Boom, Bust, Recovery Forensics of the Latvia Crisis

Olivier Blanchard, IMF Mark Griffiths, IMF Bertrand Gruss, IMF

Final Conference Draft to be presented at the Fall 2013 Brookings Panel on Economic Activity September 19-20, 2013

Boom, Bust, Recovery Forensics of the Latvia crisis.

Olivier Blanchard, Mark Griffiths, Bertrand Gruss 1 September 8, 2013 Latvia is a small country, with a population of only 2 million. Yet, it has been an object of intense attention during the crisis. Against the recommendations of many economists, it decided to maintain its peg and adjust through internal devaluation and front loaded fiscal austerity. Five years later, its proponents see it as a clear success. Its opponents see it if not as a clear failure, at least as an excessively painful adjustment. Witness the recent exchange between Latvias prime minister, Valdis Dombrovskis, and Paul Krugman: "Krugman famously said back in December 2008 that Latvia is the new Argentina, it will inevitably go bankrupt, and now he has difficulty apparently admitting he was wrong and so he tries to seek some problems in how Latvia is recovering from the economic crisis [] But I think that the mere fact that for the last two years we are enjoying rapid growth shows that it was probably the right strategy." [Dombrovskis, CNBC interview, 2013] "The adulation over Latvia really tells us more about what the European policy elite wants to believe than it does either about the realities of Latvian experience or the fundamentals of macroeconomics. [] We're looking at a Depression-level slump, and 5 years later only a partial bounce back; unemployment is down but still very high, and the decline has a lot to do with emigration. It's not what you'd call a triumphant success story," [Krugman Blog, January 2, 2013] The goal of this paper is to lay down the---often misstated---facts and some lessons of what is, on its own, a fascinating story of boom, bust, and (at least partial) recovery. The basic and striking facts to be explained are given in Figure 1: An increase in GDP of almost 90 percent from 2000:1 to 2007:4, followed by a decrease of 25% from 2007:4 to 2009:3, and a recovery, as of 2013:1, of 18 percent. A mirror image in terms of unemployment, with a decrease in the unemployment rate from 14% in 2000:1 to 6% in
1

All IMF. Prepared for the Brookings Panel, September 2013. We thank Rudolfs Bems, Martins Kazaks, Alex Klemm, Ayhan Kose, Christoph Rosenberg, for suggestions and comments, and Andreas Schaab for research assistance.

2 2007:4, followed by an increase to more than 21% in 2010:1, and a decrease since then, down to 11.4% in 2013:2.
Figure 1. Latvia: Real GDP and the Unemployment Rate (sa), 2000-13 200 180 24 20 120 16 12 8 4 2000 2002 2004 2006 2008 2010 2012 100 80 160 140

Unemployment (lhs)

Real GDP (2000q1=100, rhs)

We focus on six aspects of the story: 2 1. What triggered the boom? Our answer: EU accession and belief in convergence to EU per capita incomes, cheap funds from foreign-owned banks, and optimistic expectations. The boom was healthy at the start, but, like many booms, increasingly bubbly and unbalanced at the end. 2. What ended the boom? The end came in two steps. First, starting in 2007, a slowdown, due to rising inflation and loss of competitiveness, and tightening credit reflecting increasing worries by banks about their loan book. Then, at the end of 2008, a collapse due to the world financial crisis, leading to a sudden stop, a credit

A lot has been written on Latvia. Two very relevant references, which look at the various aspects of the

boom and the bust, and from which we have benefited, are by Aslund and Dombrovskis (2011) and the set of articles in EU (2012).

3 crunch, a sharp drop in exports and increased uncertainty. Fiscal austerity came, for the most part, later. 3. What role did the sudden stop play in explaining the sharp decline in output? Liquidity provision by both foreign banks and by the central bank and the Treasury reduced but did not eliminate the credit crunch. The decline in output was larger however than what one would expect a credit crunch to trigger. Uncertainty, after the sudden stop, and the option value of waiting, probably contributed too. 4. What was the role of fiscal consolidation? Ironically, it is hard to blame fiscal austerity for the decrease in output. Much of the fiscal adjustment was implemented after the main fall in output. There is suggestive evidence that commitment to a clear adjustment programbacked by substantial international financial support increased confidence, as reflected in lower CDS spreads. Much of the decrease in borrowing rates for households and firms was associated however with the increased credibility of the peg and the decrease in exchange rate risk. This may have been in part the indirect result of fiscal consolidation through market perception that consolidation would make the disbursement of international support more likely. 5. How did the internal devaluation work? It worked, but in ways different from the textbook adjustment. Public wages decreased sharply, but with limited effects on private wages. Much of the improvement in unit labor costs, especially in the tradable sector, came from increases in productivity. This improvement in unit labor costs was only partly transmitted to prices, leading more to an increase in margins for firms. This improvement was followed by an increase in exports, and in turn by an increase in internal demand. On the supply side, part of the adjustment has happened through emigration, in a way very similar to what happens across U.S. states. 6. Is output back to potential output and unemployment back to the natural rate? Not yet, but they may not be very far. There is no evidence that the natural rate is any higher than it was before the crisis. But the evidence is also that, given the market friendly labor market institutions, it was surprisingly high before the crisis. The difficulty is to pin down what exactly it was before the crisis. Having laid out the facts, we return to the central issue. Is the Latvian adjustment a success story? Here is the preview: In some ways, it clearly is a success. From a macroeconomic viewpoint, the Latvian economy is nearly back to where it was before the boom became unhealthy. The unemployment rate is still somewhat higher than the (too high) natural rate. Output is growing. The financial system is safer. Few other European countries can claim the same.

4 Nevertheless, the adjustment involved a very large decrease in output, a very large increase in unemployment, and substantial emigration. The question is whether an alternative strategy could have achieved a better outcome. Nobody can give a definitive answer. What can be said is that the sharp fall in output was not primarily due to the adjustment policies. That fiscal consolidation was associated with higher credibility, and did not prevent the recovery. And that the internal devaluation worked surprisingly quickly.

On the Boom Some Background After being occupied by the Soviet Union following World War II, Latvia became independent again in 1991. It is a small, middle income, country. It has been a member of the European Union since 2004. In 2012, its nominal GDP was 22 billion euros (the national currency, the lat, is pegged at 1.42 euros). Its population was 2 million. Its PPP GDP per capita was 62% of the EU28 average. It is a very open economy. In 2012, trade openness, measured by the ratio of exports to GDP was 60%, with about 30% of exports being reexports. The lat has been pegged to the euro since January 1, 2005 (and to the SDR before that), and is scheduled to join the Euro next January. Financial openness is also high. In 2012, the ratio of gross foreign liabilities stood at 135%, of which about half were bank liabilities, and there are no capital controls. After the initial drop in output due to the transition in the early 1990s, and a brief interruption due to the Russian crisis in the late 1990s, Latvia experienced very high growth until the crisis, 7.7% annually from 1996 to 2007. Based on the Penn tables, PPP GDP per capita in 2005 dollars increased from 5,500 dollars in 1993 to a peak of 14,800 dollars in 2007, just before the crisis. In general, Latvia has adopted pro market institutions. In the 2013 Ease of doing business survey conducted by the World Bank, Latvia ranked 25th out of 183 countries. The Boom Given the low income per capita in the early 1990s, relatively good institutions, and the proximity to Western Europe, the potential for catch up was clearly high. The question is whether high growth reflected healthy catch-up growth or more. Before looking at the specifics, it is useful here to get a benchmark. Using two variants of the Barro growth specification (Schadler et al 2006, Vamvakidis 2008), which include not

5 only initial income real per capita, but also population growth, partner country growth, and a number of institutional variables, suggests the following answers: Until 2000, average growth of PPP GDP, 6.5%, was close to what these panel regressions predict. From 2001 to 2004 however, average growth of 9.0% was 2-3% higher than predicted, and from 20052007, average growth of 11.6% was 4-6% higher than predicted. 3 This suggests that, starting in the 2000s, and especially from 2005 on, GDP growth in Latvia was higher than can be explained by catch-up, that the boom became increasingly cyclical, and, as we shall see, unhealthy. With these results in mind, we shall start our story in 2000, and refer to the period 2000-2007 as the boom. During that period, average annual growth was 8.8%. The unemployment rate decreased from 14% to 6%. Despite the peg, inflation increased to 10.1% in 2007 and 15.3% in 2008, the highest in the EU. Looking at it from the demand side, the boom came primarily from an increase in domestic demand. The ratio of private consumption to GDP (in constant prices) increased from 62% to 72%, the ratio of investment to GDP (also in constant prices) from 22% to 36%. 4 Investment reflected in part a housing boom, with housing investment increasing from 2% to 5% of GDP, and with 40% of the increase in employment during the period taking place in construction (of which housing accounts for roughly 1/3) and real estate. There were indeed good reasons for Latvian households and firms to increase consumption and investment: prospects of catch-up growth, prospects of joining the EU, and later joining the euro, a goal delayed by the crisis but which will now take place in 2014. As a matter of arithmetic, the result of increasing consumption and investment ratios was a steady deterioration of the current account balance, with the ratio of the current account deficit to GDP increasing from 5% of GDP in 2000 to peak at a very large 25% in mid-2007. Decomposing between exports and imports, the ratio of exports to GDP (in constant prices) remained roughly constant, at 45%, which is impressive given the high growth of the denominator---GDP growth. On the other hand, the ratio of imports to GDP (also in constant prices) increased from 51% to 71%. If we think of imports as depending on domestic demand rather than on GDP, and given that domestic demand increased much faster than GDP, a more relevant statistic is the ratio of imports to domestic demand; this ratio went up from 49% to 58%. Thus, not only was domestic demand very strong, but it was accompanied
3

See web annex for details of the two specifications.

