International Finance Synthesis PDF

3. Measuring exchange rate misalignment One EURO in Belgium has nit the same value as one EURO in South Africa. There is a value of production and a value of consumption. The most important thing is making the forecast for … Do the forecast (LR, MR, SR) for the guy that is paying you. 3.1. Th...

3. Measuring exchange rate misalignment One EURO in Belgium has nit the same value as one EURO in South Africa. There is a value of production and a value of consumption. The most important thing is making the forecast for … Do the forecast (LR, MR, SR) for the guy that is paying you. 3.1. The exchange rate movements There are several modeling approaches; time trend, moving average, Hodrick-Prescott Filter and Band-Pass Filter. 3.1.1. Deterministic Trend A very simple approach is to estimate the following regression, using OLS. It is a linear trend. Exchange rate trend is the predictable part of the regression 49  3.1.2. Moving Average 𝐾 𝑦𝑡∗ = ∑ 𝑎𝑘 𝑦𝑡−𝑘 𝑘=−𝐾 2K+1 is the window length Symmetric MA (it also exists asymmetric ones) 𝑎𝑘 are the weights, usually they sum up to 1 Each moving average constitutes a filter subject to the Slutzky Yule bias.  lower volatility  phase decay If we are below the moving average, we will buy while when we are above, we will sell. Rules of thumbs using MA for trading strategies (Examples: http://www.swing-trade-stocks.com/moving- averages.html) Focus on long positions only when the 10 SMA is above the 30 EMA. Focus on short positions only when the 10 SMA is below the 30 EMA. No clear mathematical rules, only Rules of thumbs 3.1.3. Hodrick-Prescott Filter 50  3.1.4. End-Point Bias  What if we were estimating the level of potential at the end of 2010?  What if we were estimating the level of potential at the end of 2014? Other existing filters: Band Pass filter, Christiano Fizgerald, Butterworth,... No perfect Filters 51  3.2. Exchange rate misalignments 3.2.1. PPP (Penn) PPP  Real exchange rate is constant. It is used to obtain an equilibrium level of nominal exchange rate;  Choose price index  Choose the equilibrium level of real exchange rate 3.2.1.1. PPP Adjusted for the Balassa-Samuelson and Penn Effects (1) Penn Effect (Observed fact) is a ratio of tradable goods prices to non-tradable prices tends to be lower in high- income countries than in low-income countries. PPP only concern trading good. It makes predictions about price changes rather than price levels. Ex. Suppose we have a developing country. Proportionally we need more tradable goods. On the demand side, when the country is developing, the demand for non-tradable goods is increasing proportionally more than the demand for tradable goods. The price of non-tradable goods will increase proportionally more than the one of tradable goods. From the supply side, when the country is developing, this latter adapts technology. Non-tradable goods are usually services. (2) Balassa-Samuelson hypothesis says that productivity in tradable sector rises faster than productivity in non- tradable as countries develop. They both assume that the labour forces of poor countries are less productive than those of rich countries in the tradable sector but international productivity differences in non-tradable goods are negligible. If the prices of traded goods are roughly equal in all countries, however, lower labour productivity in tradable industries of poor countries implies lower wages than abroad, lower production costs in non-tradables, and therefore a lower price of non-tradables. Rich countries with higher labour productivity in the tradable sector will tend to have higher non-tradables prices and higher price levels.  Equilibrium real exchange rate might change with the level of GDP per capita. 𝐷 𝑁𝑇 𝑃𝑁𝑇 o Demand side :  𝐷𝑇 𝑃𝑇 𝑆𝑇 𝑃𝑁𝑇 o Supply side:  𝑆 𝑁𝑇 𝑃𝑇 There is a real exchange rate appreciation for the emerging countries. PPP should be in line with the level of development of the country. 3.2.1.2. Penn Effect International Comparisons Program (ICP) was initiated in the mid-1900s and led by economists from University of Pennsylvania. PPP under ICP is measured as expenditures valued at domestic prices to expenditures valued at international prices and divided by nominal exchange rate (NCU/FCU), we obtain the real exchange rate. Penn effect – “Real exchange rates is systematically related to the ratio of GDP per capita to U.S. GDP per capita”. 3.2.1.3. Application 52 The proof here is the real exchange rate and the equilibrium is one. We can observe a regularity. Rich countries have high capital-labour ratios, while poor countries have more labor relative to capital. Because rich countries have higher capital-labour ratios, the marginal productivity of labour is greater in rich countries than in poor countries, and the former will therefore have a higher wage level than the latter. Non-tradables, which consists largely of services, are naturally labor-intensive relative to tradables. Because labour is cheaper in poor countries and is used intensively in producing non-tradables, non-tradables also will be cheaper there than in the rich, high wage countries. This international difference in the relative price of non-tradables suggests that overall price levels, when measured in a single currency, should be higher in rich countries than in poor countries. The PENN indicator means that we have to control the level of GDP for … Absolute PPP cannot be taken into account because of the presence of non-tradable goods. PPP  r = 100 but r appreciates with the development of the country following Balassa Samuelson. When a country is undervalued, it becomes cheaper and thus, more competitive. Ex. Lutvia, at that time the European Union did not accept them because it would have led to an appreciation of the EUR and thus, to a loss of competitiveness.  