{n}\left(\frac{{p}_{ki}}{{p}_{ji}}\right){s}_{ji},$$

where ${s}_{ji}= \frac{{p}_{ji}{q}_{ji}}{\sum_{l=1}^{n}{p}_{jl}{q}_{jl}}$ is the share of expenditures of good $i$ in country $j$.

We arrive at the Paasche index, departing from the prices of country k:

$${P}_{jk}^{P}=\frac{{{\varvec{p}}}_{{\varvec{k}}}^{{\prime}}\cdot {{\varvec{q}}}_{{\varvec{k}}}}{{{\varvec{p}}}_{{\varvec{j}}}^{{\prime}}\cdot {{\varvec{q}}}_{{\varvec{k}}}}=\frac{\sum_{l=1}^{n}{p}_{kl}{q}_{kl}}{\sum_{i=1}^{n}{p}_{ji}{q}_{ki}}=\frac{\sum_{l=1}^{n}{p}_{kl}{q}_{kl}}{\sum_{i=1}^{n}{\left(\frac{{p}_{ji}}{{p}_{ki}}\right)p}_{ki}{q}_{ki}}=\frac{1}{\sum_{i=1}^{n}\left(\frac{{p}_{ji}}{{p}_{ki}}\right){s}_{ki}},$$

where ${s}_{ki}= \frac{{p}_{ki}{q}_{ki}}{\sum_{l=1}^{n}{p}_{kl}{q}_{kl}}$ is the share of expenditures of good $i$ in country k.

Nevertheless, neither of these indices is preferable as there is no reason why either bundle of products should hold a privileged position. Thus, we use the Fisher price index, which is a geometric mean of both—Laspeyres and Paasche indices:

$${P}_{jk}^{F}={\left[{P}_{jk}^{L}\times {P}_{jk}^{P}\right]}^{1/2}.$$

We use the same consumption goods basket as Klimantas and Zirgulis (2020). However, we did find that a slight correction of Lithuanian 1937 GDP estimates from this paper is required. Making this correction, we express the Laspeyres, Paasche and the resulting Fisher indices for the Lithuanian estimate are expressed in SEK/Litas instead of Litas/SEK (cp. Table 3 in Klimantas and Zirgulis (2020): 248). This shifts the 1937 GDP PPP estimate by Klimantas and Zirgulis upward by 1.4% to 2018.70 GK 1990$. Drawing from further revisions and preliminary time-series estimates by Klimantas (2020), for 1935 the Lithuanian GDP per capita in 1937 prices, was 717.9 Lt. Considering that the retail price index in Lithuania increased quite significantly by 21.4% from 1935 to 1937, we convert Lithuania’s GDP per capita into 1935 prices. Thus, we get that the Lithuanian current-price GDPpc in 1935 was 591.34 Lt, which is equivalent to 1900.37 GK 1990$. We similarly convert GDPpc in Litas to 2011 international $, that is equal to 3029.01 int 2011$. The basket of priced goods and services is provided in Table 38.

Table 38 Priced goods and services basket. Sources: Klimantas (2019), Norkus et al. (2022a, 2022b), Annual Average Retail Prices of Food in Lithuania, 1913–1939 https://hdl.handle.net/21.12137/UN7JZ9, Lithuanian Data Archive for HSS (LiDA), V2; Norkus et al. (2022a, 2022b), Annual Average Retail Prices of Non-Food Goods in Lithuania, 1913–1939 https://hdl.handle.net/21.12137/QFUBCC, = Lithuanian Data Archive for HSS (LiDA), V2; Official Statistics of Sweden (1961) Consumer prices and index computations 1931–1959. Stockholm: The Swedish Social Welfare Board

We divide items from our consumption goods and services basket into four categories: food (22 items), housing (1 item), clothing (6 items) and other goods and services (11 items). Then using data on the distribution of consumption expenditure between these four categories (see Table 39), we approximate the share values ${s}_{ji}$ and ${s}_{ki}$ dividing the expenditure proportional parts by number of products in each category.