In current prices, the ratio of consumption to GDP remained nearly constant, while the ratio of investment to GDP increased by 23%, reflecting a large increase in the relative price of investment goods.

6 by a large shift towards foreign goods, leading to an even larger deterioration of the current account balance. (Figure 2 plots the evolution of exports, imports, and the trade balance). Part of the shift probably reflected the increasing real exchange rate appreciation (more on this below). Part of it may have also reflected a shift towards higher quality foreign products, an issue relevant when we look at the current account adjustment later. 5
Figure 2. Latvia: Trade Balance, 2000-13 (national accounts basis, 4 quarter moving sum, percent of GDP) 70 65 60 55 50 45 0 -5 -10 -15 -20 -25 2000 2002 2004 2006 2008 2010 2012 40

Trade Balance

Exports (rhs)

Imports (rhs)

This current account deficit was easily financed, but with a worsening in the composition of financing over time: FDI increased from around 5% of GDP in 2000 to a peak of about 8% in mid 2007, before tailing off, perhaps an indication of worries about the persistence of the boom. Even by 2007 however, the stock of FDI remained relatively low, 20% of GDP, versus for example 40% in Estonia. The rest of the financing was mostly provided by Swedish and other Nordic parent banks to their Latvian subsidiaries. Banks liabilities to foreign banks rose from 30% of GDP in 2000 to almost 90 percent in 2007.
5

See Bems and Di Giovanni (2013)

Throughout, interest rates were low. Figure 3 plots a number of interest rates from 2000 on. The 3-month money market rate decreased in the early 2000s, as Latvia repegged from the SDR to the euro, remaining around 4% until early 2007. Mortgage rates capture well the evolution of borrowing rates in general. Mortgage rates in euros remained below 6% until 2007. Rates in lats increased from 2007 on, reaching about 14% by the end of 2007, with the premium reflecting increasing perceived exchange rate risk. But, given increasing inflation, even real mortgage rates in lats (using actual CPI inflation) were roughly equal to zero from 2004 on, and real mortgage rates in euros were increasingly negative. Using wages or house price inflation, real interest rates were even more negative.
Figure 3. Latvia: Nominal Interest Rates, 2004-13 24 24

20

20

16

16

12

12

0 2004 2006 2008 2010 2012

3 month money market

Mortgage rate - lats

Mortgage rate - euro

Associated with foreign financing through banks was very high bank credit growth, increasing by end-2007 to almost ten times its 2001 level, an annual average growth rate of 33% in real terms. While the ratio of private sector credit to GDP was less than 20% in 2000, it reached almost 90% in 2007, higher than in other emerging European economies, although still lower than the Euro average. The proportion of loans denominated in foreign currency (initially mostly dollars, but by the end almost entirely euros) also increased steadily, from 50% in 2001 to more than 85% in 2007. There were few signs of overheating until 2006, consistent with the notion that high growth until then reflected mostly potential growth. Starting in 2006 however, signs became much clearer. Wages and prices of domestic goods increased rapidly. As shown in Figure 4, unit

8 labor costs (ULCs), normalized to 100 in 2000, reached 135 at the end of 2005, 164 at the end of 2006, 211 at the end of 2007, and 245 at the end of 2008. 6 The GDP deflator increased, although by less, to reach around 200 at the end of 2008 (with the smaller increase in the price deflator than in the ULC implying a substantial increase in the labor share, a point to which we shall come back later). The CPI also increased, although by less than the GDP deflator, reflecting the large share and the stable price of imported goods in the consumption basket. The price per square meter of an apartment in Riga, a good index of housing prices, increased from 400 euros in early 2004 to 1700 in early 2007. 7 Thus high inflation, increasing overvaluation, and exploding housing prices all pointed to an increasingly unhealthy boom.
Figure 4. Latvia: Price Deflators and Unit Labour Cost (2000=100) 260 240 220 200 180 160 140 120 100 80 2000 2002 2004 CPI 2006 2008 2010 ULC 2012 260 240 220 200 180 160 140 120 100 80

GDP deflator

Through 2007, the policy response was however limited.

Along the paper the reported ULC for the whole economy or for individual sectors are constructed as the ratio of aggregate nominal compensation of employees over real gross value added (GVA) in the respective category. This ensures consistency when comparing ULC for the whole economy with ULC for individual sectors, as GDP is not available at the sector level. Eurostat official data on house prices starts only in 2006, but even then from 2006:1 until the peak in 2008:1 shows a 75% increase.

9 Although there were some increases in reserve requirements, monetary policy was run as a quasi currency board, with the implication that the refinancing rate of the Bank of Latvia followed very closely the low ECB lending rate. A small fiscal headline deficit turned into a small headline surplus in 2007. Was it the appropriate fiscal stance? This is where hindsight comes heavily into play. As of 2007, output was assessed to be only slightly above potential, the output gap perceived to be only slightly positive. In hindsight, it has become clear that the output gap was in fact larger, and thus the ``cyclically adjusted fiscal balance was much worse. The point is made in Figure 5, which plots the cyclically adjusted balance for each year from 2006 to 2009, using the output gap for that year, calculated by the European Commission as of different dates. 8 For 2004 for example, the adjusted balance is roughly similar across calculation vintages, reflecting the fact that, then and as of now, the output gap was perceived as small. For 2007 however, the adjusted balance, which was then perceived to be a small surplus, is now estimated to have been a deficit of close to 3%. This reflects the current perception that what was seen then as a small positive output gap (3% according to the European commission) is now estimated to have been a much larger one (12% as of 2013)
Figure 5 - Real-time Cyclically Adjusted Fiscal Balances (percent of GDP, using real-time output gap estimates and elasticity of 0.3) 1 0 -1 -2 -3 -4 -5 2004 2005 2006 2007
2006 2007 2008 2009 Latest (2013) Headline Fiscal Balance

See web annex for details. The cyclically adjusted fiscal balance is constructed as the headline balance minus 0.3 times the output gap, following a method proposed by the OECD and estimates of elasticity for Latvia in European Commission (2005). The different vintages of output gap series are those estimated by European Commission (Autumn Economic Forecast).

10

In March 2007, the authorities introduced an ``anti inflation plan. The main measures were balanced budget targets in 2007-08, and budget surpluses for 2009-10, the introduction of a capital gains tax for real estate, and attempts to restrain bank lending (making loans exclusively on clients legal incomes (as opposed to stated income), mandatory 10 percent first installment repayment, and fixing a maximum loan to value ratio). It was however too little too late. In short, the anticipation of a large scope for catch up growth, together with cheap external financing, led to an initially healthy boom. As time passed, the boom turned unhealthy, with overheating leading to appreciation and large current account deficits, with lower credit quality, and with balance sheet risks associated with FX borrowing. It is no great surprise that the government was reluctant to acknowledge the changing nature of the boom, and thus was unwilling to dramatically slow it down. Even as late as mid-2006, government officials argued that macroeconomic developments were largely benign: rapid growth was essential for income convergence, inflation was due more to wage and price convergence than demand factors, increased infrastructure investment would prevent growth bottlenecks and enhance competition. The financial regulator saw potential for further credit growth, arguing that household debt ratios were low and the strong and liquid housing market provided adequate loan collateral; it saw its responsibility as one of ensuring that individual banks had sufficient capital, rather than having a macroprudential role. In contrast, the Bank of Latvia was more concerned about overheating, and the risks from high debt levels and large currency and real estate exposures. Though it supported fiscal tightening, aside from raising reserve requirements the Bank of Latvia lacked the instruments to respond, given the quasi-currency board and the open capital account. An interesting question, with obvious implications beyond Latvia, is whether outside observers, with no obvious political stake, sounded the alarm bell more strongly. The answer, at least referring to the IMF, is a qualified yes. In its 2005 annual review (the so called ``Article IV annual review), the IMF pointed out problems arising from rapid credit growth, domestic banks increasing reliance on non-resident deposits for funding, and saw the authorities decision to remove limits on banks open positions in euros (because of the peg and the goal of euro adoption) as premature. In 2006, it renewed its warnings, recommending stronger fiscal tightening, and macro prudential measures to limit credit. In its concluding statement for the 2007 Article IV mission, the warnings were even more explicit. The record of credit agencies was definitely more mixed. Though they lowered their outlooks earlier, ratings agencies were slow to react with ratings downgrades. For example, while