Not the same level of technology! 𝑃𝑁𝑇 → 𝑖𝑛𝑡𝑒𝑟𝑛𝑎𝑙 𝑒𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝑟𝑎𝑡𝑒 𝑃𝑇 𝐸𝑃𝑇 → 𝑒𝑥𝑡𝑒𝑟𝑛𝑎𝑙 𝑒𝑥𝑐ℎ𝑎𝑛𝑔𝑒 𝑟𝑎𝑡𝑒 𝑃𝑇∗ 3.2.2. Macroeconomic Balance Approach – CGER/EBA (IMF) Find exchange rate that equates underlying current account with assessment of equilibrium current account (called the norm). 3.2.2.1. Definition of Underlying Current Account Balance (UCUR) (1) Relevant concept for thinking about medium-run projection once temporary and cyclical factors have run their course. (2) Current account position that would prevail if output was at potential at home and abroad and the effects of past exchange rate changes had been fully realized. (3) Should also adjust for other factors temporarily influencing current account (e.g. one- off large FDI) 3.2.2.2. Basic Steps in the Macroeconomic Balance Approach (CGER) Step 1: Calculate the underlying current account (UCUR) at prevailing R. Call these UCUR1 and R1. 53  CA = TB + rF* + Rem  depending on the flows Income of foreign assets Closed economy: Y = C + I + G (CG+IG) Y–C=I 2 types of definition S=I Opened economy: Y – C = I + CA S – I = CA Positive CA  S > I Is it good or bad to have a deficit? No answers, it depends on the economy. UCUR means a structural Current Account and the term “structural” means that we remove the fluctuations/cycles. Step 2: Calculate the current account “norm” (or equilibrium level of S-I) Step 3: Calculate the GAP Here, we have to calculate the difference between UCUR and the Norm CA and if UCUR > Norm CA, that means that we make a diet. Step 4: Calculate change in R needed to make UCUR equal to equilibrium level of S-I. The adjusted level of R is the equilibrium REER (R*). The elasticity is the reaction to exchange rate: # Very elastic  small diet # Inelastic  large diet Step 5: Analyze sensitivity of R* to assumptions about equilibrium S-I and other considerations. 3.2.2.3. External Balance Assessment (EBA)  New generation EBA is successor to “CGER” methodology. It has a broader set of determinants and focuses on present position rather than on the medium term. Moreover, it takes into account the economic cycle and has a normative part (actual and “recommended” policies). See http://www.imf.org/external/np/res/eba/ Ex. To know how much the exchange rate should vary to close the gap you need to estimate elasticities.  Difficult part TB = X – eM domestic demand = X(Y*, e) – eM(Y, e) 54 𝑑𝑇𝐵 𝑑𝑒  impact of a change in the exchange rate on the trade balance 𝜕𝑋(.) 𝜕𝑀(.) = 𝜕𝑒 de – [ M(.) + 𝜕𝑒 de ] = 0 𝜕𝑋(.) 𝑑𝑒 𝑀 𝜕𝑀(.) 𝑑𝑒 = 𝜕𝑒 𝑀 -𝑀- 𝜕𝑒 𝑀 =0 [divide everything by M] 𝑑𝑇𝐵 𝜕𝑋(.) 𝑑𝑒 𝜕𝑀(.) 𝑑𝑒 = –1- =0 𝑑𝑒 𝜕𝑒 𝑀 𝜕𝑒 𝑀 𝜕𝑋(.) 𝑋 𝜕𝑒 𝑒 𝑑𝑒 – 1 – ηM/e eM = X and M = X/e Elasticity export : ηX/e 𝑑𝑇𝐵 𝑑𝑒 = ηX/e - ηM/e – 1 = 0  Marshall-Lerner What does it mean? Depreciation = positive impact on TB if the elasticity of exports is bigger than the elasticity of imports. 𝑑𝑇𝐵 > 0 → ηX/e - ηM/e > 1 𝑑𝑒 𝑑𝑇𝐵 𝑑𝑒 > 0 → ηX/e - ηM/e < 1 Ex: In Belgium, if we go back to the Belgian Franc, we would devaluate our currency to be more competitive. If weak BF → the value of IM ↑. The curve on the left is called the J-curve. For small economy, it’s the second relation that works. Marshall-Lerner condition only works for big opened economy. Trade barrier in the U.S. (Trump) create inflation because IM & EX do not substitute so there is an increase in the prices. Most of the time depreciation improve your CA but sometimes not. 55  3.3. Forecasting exchange rate 3.3.1. Using the deterministic trend Forecasting with the trend  Trend equation is: 0.0027x+0.9717 So for June 2014 (observation 185) forecast is 0,0027*185+0.9717= 1,4712 Be aware that there is uncertainty due to estimation error (R2=0,61). So the presence of confidence bound is advice able. Is there really a deterministic trend? This hypothesis should be tests. Is 0.0027 statically different from 0? T-stat analysis (via toolkit): 3.3.2. Using the stochastic trend Summary It is not possible to forecast a variable with a unit root. The unit root test are necessary to be implemented beforehand. If unit root is discovered, then use change in exchange rate. The unit root test Is it possible to forecast a random walk? In practice, the regression is used: ∆𝑦𝑡 = 𝑦𝑡−1 + 𝑢𝑡 and the test is =0. The distributions of the DF test does not follow standard distributions and have been tabulated. Different DF tests are available according to the presence of a drift or/and a deterministic trend. Shocks never die = stochastic trend Critical Values for the DF Test 3.3.3. Using the moving average Forecasting with the moving average  Use an AR(k) process 𝑦𝑡 = 𝑎1 𝑦(𝑡−1) + ⋯ + 𝑎𝑘 𝑦𝑡−𝑘 + 𝑢𝑡. “k” is determined by information criteria (AIC, BIC,...). There Structural models is an alternative method which is the fact of using the lag until which ak is significant. (1) Using theoretical models and fundamental variables (2) Using future markets 56