Table 39 Ratio of goods and services in the goods and services basket in each country in certain years. Sources: Official Statistics of Sweden (1961); Consumer prices and index computations 1931–1959. Stockholm: The Swedish Social Welfare Board; Klimantas (2020); Darba statistika (1936, p. 8)

Data on the composition of consumption expenditure were published in Darba statistika by Latvian national statistical office, which did conduct the research on the budgets of Latvian families (see also Davidsons 1940). This research was limited to urban population. Publishing its data, Latvian statisticians did distinguish families from capital city (Riga) and from province cities. Both categories were then divided into families of workers and those employed in service sector (officials and servants). Estimating the composition of consumer expenditure on the national level, we assumed that composition of consumer expenditure of agricultural population was close to that of urban workers, merging them with agricultural population into category of manual workers. We included into this category also home servants, considering the remaining part of population employed in the service sector as non-manual workers. Then, we used the shares of manual and non-manual, capital city and province population according to 1935 population census as group weights to compute weighted average of expenditure on food, housing, clothing and other products shares in total consumption expenditure on national level.

PPP indices (SEK/Ls, Lt/Ls conversion rates) are provided in Table 10 in the main text. Further, we convert GDP per capita values for Sweden (1558.02 Kr) and Lithuania (591.34 Lt) into values in Lats using the computed indices (see Table 40). Then, using Latvia’s GDP per capita 584.27 Ls at market prices, we find that Sweden had a 1.63, 1.94- and 1.36-times larger GDP per capita than Latvia in 1935 by the Fisher, Paasche, and Laspeyres index, respectively. Lithuania’s GDPpc at PPP made 68% (by the Fisher index), 81% (by the Paasche index), 57% (by the Laspeyres index) of Latvia’s GDPpc.

Table 40 GDP per capita conversion to GDP PPP

Finally, we use these ratios to convert Latvia’s GDP per capita into international dollars by using the Swedish GDP per capita 4491.73 GK 1990$ (MPD 2013) and 7160,2011$ (MPD 2020) and Lithuanian GDP per capita 1900.37 GK 1990$ and 3029.01 int 2011$. The final results are given in Table 11 in the main text.

Appendix 3: A comparison of our GDP estimate with calculations of national income of Latvia in 1935 by its national statistical office

There are two issues with this comparison. Firstly, GDP in interwar time was still uninvented, so interwar authors did estimate national income. GDP is the total market value of all finished goods and services produced within a country; national income is the total income received by the country from its residents and businesses regardless of whether they are located in the country or abroad, with the income of foreign residents subtracted. Interwar Latvia’s residents did not invest abroad, and there were no noticeable remittances from abroad (Aizsilnieks 1968). However, there were significant investments of foreign capital in Latvia, and there was considerable private and public foreign debt. According to data on Latvia’s payment balance data (Valsts statistiskā pārvalde 1936b: 273), it received only 0.2 mil. Ls of interest from foreign debtors and for capital invested abroad, but paid 4.45 mill. Ls. There is no data on remittances from Latvia’s residents employed abroad. However, Latvia was dependent on seasonal agricultural workers from Lithuania and Poland, who jointly with few foreign employees in Latvia’s industry did transfer 3.08 mill. Ls. In 1935 to their homelands. So Latvia’s gross national income (GNI) was at least 7.28 mill. Ls less than its GDP. However, this figure is not so large to make both figures incomparable.

Secondly, the sources provide two different figures for 1935: 980.9 mill. Ls in Valsts statistiskā pārvalde (1936b) and 896.1 mill. Ls in later publications (Reichskommmissar für das Ostland 1942). However, only the first publication provides a complete breakdown of national income by industries. Valsts statistiskā pārvalde (1936b) published the 1935 estimate for 11 industries. But for 1933–1934 some of them are merged, and the same applies to all publications of national income 1936–1938 (see Table 2 in main text), where some industries are merged into the residual “other”. The same applies to the Reichskommmissar für das Ostland (1942), which provides the longest (1933–1938) national income series, but only with shallow and somewhat idiosyncratic breakdown by five industries, i.e. primary, manufacturing, trade and finance, transport, and others.