11 Moodys recognized that rising inflation and current account deficits posed risks, it pointed to the governments low debt ratio and stable external funding of the financial system (longterm loans from parent banks plus non-resident deposits which it viewed as stable) as mitigating factors. S&P kept its A- rating until May 2007, and BBB+ until October 2008. Fitch kept an A- rating until August 2007, BBB+ until October 2008. Moodys kept its A2 rating until November 2008. S&P and Fitch only lowered their ratings to BB+ (below investment grade) in February and April 2009; Moodys kept its investment grade rating throughout.

On the Bust. Part I. Fatigue, and the Global Financial Crisis

The end of the boom happened in two distinct phases: First, a slowdown before the global financial crisis. Then, a collapse due to the impact of the global financial crisis, through a sudden stop of capital inflows, a credit crunch, and a sharp drop in exports.

Fatigue and the Slowdown Booms sometimes die a natural death. The stock adjustment process that initially increased investment and durable consumption comes to an end. Increasing overvaluation reduces exports. Expectations of sustained fast growth turn out to be optimistic and are revised downwards, leading to lower domestic demand. Credit quality deteriorates, leading banks to eventually tighten credit. The first signs of such fatigue appeared in Latvia in 2006. Consumer confidence peaked in 2006:3, business confidence in 2007:1. Starting in February 2007, worries about the exchange rate peg led to a large jump in the Rigibor, the Latvian interbank rate, relative to its Euribor counterpart, with the spread increasing from 0.5% to 6% within two months. By early 2007 on, credit standards were tightened, with subsidiaries of Swedish banks in the lead. Borrowing rates increased from around 7% at the start of the year to peak at around 15% in November ---although, as we saw earlier, real interest rates declined, because of the sharp increase in inflation. Output peaked in 2007:4. Had there been no global financial crisis, Latvia might have gone through a slump similar in nature (but probably worse in size, given the imbalances) to what happened in Portugal in the early 2000s: Weak foreign demand because of the overvaluation triggered by the earlier boom, and weak domestic demand, due in part to tighter credit. But, starting in 2008, this adjustment process was overtaken by the effects of the world crisis.

12

The Global Financial Crisis and the Bust

As in other emerging markets, the crisis affected Latvia through two main channels, trade and financial. 9 It is useful to look at the evolution of the different components of GDP during the bust. Figure 6 shows the evolution of each component of GDP from the peak in 2007:4 to the trough in 2009:3, as a percentage of 2007:4 GDP.
Figure 6. Evolution of GDP Components During the Crisis, 2007:4-2009:3 (percent of 2007q4 real GDP) 30 30

20

20

10

10

-10

-10

-20

-20

-30 2007q4 GDP 2008q2 Consumption 2008q4 Investment 2009q2 Exports

-30

Imports

Over those eight quarters, GDP declined by 25%. Foreign demand (X) accounted for 8% of the decrease. But, much more dramatic was the decrease in domestic demand (C+I+G), which declined by 43% of GDP! Fixed investment itself fell by more than half. This decrease in domestic demand was partly offset, in terms of its effect on the demand for domestic goods, and, by implication, its effect on GDP, by a decrease in imports (M) of 26%

A first pass at the effects of the crisis on emerging market economies in general was presented in an earlier Brookings paper (Blanchard et al, 2010)

13 of GDP. 10 These numbers have a clear implication: The bust was due in part to a decrease in foreign demand, but much more so to a collapse in domestic demand. Looking more closely, we start with foreign demand. Exports started declining in 2008:1, so while world demand was still increasing, before the global crisis. This decrease was probably due to the increasing overvaluation we saw earlier. But the major decline took place during 2008:4 and 2009:1. During those two quarters, exports decreased by 8% of GDP, clearly due to the global crisis. Dynamic simulations, using estimated an export equation specifying log exports as a function of the log of the partner countries GDP (using trade weights) and the real exchange rate, suggest that the adjustment was faster than usual, but by 2009:2, roughly in line with what would have been predicted. 11 How much of the decrease in output can be explained by the decrease in foreign demand? About 30% of exports are re-exports. So a decrease in exports of 8% of GDP implies a decrease in net external demand of just over 5%. In our earlier BPEA paper on emerging markets during the crisis, we found that an (unexpected) decrease in exports of 1% of GDP led to a 1.5% (unexpected) decrease in GDP, so, in this case, 5x1.5= 7.5%. Given the size of the decrease in domestic demand (43%), it is clear that the dominant factors have to be found elsewhere, namely with the credit crunch and the sudden stop, starting in 2008: Fiscal policy did not play much role until the middle of 2009. The sudden stop and credit crunch, and the role of fiscal policy are the topics of the next two sections.

On the Bust. Part II The Sudden Stop and the Credit Crunch.
The relative simplicity of the Latvian financial system makes it easier to trace the effects of the sudden stop. Some background is in order: In 2008, Latvian subsidiaries of Nordic banks accounted for 60% of total bank assets. The largest domestic bank, Parex, accounted for another 14%. Other domestic banks accounted for the remaining 26%. The banks had different business models: On the liability side, Nordic subsidiaries were financed 1/3 by domestic deposits,
The decrease in imports seems large as a proportion of GDP, but the relevant denominator is domestic demand. The decrease in imports is equal to 60% of the decrease in domestic demand, roughly equal to the ratio of imports to domestic demand in 2008:1. See web annex. The estimated elasticity with respect to partner country GDP is 1.5, the estimated elasticity with respect to the exchange rate is -0.27.
11 10

14 2/3 by their parent banks. Parex, in contrast, was financed, in roughly equal proportions, by resident deposits and nonresident deposits---mostly from CIS countries. 12 Thus, funding for Nordics depended very much on the decisions of their parent banks, funding for Parex on the behavior of nonresident depositors and lenders from abroad. As we saw earlier, credit growth had slowed down already before the global crisis. New loans peaked in 2006:4 (so before the decrease in output), and real credit growth slowed from 12% quarter on quarter in 2006:4 (56 percent annualized) to being virtually flat one year later. But the trigger for the financial crisis was a run on Parex, in the wake of the Lehman collapse. Parex was exposed to high rollover risk. In addition to funding by nonresident deposits, large syndicated loans, amounting to 16% of its end-2008 liabilities, were coming due in early 2009, and a Euro bond, accounting for another 4% of its liabilities, was potentially callable, for example in the event of default on syndicated loans. In contrast to Nordic subsidiaries, Parex had no parent bank, thus no deep pockets. And, under a currency board, there was, at least in principle, no room for liquidity provision by the central bank. A bank walk had started in late July. Then, starting in early October (a few weeks after Lehman), the walk turned into a run. By the end of 2008, total deposits in Parex were down by 34% relative to June. Various measures were taken by the Latvian supervisory authority, and by the government. A partial public takeover in November was followed by full nationalization later in the month. These however did not stop the run. Restrictions on deposit withdrawals were imposed in early December. As nonresidents closed their accounts and sold lats, the central bank defended the peg. Reserves (ignoring valuation effects from depreciation of the euro and thus the lat) fell. In the last three months of 2008, the Bank of Latvia sold 1.15 billion, or roughly one quarter of its end-September reserves. Under a strict currency board, there would have been no further central bank intervention, and would have implied a decrease in the monetary base equal to the decrease in reserves. There was however strong pressure to provide funds to Parex. This was eventually done, but indirectly: The government placed treasury bills and increased Treasury deposits at Parex (to 1/3 of total deposits by end-2008), which in turn used the bills to obtain financing from the central bank to fund deposit outflows. In turn, the international donorsat first a swap line from the Swedish and Danish central banks, as a bridge to an IMF/EU/Nordic programreplenished the reserves of the central bank.
12

Non resident deposits in Latvia were mostly by CIS corporations, for ease of transaction, geographical proximity, and language.