The bulk of the archives of the national statistical office of interwar Latvia (Valsts statistiskā pārvalde) did not survive. However, searching in the surviving part in the Latvian State Historical Archive (Latvijas Valsts vēstures arhīvs), we traced national income by 11 industries for 1934–1938. These are more detailed than those by the Reichskommmissar für das Ostland (1942, 21) and similar to the publication of the Latvian statistical office for 1935 (Valsts statistiskā pārvalde 1936b: 276–277). We report our findings in Table 37, claiming that it contains the first complete publication of national income by the Latvian national statistical office for 1933–1938 by 11 industries.

This table contains the final version of the national statistical office’s income estimate for 1935, while 980 mil. Ls total figure in Valsts statistiskā pārvalde (1936b) should be considered as a preliminary estimate (see Table 42 below). The difference is basically explained by the larger preliminary estimates of manufacturing and crafts (239.0 mil. Ls), and trade (128.1 mil. Ls) compared to our source (183.6 and 104.9 mil. Ls correspondingly, see Table 41). The series suggests that a full set of data was not accomplished before 1939 or 1940. In consequence, we conclude that the figures for 1933–1935 in Table 41 contain the final version of the estimates also for other years. The coincidence of our figures for 1935 and those from the Reichskommmissar für das Ostland (1942) provides further support for this argument.¹⁹

Table 41 National income of Latvia 1933–1938 in mil. Ls. Source: Latvian State Historical Archive (1308a, 1308b, inv. 11, file 18,987, 16)

The first publication of Latvia’s national income in Valsts statistiskā pārvalde 1936b contains no methodological information. However, we have traced another document providing some information on how Latvian statisticians did their calculations of national income for 1933–35 (Latvian State Historical Archive 1308a, 1308b, inv. 43, file 4). They defined national income as “the total value of wealth and services, which a national economy can use during a year for consumption or creation of capital without decreasing the preceding wealth condition” (Latvian State Historical Archive 1308a, 1308b, inv. 43, file 4, p. 4). We find they used market prices without subtracting taxes less subsidies. However, they did their best to subtract consumption of fixed capital (“amortisation”) from gross added value. They also subtracted dividends paid to foreign investors and interest to foreign creditors. Thus, it corresponds to net national income (NNI) at market prices according the SNA 2008 terminology.

To compare our estimates of Latvian total output in 1935 with the interwar time estimates in detail (see Table 42), we merge 20 ISIC, rev 4. sectors into 11 units, roughly corresponding to those used by Latvian statisticians. Our estimates of output of agriculture are only slightly higher than both interwar time estimates. There are no noticeable differences between our estimates for communications and transport as well as the public sector. Our estimate for industry and crafts coincides with preliminary estimate of interwar national statistical office but is larger in comparison with its final downwardly revised version.

Table 42 Comparison of the estimates of Latvia’s national income and GDP in 1935. Sources: Valsts statistiskā pārvalde (1936b, 276–277), Tables 4 and 41

To arrive at our estimate of Latvia’s NNI at market prices, we reduce our figures of GDP at market prices (1139.63 mil. Ls) by 1/12 (or 8.33%). This was a rough fix to calculate CFC following Klimantas and Zirgulis (2020) in situations when data on the CFC were missing. Then, from the intermediate figure of 1044.66 mil. Ls we subtract 7.28 mill. Ls as negative saldo of payments to foreign creditors, investors and remittances by foreign workers. Final Fig. (1037.38 mil. Ls) is not far away from the initial semi-official estimate (980.9 mil. Ls) published in Valsts statistiskā pārvalde (1936b), which was then downwardly revised (to 896.1 mil. Ls).

Appendix 4: Sensitivity analysis

We conduct sensitivity analysis for the GDP at factor costs by making changes for GVA of different sectors. Table 43 contains the list of possible sources of bias and which sectors they affect. Also, we indicate the direction of the bias and we assume that the possible size of the bias is 10% for each sector, except for inaccurate estimation of value added for the small industrial companies and incomplete dataset on services, since it affects quite small parts of GDP. Depending on direction, we either increase, or decrease the GVA of the sectors mentioned in the source of bias by the size. Then, we compute new GDP at factor cost and estimate the size of bias impact.

Table 43 Sensitivity analysis by source of bias

For example, the bias (2) affects sectors B, C, D, E, F in both directions. Thus, the GVA values (1.92 (B), 199.13 (C), 13.85 (D), 1.62 (E) and 23.59 (F) mil. Lats) are increased and decreased by 5%. In such a case, the reduced GDP at factor costs is 1042.99 mil. Lats and increased GDP at factor costs is 1080.71 mil. Lats. Comparing with the computed GDP at factor cost (1061.85 mil. Lats), we obtain the possible bias of 1.78%.

The sensitivity analysis (see Table 43) shows that our estimate of GDP at basic prices is most sensitive to possible bias in the composition of the sample of farms participating in the regular survey of agricultural production and costs, conducted by Latvian statistical office. The sample was small (N = 110 in the 1935–1936 agricultural year). It can be surmised that farms which were able regularly submit their complete accounts were husbanded by more efficient (in comparison with the mean) farmers, and so there is risk of upward bias estimating the output of whole agricultural sector from this sample. The estimations of the GVA of sectors from the “trade in broad meaning” are based on assumptions about compensation of employees, net operating surplus and consumption of fixed capital. These assumptions affect many sectors, and final GDP values reacts quite sensitively to these assumptions. Next, estimation of value added for small industrial companies on the basis of big companies might affect the GDP at basic prices about 1.7% to both directions. Similar effect might have missing information on various small services. Other effects like estimation of rural imputed rent or compensation of employees in public sectors have negligible impact on GDP values.

Appendix 5: The list of abbreviations

Abbreviation	Meaning
1990 GK$	1990 Geary Khamis International Dollar
A	ISIC rev 4 industry agriculture, forestry and fishing
B	ISIC rev 4 industry mining and quarrying
C	ISIC rev 4 industry manufacturing
CE	Compensation of employees
CFC	Consumption of fixed capital
D	ISIC rev 4 industry Electricity, gas, steam and air conditioning supply
E	ISIC rev 4 industry water supply; sewerage, waste management and remediation activities
F	ISIC rev 4 industry construction
FDI	Foreign direct investments
G	ISIC rev 4 industry wholesale and retail trade; repair of motor vehicles and motorcycles
GDP	Gross domestic product
GDPpc	Gross domestic product per capita
GK$ 1990	Geary Khamis 1990 international dollars
GOS	Gross operating surplus
GVA	Gross value added
GVO	Gross value of output
H	ISIC rev 4 industry transportation and storage
I	ISIC rev 4 industry accommodation and food service activities
IC	Intermediate consumption
ISIC 2008 or ISIC rev 4	The International Standard Industrial Classification of All Economic Activities, revision 4 (adopted in 2008)
J	ISIC rev 4 industry information and communication
K	ISIC rev 4 industry financial and insurance activities
L	ISIC rev 4 industry real estate activities
Ls	Lats (currency of Latvia in 1922–1940 and 1993–2014)
Lt	Litas (currency of Lithuania in 1922–1940 and 1993–2015)
M	ISIC rev 4 industry professional, scientific and technical activities
MPD	Maddison project database
N	ISIC rev 4 industry administrative and support service activities
NOS	Net operating surplus
O	ISIC rev 4 industry public administration and defence; compulsory social security
P	ISIC rev 4 industry education
PPP	Purchasing power parity
Q	ISIC rev 4 industry human health and social work activities
R	ISIC rev 4 industry arts, entertainment and recreation
S	ISIC rev 4 industry other service activities
S	Subsidies on products
s	Subsidies on production
SEK	Swedish krona (currency of Sweden)
SNA 2008	System of national accounts, 2008 version
T	ISIC rev 4 activities of households as employers; undifferentiated goods- and services-producing activities of households for own use
T	Taxes on products
t	Taxes on production
U	ISIC rev 4 activities of extraterritorial organizations and bodies
α	Latvian national statistical office: retail and wholesale trade in goods
β	Latvian national statistical office: trade support enterprises
γ	Latvian national statistical office: accommodation and food service activities
δ	Latvian national statistical office: communication and transport
ε	Latvian national statistical office: sports, arts and education
ζ	Latvian national statistical office: health care and hygiene
η	Latvian national statistical office: insurance and credit