15

Fortunately, Nordic parent banks absorbed losses by recapitalizing their subsidiaries and committed not to cut funding to their subsidiaries, both implicitly at the start of the program and more formally later on. Despite this, loans from Latvian subsidiaries to residents still went down from a peak of 10.5b lats in 2008:4 to 8.1b lats at the end of 2011, but it was a smooth decline, and it is likely that much of it was explained by lower credit demand rather than just tighter credit supply. 13 Thus, in the end, continued funding by Nordic parent banks limited the size of the sudden stop (or at least made it less sudden), and liquidity provision by the Treasury and the central bank limited the size of the credit crunch. Still, in the year to June 2009, the monetary base decreased by one third, or about 5% of GDP. This monetary tightening (linked to another round of devaluation rumors and thus speculative attack) and higher perceived counterparty risk were reflected in an increase in the 3-month money market rate, which went from 6.3% in September to 21% in June 2009. Borrowing rates for households and firms moved in the same way, with much evidence of strong credit rationing. Interestingly, rates on loans in euros increased by much less, suggesting that investors were more worried about the risk of euro depegging rather than about credit risk. How large was the effect of the credit crunch on activity? There are now a number of studies looking at the effect of such credit supply shifts on output, both in normal times and in crisis times. The studies looking at crisis times give a range for the one-year response of output growth to credit growth of 0.3 to 1.1. 14 15 This suggests the following back of the envelope computation: Loan growth from 2008:3 (when loans peaked) to 2009:3, in real terms, was around minus 5%, compared to loan growth over the 4 quarters to 2008:3 of around 3% (already down from close to 50% eighteen months earlier). The parameters above suggest that this decrease in loan growth of 8% may explain a decrease in domestic demand growth between 3% and 9%. 16 This is large, but is still substantially less than the decrease in
Latvian subsidiaries of Nordic banks include Swedbank, SEB, DNB and Nordea (which operates as a branch rather than as a subsidiary). Calomiris and Masson (2003) focus on the Great Depression, Peek and Rosengreen (2000) on Japan, Greenlaw et al (2008), Duchin et al (2010) on this crisis, Claessens et al (2009) on a number of crunches. It is not clear whether the right parameter one should look at is the response of output growth to credit growth, or instead the response of output growth to the change in the credit to GDP ratio. The studies results we mention are stated in terms of the first. It is not clear that credit growth is the right metric to relate to demand growth. As one might expect, the decrease in new loans was much larger. After peaking in 2006:4, the flow of new loans[(constructed from the change in the credit stock and an estimated amortization flow based on the maturity structure of bank loans) had decreased by 65% in real terms by end-2008 and by 85% by end-2009. A related metric, constructed as the
16 15 14 13

(continued)

16 domestic demand of 27% over the same period. So what else can explain the collapse in demand?

Figure 7. New Car Registrations (2007=100) 140 120 100 80 60 40 20 0 2005 2007 2009 2011 2013 140 120 100 80 60 40 20 0

By elimination, the answer seems to be uncertainty, and the option value of waiting. Suggestive evidence is given by the behavior of car sales during the period. As shown in Figure 7, new car registrations, normalized to be 100 in 2007, collapsed to 18 in January 2009 and fell gradually during the year to reach one tenth of their 2007 levels by January 2010. 17 18 Some of the decrease came from credit rationing: Partly because of the uncertainty, partly because of legal uncertainty about the ability of banks to repossess the collateral, banks simply stopped offering car loans. But much of it clearly came from the high uncertainty facing consumers, be it about the peg, the soundness of the banking system, or the size of the ongoing recession.

difference in the flow of credit relative to GDP (see Biggs et al., 2009) would suggest that the contribution of credit to demand growth, or credit impulse, was -19% of GDP over 2008:3 to 2009:3. For comparison, in her study of the effects of uncertainty on spending at the onset of the Great Depression, Christina Romer (1990) found that car registrations declined by 24% from September 1929 to January 1930. Anecdotally, Latvia, which does not produce cars, became a car exporter for a few months in late 2009 and early 2010, as dealers, unable to sell their inventory of cars at home, sold them to foreign dealers.
18 17

17

On Adjustment Choices. Fiscal Consolidation.

In early 2009, after the large decrease in output, what was needed to return Latvia to health? Again, it is important to keep in mind the two sources of the crisis: The natural (or unnatural) end of a boom, and the later collapse due to the world financial crisis. The world financial crisis had led to a sudden stop and a credit crunch, a large drop in exports, and, as a result, a large decrease in output. Recovery of the world economy was needed to resolve the second. Financial cleanup was needed to repair the first. It was clear however that, even without the financial crisis, Latvia would have needed a large macro adjustment. Latvia had been operating above potential output before the crisis. This had led to a clear overvaluation, reflected in an unusually large current account deficit. And, while the fiscal deficit had remained small, the fiscal balance at anything like potential output required substantial consolidation. Many economists recommended a nominal devaluation (with some variations, for example, a onetime devaluation and then use of the ERM II bands (backed by the potential support of ECB intervention) or even accelerated adoption of the euro), and a steady but smooth fiscal adjustment. The argument in favor of devaluation was an old one, namely that external devaluations solve a coordination problem and are easier to achieve than internal devaluations: A nominal exchange rate adjustment automatically coordinates real price and wage adjustments. The argument in favor of a steady but smooth fiscal adjustment was that the increase in the deficit was mostly due to the crisis and would largely go away as output recovered, and that the ratio of net debt to GDP, while it would likely increase because of bank recapitalization costs and future budget deficits, at end-2008 was still very low at less than 15%. The Bank of Latvia and the Latvian government rejected this advice, decided to maintain the peg and to proceed with front loaded consolidation. It saw any devaluation as inconsistent with the goal of euro adoption, a goal delayed first by the boom (because of the induced inflation) and then by the crisis. It believed that only a strong fiscal consolidation would send the signal that the government was committed to fiscal sustainability. 19 The European Union, the European Central Bank, and the Nordics supported that strategy, all be it for their own reasons. While views within the IMF differed, it is fair to say that the IMF was more skeptical of the choice to maintain the peg, but went along with the overall strategy.
19

See Aslund and Dombrovkis (2011).

18

This section looks at the fiscal consolidation, the next at the adjustment under the peg---the so called internal devaluation. In 2008, the headline general government balance turned from a small surplus to a large deficit of 3.4% of GDP (excluding bank restructuring costs of around 4 percent of GDP), reflecting the large decrease in activity. Little fiscal consolidation took place until 2009---which is why we did not focus on fiscal policy when explaining the initial decline in output. A revised 2009 budget, passed in December 2008, included measures adding up to 7% of GDPalthough some estimates suggest only 4% was actually implemented. 20 In February 2009, the government fell, and, in March, a new government was put in place, with the challenge of implementing the fiscal consolidation agreed to by the previous government but which it had been unable to deliver. At the same time, the deepening recession was blowing wide open the deficit. In June, after local government and European elections, and following long discussions with social partners and with the involvement of the President, the new government announced a consolidation program, with new measures adding up to 3.4% of GDP in 2009 (with a full year effect of 6.5% of GDP). New measures included a further 20% cut in the government wage bill, controversial cuts to pensions (later ruled unconstitutional), and reductions in personal income tax allowances, which made the personal income tax less progressive. 21 Netting out expansionary measures (like pensions and benefits increases embedded in earlier versions of the 2009 budget, and the approval of additional spending of 1% of GDP in social safety nets, as part of the program) and the likely partial implementation of early measures, fiscal consolidation in 2009 is estimated at about 8% of GDP, of which only about 2%-3% of GDP happened in the first half. These measures and others introduced in the following budgets implied a further adjustment of 5.4% of GDP in 2010, and 2.3% in 2011. The evolution of headline deficits and of cyclically adjusted fiscal balances (the motivation for the use of quotes will be clear below), from 2008 on, is given in Figure 8. We plot three series, the change in the headline deficit, the change in the cyclically adjusted deficit including bank restructuring costs, and the change in the cyclically adjusted deficit, excluding bank restructuring costs. 22 Two aspects of the figure are particularly striking:
Although all the revenue measures (including an increase in the headline VAT rate from 18 to 21 percent, in the reduced VAT rate from 5 to 10 percent, and excise tax increases) adding to 2.5% of GDP were introduced, only about 1.5% of GDP in expenditure cuts are estimated to have been implemented.
21 20

Public sector wages had increased by roughly 20% more than private sector wages during the boom.

The cyclically adjusted fiscal balance in Figure 8 is defined as the headline balance-to-GDP ratio minus the output gap times an overall budgetary sensitivity parameter, based on European Commission (2005) estimates, of 0.3.

22

19

The first is the increase in the headline deficit in 2009; but given the very large decline in output, this corresponded to a reduction in the cyclically adjusted deficit.
Figure 8. Fiscal Impulse: Change in Headline and Cyclically Adjusted Balances (percent of GDP, using latest EC output gap estimate) 6 4 2 0 -2 -4 -6 -8 2008 2009 2010 2011 6 4 2 0 -2 -4 -6 -8

Headline (change, excl. bank restructuring costs) Cyclically Adjusted (change, excl. bank restructuring costs) Cyclically Adjusted (change, incl. bank restructuring costs)

The second is the small decrease in the cyclically adjusted deficit in 2009, computed excluding bank restructuring costs (which affected mostly the budget in 2008): 1.4% versus the bottom up 8% number given earlier, based on government measures taken in 2009. This points to the difficulties in measuring cyclically adjusted deficits under such conditions---and thus the extreme care that must be exercised in quantitative exercises. The problem in this case is not so much the measurement of the size of the output gap, which we discussed earlier; so long as measures of potential output move smoothly from year to year, changes in the fiscal position are not very much affected by potential output measurements. The problem comes from two sources. First, the elasticity of various budget items to activity: For example, the adjustment for the output gap can be misleading if some taxes depend on domestic demand, and domestic demand and output move very differently (as they did in 2009, with domestic demand contracting much more than output). Di Comite et al (2012) conclude that correcting for the right elasticities implies up to 3 to 4% of GDP in 2009 more consolidation than the 1.4% reported in the figure. Second, and going the other

20 way, measures which are neither cyclical nor explicit but which still affect the budget: In the case of Latvia for example, Di Comite et al. show that indexation of public wages to past inflation and non cyclical social benefits may have led to higher expenditures, subtracting 2% from the bottom up number reported above. In any case, the numbers imply a substantial fiscal consolidation, with much of the adjustment starting in the second half of 2009, and continuing in 2010 and 2011. Determining with any certainty its effects on output is impossible. Surely, it is unwise to argue, as some have done, that the return to growth from 2009:4 on was due to the expansionary effects of fiscal consolidation. Many other factors were at play. As we have seen, much of the earlier sharp decrease in domestic demand was probably due to uncertainty and the option of waiting. It is likely that, as uncertainty decreased, the economy would have recovered, independently of the path of fiscal policy. One of the channels through which fiscal consolidations can have limited adverse effects on activity, and even sometimes lead to an increase in output, is through decreases in interest rates. The announcement of a credible consolidation program may lead investors to decrease the risk premium, leading in turn to an increase in demand that may partly or fully offset the direct contractionary effects of consolidation. And it is indeed the case that, during that time, there was a dramatic decrease in interest rates. As was shown in Figure 3, the 3-month money market rate went from a high of 21% in June 2009 to 11% in September, down to 3% by February 2010. Can this decrease be attributed to fiscal consolidation? Two arguments suggest that the answer is probably not, or at least not directly. 5-year CDS spreads on Latvian public debt, which had increased from 800 at the start of 2009 to 1200 after the fall of the government, indeed decreased to 700 by the end of April, and the installation of the new government, and to 500 at the end of September, thus reflecting increasing confidence about fiscal sustainability. 23 But, there is little relation between the evolution of these spreads and the rates relevant to private borrowers, such as the 3-month rate or mortgage rates. The decrease in the 3-month rate was accounted for, nearly fully, by the decrease in the spread between lat rates and euro rates, suggesting a decrease in exchange rate rather than fiscal risk. And, indeed, there were good reasons for investors to believe that the peg would be maintained. Despite intense debates until June about the pros and cons of a devaluation, the government reiterated its commitment, and, by July, both the
23

One has to wonder about such large spreads for a country where the ratio of net debt to GDP was still around 10% of GDP (1200 basis points imply an annual payment of 12% of the nominal value of the debt.)

21 EU and IMF had agreed to disburse funds, removing one major source of uncertainty about the ability of the government and the central bank to keep the peg. In short, it may well be that a credible fiscal plan was part of what made the overall program credible, and, together with other measures, restored confidence in the peg and led to the decrease in interest rates. If so, the effect was indirect. What cannot be established is whether major front loading was needed for credibility. Whatever the case, the fact is that fiscal consolidation coincided with growth, from a very low starting point.

On Adjustment Choices. The Internal Devaluation

Before the global crisis, Latvia had been running very high current account deficits. This reflected in part output above potential and correspondingly high imports, in part increasing overvaluation. With the global crisis came a dramatic improvement in the current account. While exports further decreased, domestic demand collapsed, leading to a collapse in imports. During 2009, the current account actually swung into surplus. This was not completely good news however. It was likely that with the return of output to potential---whatever the precise value of potential output was, more on this later---the current account deficit would reappear. To maintain current account balance as growth came back, there was a need for a real depreciation. Estimates varied, but it was generally believed that at least a 20% real depreciation was needed. While many argued that a devaluation of the lat was the best solution, the authorities, seeing euro entry as a clear goal for the adjustment program, emphatically rejected this approach and decided to adjust through an internal devaluation, an adjustment of nominal prices and wages, rather than an adjustment of the nominal exchange rate. So how did it work? The evolution of unit labor costs. Figure 9 shows what happened to ULCs. It plots cumulative changes in ULCs and its two components, productivity and wages, for the economy as a whole, from 2008:3 on. 24 The
We use occupied posts, instead of number of people employed, to decompose ULC into productivity and wages. The reason is that the Labor Force Survey (LFS) data from 2011 onwards is not comparable with historical data: results of the 2011 population and housing census revealed that population in Latvia was significantly different from the one used for the extrapolation of LFS data, and 2000-11 data has not been
24

(continued)

22 evolutions are quite striking. The adjustment of ULCs was fast and substantial. By the end of 2009, ULCs had declined by close to 25%, and they have remained roughly stable since. While wage cuts played a role initially, over time the reduction in ULCs has come entirely from productivity improvements.
Figure 9. Cumulative Change in Wages, Productivity and ULCs Whole Economy since 2008:3 (2008q3=100) 120 120

110

110

100

100

90

90

80

80

70 2008 ULC 2009 2010 Output per occupied post 2011 2012 2013

70

Wage per occupied post

What matters however for competitiveness is the evolution of ULCs in the tradable sector. The 20% decrease in public sector wages we mentioned earlier may have been essential for the fiscal adjustment, but was of no direct relevance for competitiveness. So Figure 10 plots the evolutions of the same three variables, but now just for manufacturing. The picture is again of a substantial and fast adjustment, but with much less of a decline in wages: Wages in manufacturing barely fell initially, and then increased. The adjustment has come mostly from an increase in productivity. Where did the increase in productivity come from? It clearly came initially with labor shedding: Employment decreased in nearly all sectors, an outcome reflected in the large increase in unemployment. This raised the question of whether the productivity improvement would be long lasting, or reflected for example credit constraints forcing firms to take decisions they would reverse when credit improved. This does not appear to have
revised yet. Instead, data on occupied posts are obtained by surveying enterprises and government institutions and its extrapolation is not affected by the recent correction in population figures.

23 been the case. Productivity gains have remained, indeed, as Figure 10 shows for manufacturing, have continued. Looking across subsectors within manufacturing, productivity continued increasing even in subsectors where employment growth has resumed.
Figure 10. Cumulative Change in Wages, Productivity and ULCs Manufacturing since 2008:3 (2008q3=100) 150 140 130 120 110 100 90 80 70 60 2008 ULC 2009 2010 Output per occupied post 2011 2012 2013 150 140 130 120 110 100 90 80 70 60

Wage per occupied post

Before the adjustment started, one of the main worries was that large nominal wage cuts would be needed, and, judging from the evidence from other advanced countries, this would be a slow and difficult process at best (as it has indeed proven to be in periphery Euro countries). The increase in productivity made this less central of an issue, as, other things equal, smaller nominal wage cuts were needed---and smaller nominal wage cuts were indeed achieved. Still, the large divergence between productivity and wages raises the question: Why were productivity gains not matched by wage increases? Clearly, the large increase in the unemployment rate, weaker unions, and limited employment protection, must have played a central role. So must have the 20% decrease in public sector wages which was part of the 2009 fiscal adjustment. 25 Other factors, specific to Latvia, were also likely at play, although

25

See the next section and the web annex for the results of estimation of Phillips curve relations. The time series is however too short to reach strong conclusions. A Phillips curve specification, allowing for an effect of

(continued)

24 they are impossible to quantify. One is the earlier boom: Latvians probably knew that the earlier large increases in wages were excessive. Looking at the Baltics and Euro periphery countries, Kang and Shambaugh (2013) find that countries that had more of a wage increase in the boom (since 2000) had more of a decrease in the wage later. Another is the recent history of Latvia, the painful transition to a market economy in the early 1990s, the sense of national unity in the face of its Russian neighbor. Yet another is the determination to join Europe and the Euro.

From unit labor costs to prices One would have expected this decrease in ULCs to be reflected in lower prices, and thus higher competitiveness. This however has not been the case. As was shown in Figure 4, the GDP deflator declined initially less than ULCs, and is now higher than it was in 2008. Put another way, the adjustment has come with a large decrease in the labor share (recall that the labor share can be expressed as the ratio of the ULC to the GDP deflator). This is shown in Figure 11, which plots the labor share both for the economy as a whole, and just for manufacturing, starting in 2000.
Figure 11. Latvia: Labour Shares in Whole Economy and Manufacturing (2000-12) 65 65

60

60

55

55

50

50

45

45

40 2000 2002 2004 2006 2008 2010 2012

40

Whole Economy

Manufacturing

public sector on private sector wage inflation, does not yield conclusive results. But it may be that there was a one-time strong effect in 2009.

25 For the economy as a whole, the labor share has fallen from 58% at the peak to 46%. For manufacturing, the share has gone from 64% to 44%. Figures 4 and 10 make clear that this is the mirror image of what had happened during the late part of the boom. Wages had increased faster than the GDP deflator, the labor share had steadily increased; the adjustment has undone this increase---in the case of manufacturing, more than undone it. In short: The adjustment of ULCs and prices was surprisingly fast. And, in contrast to expectations and the textbook adjustment, it came largely from productivity increases, and has been reflected in larger profit margins rather than lower prices. External and internal demand Increases in profit margins typically lead to a supply response, but a standard assumption is that this response may be slow. Exports however increased quickly, and have increased by more than 40% since the output trough in 2009:3. This increase is remarkable given that it was achieved during a period of weak foreign demand growth: The GDP of partner countries has increased by only 11% since 2009:3. Indeed, after falling slightly during the crisis, Latvias export market share has increased from 0.07 to 0.08 percent, higher than before the crisis, a large increase for a small country. Figure 12 shows the evolution of the different components of GDP since the trough, as ratios to 2009:3 GDP. (Figure 12 corresponds to the earlier figure 6, which showed the evolution from peak to trough). It shows a recovery driven by foreign demand: The increase in exports is actually larger than the increase in GDP, a contribution of 23% versus an 18% GDP increase. These two numbers imply that the contribution of domestic demand for domestic goods has actually been negative, -5% of GDP. Domestic demand itself has been reasonably strong, with increases in consumption and investment accounting for 13% and 6% of GDP respectively. But imports have increased by a surprising 24% of GDP. If one takes out that part of imports which is re-exported, about 1/3 of exports, the remaining increase in imports is still 17% of GDP (24% minus 0.3 times 23%), thus nearly equal to the increase in domestic demand---a surprisingly large increase. Part of the explanation must be a rebound from the exaggerated import collapse of 2009. Another, more intriguing, explanation is the reversal of a phenomenon analyzed by Bems and Di Giovanni who, based on supermarket data during the bust, found that consumption had shifted towards lower quality and lower price goods, which tended to be domestic goods. As income has recovered, the reverse effect may have taken place.

26
Figure 12. Evolution of GDP Components Since 2009:3 (percent of 2009q3 real GDP) 30 30

20

20

10

10

-10

-10

-20

-20

-30 2010q1 GDP 2011q1 Consumption 2012q1 Investment Exports 2013q1

-30

Imports

Balance sheet effects, investment and consumption During the boom, as loans were increasingly set in foreign currency, an increasing worry (expressed for example in a number of IMF reports), was that eventual adjustment would lead to strong adverse balance sheet effects. Estimates are that, as of 2008, FX exposure amounted to 25% of GDP for consumers, and 44% of GDP for firms. (Both the public sector (government and central bank consolidated), and banks had a small positive net FX position) Thus, a back of the envelope computation suggested that, say a 20% real devaluation, whether achieved through external or internal devaluation, would decrease the net worth of consumers by 5% of GDP, and the net worth of firms by 8.8%. 26 Balance sheet effects could also come from declines in housing prices. Although only around 25% of households had mortgages, a large decline in housing prices would lead to an increase in the number of households underwater. Mortgages are full recourse, and often backed by personal guarantees (of family members). These adjustments could lead to large wealth effects and a large increase in the proportion of nonperforming loans, preventing the recovery (again, a worry that has proven relevant in a number of Euro periphery countries).
While both internal and external devaluations imply the same FX induced balance sheet effects, their timing can be different. The balance sheet effect is instantaneous in the case of a nominal exchange rate adjustment. But it happens over time in the case of an internal devaluation, and thus gives more time to adjust.
26

27

What happened? Again, the script has deviated somewhat from the feared scenario and from the textbook. The fact that the adjustment has mostly taken the form of productivity increases rather than wage decreases has limited the increase in the ratio of nominal debt to wages. The fact that prices have adjusted less than wages implies that balance sheet effects have affected firms less than households. Housing prices however have halved between 2008:1 and 2010:1 (private estimates suggest the decrease may have been as large as 70%). The results have been an increase in nonperforming loans (NPLs), but an increase that the banks have been able to manage. Firms NPLs peaked at 22% of loans in early 2010, and are now down to 8.5%. Households NPLs stabilized for some time at a high 20%, but have now decreased to 14.1%. There has been substantial restructuring of loans: about 1/3 of the end-2009 stock of loans was restructured between 2010:1 and 2013:2. Cumulative write-offs during that period amounted to 7% of the end-2009 stock of loans. By June 2013 still 10% of the stock of loans were in the work-out process, 14% in the case of loans to households. Despite the still high NPLs, the banking system is in decent shape. As of June 2013, Latvian banks reported a capital adequacy ratio of 18.6%, up from 11% at end-2007 and 14.6% at end 2009. 71% of NPLs were provisioned. Parex no longer exists. It was recapitalized through a conversion of the Treasury deposits into equity and subordinated debt in May 2009. It was then split into a good bank/bad bank in August 2010. Core and some non-core performing assets were transferred to a new bank, Citadele Bank. Remaining assets and liabilities were put in a special purpose vehicle in March 2012. Except for Parex and a small public bank, MLB, the banking system received no public help. 27 Is this the end of the story? Are the macro and financial adjustments achieved? This is the topic of the next section.

Is Most of The Adjustment Complete?

Is most of the adjustment complete? On the one hand, output has increased by 18% since the trough, the current account and the fiscal accounts are roughly in balance, and the financial system seems to be in decent shape. On the other hand, output is still 11% below its peak. Employment has fallen by about 12%. 28 Unemployment is still around 12%. Population is 7.7% lower than it was at the
More recently Latvijas Krajbanka, a medium-size bank, was intervened in November 2011 after fraud was discovered. But the bank did not receive state aid and was liquidated in early 2012. Official data suggests that employment has fallen by about 25% since the peak. But this figure overstates the decline due to a discontinuity in LFS data explained before: the estimate of the number of people employed in
28 27

(continued)

28 beginning of 2008, reflecting mainly a net emigration of 5.7% of the population over the period. 29 This raises at least two issues. How far is unemployment from the natural rate? How should we think about emigration as part of the adjustment process? We take those in reverse order. Emigration While we did not focus on it so far, an important part of the adjustment has taken the form of emigration. Figure 13 gives numbers for net emigration since 2000. It makes clear that Latvian emigration long predates the crisis. The average net emigration rate was 0.5% from 2000-2007. It increased to an average 1.3% from 2008 to 2011, but by 2012, was roughly back to its pre-crisis average. Given the low income per capita, and the fact that Latvia is part of the Schengen agreement on the free circulation of persons within the European Union, such steady emigration is easily explained. Nevertheless, the crisis clearly led to a temporarily higher emigration rate. 30

2011:1 was lowered by as much as 11.5%but data prior to 2011 has not been revised down yet. Splicing the new and old employment series using growth rates for the overlap would suggest that the number of employed people in 2013:2 was 12.4% less than in 2008:2.
29

Interestingly, Eurostat estimates PPP GDP per capita to now be at 9% above its 2008 peak.

Participation in the Schengen agreement started in 2007, although Ireland, the U.K. and Sweden opened their labor markets to those from the new EU member states in 2004. Thus, part of the post 2007 increase may also reflect easier access to labor markets in continental Europe.

30

29
Figure 13. Latvia: Net Emigration 2000-12 45 40 35 30 25 20 15 10 5 2000 2002 2004 2006 2008 2010 2012 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2

Thousands of People (LHS) Emigration Rate (percent of population) (RHS)

What was the effect on unemployment? Two crude back of the envelope computations give plausible upper and lower bounds. We can compute excess emigration as the increase in emigration in 2008-11 over the normal emigration trend, so a cumulative 3.3% over four years. If we assume that, had they stayed, all the emigrants of working age would have remained unemployed, and given a ratio of the labor force to population of about 50%, the unemployment rate would be about 6% higher. If we assume instead that only those emigrants who were unemployed at the time of emigration had remained unemployed, but that the unemployment rate among emigrants was 3 times higher than among nonemigrants, the unemployment rate would be about 3% higher. The question however is whether this emigration is, in some sense, a failure of the adjustment program. In the United States, migration rather than unemployment is the major margin of adjustment to state specific shocks (Blanchard and Katz 1992, with an update and extensions by Dao et al 2013). These adjustments are typically seen as good, indeed as the main reason why the United States functions well as a common currency area: If there are jobs in other states, and if moving costs are low, it is better for workers to move to those jobs than to remain unemployed. Is the answer different for a small country than for a US state? Some economic aspects are different: Some of the costs of running a country are fixed costs, and thus may not be easy to support with a smaller population. In the United States, many of those costs are picked up by the Federal government (although, as we have seen for Detroit, the remaining fixed costs per capita may become too large for a state or a city to function). This is not the case for a

30 country, which must for example finance its defense budget alone. Political aspects may also be relevant, and a country may care more about its size than does a U.S. state. In that respect, an important issue is who the emigrants are, and whether they may return home. The evidence suggests that emigrants during the crisis were slightly younger and slightly more educated than the average population (Hazans 2011). And, from a survey of the relatives of emigrants, only 20% of the emigrants who left during the crisis report an intention of coming back within 5 yearsalthough this rate is higher among more-educated than among less-educated emigrants. Thus, the largely permanent departure of the younger and more educated workers may indeed be costly for those who stay. Unemployment The unemployment rate as of June 2013 was a still high 11.4%. The question is how far this rate was from the natural unemployment rate, and thus how much remained to be done. To think about the question, we have two, admittedly imperfect, tools, the Beveridge curve, and the Phillips curve. Figure 14 plots the Beveridge curve, i.e. the relation between the unemployment rate and the vacancy rate, from 2005:1, the first quarter for which data on vacancies (from an official survey) is available. After an early inward shift of the curve towards the end of the boom, the relation appears quite stable. Indeed, as unemployment starts declining from its peak, it appears to be retracing its movement in the slump. In short, there is no evidence of an adverse shift in the Beveridge curve due to the crisis. 31

To a first approximation, emigration should have no effect on the long run Beveridge curve. In the short run however, it could lead to a shift inwards, as unemployment declines given vacancies. There is however no evidence that, as the emigration rate has returned to pre-crisis levels, the Beveridge curve has shifted back out.

31

31
Figure 14. Latvia: Beveridge Curve (2005-13) 2.4

2.0

1.6
Vacancy Rate
2005q 1

1.2
2013q 1

0.8

2010q 1

0.4

0.0 0 5 10 15 20 25

Unemployment Rate

This does not however settle the issue of what the natural rate might be. For this, we must look at the relation between inflation and unemployment, the Phillips curve. Figure 15 plots core inflation minus average core inflation over the previous 4 quarters against the unemployment rate, as well as the corresponding regression line.
Figure 15. Phillips Curve: Unemployment and Core Inflation (quarterly, 1990-2013) 8 6
Difference in Inflation 1/

4 2 0 -2 -4 -6 -8 0 5 10 15 20 25

Unemployment Rate (SA, percent)

1/ Measured as inflation y/y minus four-period moving average (lagged).

32

The point estimate for the unemployment rate at which core inflation remains constant is about 12%. The figure makes clear however that an unemployment rate below 8-10% has been typically associated with large increases in inflation. This is indeed what we saw earlier when looking at inflation in the boom. While the approach underlying Figure 15 is very crude, most econometric estimates of the Latvian natural rate find it to be quite stable, and to be around 10%. 32 Thus, using this metric, the actual unemployment rate is rapidly approaching the natural rate. This raises a final question, which, while not central to the issues of this paper, is nevertheless intriguing: How can a country with a low minimum wage, weak unions, limited unemployment insurance and employment protection, have such a high natural rate? We do not have a good answer. High unemployment appears to reflect high duration rather than high reallocation and high flows through the labor market. The ``Lilien index, defined as the standard deviation of sectoral employment growth rates for 10 sectors for the decade 2001-2010, is substantially higher in Latvia than in Germany or France. But job turnover, defined as the sum of job creation and job destruction, appears to be similar to that of Germany or France (Boeri and Garibaldi 2006). And unemployment duration appears similar to that for the European Union, with a large upper tail. High reservation wages due to a still extensive informal economy (including home production and barter networks, especially in rural areas), or skill mismatches, may be the cause. To summarize, the actual unemployment rate is still probably higher than, but close to the natural rate of unemployment. Latvia may well want to take measures to reduce its natural rate, but the recovery from the slump is largely complete.

Conclusions and Tentative Lessons

The Latvian boom-bust recovery story is a striking one. We have tried throughout the paper to stick with the facts and avoid normative statements. In the conclusion, we venture to draw some lessons. Some small lessons first: The first is an old one: Healthy booms often turn unhealthy, and policy makers are too late in reacting. The boom in Latvia was healthy until 2006, unhealthy thereafter. Clearly

32

See web annex for the estimation of Phillips curve like specifications.

33 Latvia could not have avoided the adverse effects of the global crisis, but had policy be tighter from 2006 on, the adjustment would have been easier. The second is that, if your financial sector is largely composed of foreign subsidiaries, it is a good idea to be friends with the parent banks. Had Nordic banks not largely maintained funding for their Latvian subsidiaries, the outcome for Latvia would have been much worse. The third is that strict currency boards do not mix well with sudden stops. Had the Latvian Treasury and central bank not deviated from strict currency board rules and not provided funding to the banks, the decline in output would have been much larger. The fourth, and to us perhaps the most interesting, is that productivity gains can happen quickly. The general assumption in European adjustment programs has been that competitiveness would first be achieved by wage cuts, and that productivity increases, spurred by structural reforms, would come over time. In Latvia, productivity increases happened quickly, suggesting substantial X-inefficiency to start. Given that X-inefficiency must be present in other countries as well, an interesting question is why this has not happened elsewhere. The fifth is that, for political purposes, output growth may matter as much or more than its level. One of the striking aspects of the story is that, while most European governments lost to their opposition during this period, Latvia has had the same Prime Minister since March 2009. True, he could blame the previous governments for the fall in output. But for most of his tenure, unemployment was very high, and still he remained in power. The fact that growth was positive, even if output was low, was seen as a sign of success. 33 Turning to larger issues: The debate about Latvia has focused on two aspects of the adjustment program. The first is the large front loaded fiscal consolidation. As we saw, the timing of events makes clear that fiscal adjustment was not responsible for much of the output decrease. And it is also a fact that much of the fiscal adjustment coincided with a return to growth in Latvia. This however does not settle the issue of whether fiscal consolidation had adverse effects on demand and output; as we saw, much of the output collapse was due to a credit crunch and to the option value of waiting. As both were reduced, the impulse for growth may well have been strong enough to offset the adverse direct effects of the consolidation.
In related work, we have explored how trust in government depends on both the level and the change in the unemployment rate, as well as other controls. Using survey data for EU countries from the Euro barometer, we have found roughly equal coefficients on the two variables. A decrease of one percentage point in the unemployment rate can offset a one percentage point higher level of the unemployment rate.
33

34

There is some evidence that the announcement of a clear fiscal path was associated with increased confidence, reflected in lower CDS spreads. But the evidence also suggests that the large decrease in the rates relevant to private borrowers was associated with the increased credibility of the peg and the decrease in exchange rate risk. Whether the front loading aspect of the fiscal adjustment made the whole adjustment program more credible cannot be settled. Adjustment in the 2011 budget was minimal: this may be interpreted as adjustment fatigue, justifying the earlier front loading; or it may be interpreted as coming from the perception that, given front loading, enough had been done already. The second issue is the choice to go for an internal rather than an external devaluation. It is fair to say that it worked better than most analysts had expected. The main argument for an external devaluation, an adjustment of the nominal exchange rate, is that it would work faster; the internal devaluation worked surprisingly quickly as well. It did not quite work in the way one might have predicted. As we saw, much of the improvement in the tradables sector came from productivity increases rather than nominal wage decreases. And much of the decrease in unit labor costs was reflected in an increase in profit margins rather than a decrease in prices. Still, exports increased rapidly, pulling the recovery. One might argue that a nominal devaluation might have led to less pressure on firms to increase productivity. Do these lessons extend beyond Latvia? The evidence from adjustment in Euro periphery countries suggests great caution. Front loaded (although less so than in Latvia) consolidation has been associated with negative growth. Internal devaluations have been associated, at least initially, with labor hoarding, and decreases rather than increases in productivity. The issue is to identify what factors explain these differences. And finally: Is it a wise decision for Latvia to join the Euro next year? Based on the evidence, the answer is that the case for Latvia joining is a strong one, perhaps more so than for some of the existing members. Small open economies are very sensitive to capital flows. Floating may lead to large movements in the exchange rate. One option is to reduce capital mobility. The other is to peg. Pegging may be consistent with sufficient real exchange flexibility if prices, wages, productivity, or migration, can adjust fast enough, as appears to have been the case for Latvia. And if Latvia is going to peg, it is probably better off becoming a member of the Euro and thus having access to a lender of last resort.

35

References
Ashcraft, Adam, 2005, Are Banks Really Special? New Evidence from the FDIC-Induced Failure of Healthy Banks, American Economic Review 95, pp. 1712-1730. Ashcraft, Adam, 2006, New Evidence on the Lending Channel, Journal of Money, Credit and Banking 38(3), pp. 751-775. Aslund, Anders and Valdis Dombrovskis, ``How Latvia Came Through the Financial Crisis, Peterson Institute, May 2011. Bakker, Bas, and Christoph Klingen, eds, ``How emerging Europe came through the 2008/2009 crisis, IMF 2012 Bayoumi, Tam and Ola Melander, 2008, Credit Matters: Empirical Evidence on U.S. Macro-Financial Linkages, IMF Working Paper No. 08/169. Bems, Rudolfs, and Julian Di Giovanni, 2013, Sudden Stops and Expenditure Switching: Scanner Data Evidence from the 2008-09 Global Financial Crisis, mimeo IMF. Bekovskis, K.(2012) Competitiveness of Latvias exporters, Bank of Latvia Working Paper 3/2012. Bernanke, Ben, and Cara Lown, 1991, The Credit Crunch, Brookings Papers on Economic Activity 2, pp. 205-247. Berrospide, Jose and Rochelle Edge, 2010, The Effects of Bank Capital on Lending: What do We Know, and What Does it Mean? International Journal of Central Banking 6(4), pp. 5-54. Blanchard, Olivier 2007,``Adjustment Within the Euro: The Difficult Case of Portugal, April, Portuguese Economic Journal, 6-1, 1-22 Blanchard, Olivier, Mitali Das, and Hamid Faruqee, The Initial Impact of the Crisis on Emerging Market Countries, Brookings Papers on Economic Activity, Fall 2010 Blanchard, Olivier and Lawrence Katz, ``Regional Evolutions, Brookings Papers on Economic Activity, Spring 1992, 1-75. Boeri, Tito and Pietro Garibaldi, 2006, Are Labour Markets in the New Member States Sufficiently Flexible for EMU?, Journal of Banking and Finance, 30-5, 1393-1407, May

36 Biggs, Michael, Mayer, Thomas and Andreas Pick, 2009, Credit and economic recovery, DNB Working Papers 218, Netherlands Central Bank. Claessens, Stijn, Ayhan Kose, and Marco Terrones, (2009), What happens during recessions, crunches and busts? Economic Policy, 24: 653700. Calomiris, Charles and Joseph Mason, 2003, Consequences of Bank Distress during the Great Depression, American Economic Review 93(3), pp. 937-47. Datia, Anosova, et al, 2013 Structural or cyclical? Unemployment in Latvia Since the 2008-2009 Financial Crisis, mimeo Di Comite, Francesco, Gabriele Giudice, Radoslav Krastev, Daniel Monteiro, ``Fiscal Consolidation in the Midst of the Crisis, EU OP 120, November 2012, pp 81-103 Driscoll, John, 2004, Does Bank Lending Affect Output? Evidence from the U.S. States, Journal of Monetary Economics 51(3), pp. 451-71. Duchin, Ran, Oguzhan Ozbas, and Berk Sensoy, 2010, Costly External Finance, Corporate Investment, and the Subprime Mortgage Credit Crisis, Journal of Financial Economics 97, pp. 418435. European Commission, 2005, ``New and updated budgetary sensitivities for the EU budgetary surveillance, http://ec.europa.eu/economy_finance/economic_governance/sgp/pdf/budg_sensitivities_0920 05_v02_en.pdf European Commission, 2012, ``EU balance of payments assistance for Latvia. Foundations of success, Occasional paper 120, November Greenlaw , David, Jan Hatzius, Anil Kashyap, and Hyun Song Shin, 2008, Leveraged Losses: Lessons from the Mortgage Market Meltdown U.S. Monetary Policy Forum Report No. 2, Rosenberg Institute, Brandeis International Business School and Initiative on Global Markets, University of Chicago Graduate School of Business. Hazans, Mihails, ``Coping with Growth and Emigration: Latvian Labor Market Before and After EU accession, mimeo (date unknown, 2007?) Hazans, Mihails, 2011, Chapter 2, in ``Latvia: Human Development Report 2010/2011, Advanced Social and Political Research Institute Hazans, Mihails, and Kaia Phillips, 2011, ``The Post-Enlargement Migration Experience in the Baltic Labor Markets, IZA DP 5878, July

37 IMF, 2010, ``Republic of Latvia 2010 Article IV Consultation, December, IMF Country Report 10/356 IMF, 2012, Policy note for the article IV Latvia consultation. Nov 5, 2007 (can probably not be listed) IMF, 2012, Ex post evaluation of the 2008 Stand-By Arrangement (SBA). (Annex 1: The Latvians authorities view on the ex post evaluation) Kang, Joon Shik, and Jay Shambaugh, 2013, ``Many Paths to the Same End Point, IMF working paper, 13/69 Knack Stephen and Philip Keefer, ``Institutions and Economic Performance; Cross country Tests Using Alternative Institutional Measures, Economics and Politics, 1995 Lown, Cara and Donald Morgan, 2006, The Credit Cycle and the Business Cycle: New Findings Using the Loan Officer Opinion Survey, Journal of Money, Credit and Banking 38, pp. 1575-1597. Dao, Mai, David Furceri, and Prakash Loungani, 2013, ``Regional Labor Adjustment in Curency Unions; Evidence from Europe and the United States, slides for a Presentation at the European Regional Science Association Meihovs, Aleksei, and Anna Zasova, 2009 Assessment of the natural rate of unemployment and capacity utilisation in Latvia, Baltic Journal of Economics 9(2) (2009) 25-46 Mody, Ashoka, 2013, ``Latvia in the Eurozone: A Bet with No Upside, Bruegel, June Peek, Joe, and Eric Rosengren, 2000, Collateral Damage: Effects of the Japanese Real Estate Collapse on Credit Availability and Real Activity in the United States, American Economic Review 90, pp. 30-45. Purfield, Catriaona, and Christoph Rosenberg, Adjustment under a Currency Peg: Estonia, Latvia and Lithuania during the Global Financial Crisis 200809, IMF WP 10/213 Romer, Christina, 1990, ``The Great Crash and the Onset of the Great Depression, Quarterly Journal of Economics, 15-3, 597-624 Rosenberg, Christoph, 2009, ``Why the IMF Supports the Latvian Currency Peg, Economonitor.com, January 6th (www.economonitor.com/blog/2009/01) Rosenberg, Christoph, and Marcel Tirpak, 2008, Determinants of Foreign Currency Borrowing in the New Member States of the EU, IMF WP/08/173

38 Schadler, Susan, Ashoka Mody, Abdul Abiad, and Daniel Leigh, Growth in the Central and Eastern European countries of the European Union IMF occasional paper, 2006. Shambaugh, Jay, and Joong Shik Kang, The Evolution of Current Account Deficits in the GIPS and the Baltics; Many Paths to the Same Endpoint, Working paper, IMF, May 2013 Vamvakidis, Athanasios 2008, Potential Growth Estimates in Emerging Europe Based on a Growth Model, Box 8 of the EUR Regional Economic Outlook, (April). http://imf.org/external/pubs/ft/reo/2008/eur/eng/ereo0408.pdf Vanags, Alf (2013), Latvias Exports: The Real Success Story, BICEPS Research Report http://biceps.org/assets/docs/izpetes-zinojumi/Latvias_exports