## Exchange Rate Misalignment and Growth: A Myth?

## Source details

**Canonical URL:** [Exchange Rate Misalignment and Growth: A Myth?](https://www.imf.org/-/media/files/publications/wp/2017/wp17283.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17283.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17283.pdf.json)

---

### Abstract and central question
- The paper investigates whether real exchange rate deviations ("misalignments") are associated with GDP per capita growth across a broad panel of countries.
- Key claim: The association between exchange rate misalignment and growth is very weak; any positive association of a more depreciated currency with growth vanishes after controlling for the savings rate. This holds for both the full sample and the emerging-economies subsample.
- JEL Codes: F31, F43, O47. Keywords: Real exchange rate, growth, misalignment.

### Data and methodology
- Data: Penn World Table (PWT 9.0), period 1950 to 2014, divided into 13 five-year periods (1950-54, 1954-60, ..., 2010-14); variables are averages across each period.
- Two-stage estimation (Rodrik (2008) style):
  - First stage: lnRER_it = a + b:lnY_it + t + e_it
    - RER: nominal exchange rate (local currency/USD) divided by local price level (PPP); U.S. price level normalized to 1.
    - Controls: time fixed effects and ln GDP per capita to capture Balassa-Samuelson and global shocks.
    - Residuals Dev_RER_it are used as measure of deviations/misalignments.
  - Second stage: g_it = β + γ:lnY0_it + δ:Dev_RER_it + κ:Savings_it + μ_i + τ_t + ε_it
    - g_it: average annual growth rate of GDP per capita in period t (multiplied by 100 in reported regressions).
    - lnY0_it: log of initial GDP per capita (first year of period t).
    - Savings_it: average savings rate over period t, defined as one minus the shares of private consumption and government consumption in GDP.
    - Country and time fixed effects included. Regressions run with and without the savings control; also robustness checks exclude outliers.

### First-stage (RER) results
- Estimated equation (with full set of time dummies):
  - lnRER_it = 3.778 (34:17) + (−0.187) (−16:95) lnY_it
  - N = 1899; R^2 = 0.456
  - Interpretation: GDP per capita negatively associated with log-level RER (more developed economies tend to have stronger currencies — Balassa-Samuelson effect).
- Correlation between RER deviations and savings (panel regressions with country and time fixed effects):
  - All countries: Savings_it coefficient = 0.559 (4.84); N = 1899; #countries = 182; R-squared = 0.140.
  - Emerging economies: Savings_it coefficient = 0.477 (3.82); N = 1401; #countries = 166; R-squared = 0.123.
  - Conclusion: Dev_RER and savings are positively and highly significantly correlated in the sample.

### Second-stage (growth) results — main findings and robustness
- Dependent variable: average annual growth rate of GDP per capita (multiplied by 100).
- Key estimates (selected coefficients reported exactly as in the paper; z-statistics in parentheses):
  - lnY0_it (convergence term) across specifications is negative and highly significant:
    - Examples: −4:261 (−10:1), −4:405 (−10:3), −3:161 (−6:31), −3:819 (−8:20).
  - Dev_RER_it coefficients (columns (1)–(8) of Table 3):
    - 1:318 (2:54)
    - 0:694 (1:16)
    - 0:749 (1:45)
    - 0:380 (0:75)
    - 1:058 (1:72)
    - 0:319 (0:45)
    - 1:068 (2:35)
    - 0:625 (1:32)
  - Savings_it coefficients (selected):
    - 3:028 (1:84)
    - 10:21 (6:44)
    - 3:224 (1:29)
    - 9:447 (4:45)
- Main empirical patterns:
  - Full sample, without savings (column (1)): Dev_RER_it positive and statistically significant (1:318 (2:54)). A one standard deviation increase in Dev_RER is associated with a growth rate higher by 1:318 × 0:454 = 0:60 percentage points (paper notes: "A one standard deviation increase in this variable is associated to a growth rate higher by 1:3180:4540:60 percentage points"; mean growth = 0:023).
  - Once Savings_it is included (column (2)), the estimated Dev_RER effect falls substantially and becomes statistically insignificant (point estimate declines to 0:694 (1:16); magnitude roughly halved).
  - Excluding outliers (columns (3) and (4)) the Dev_RER coefficient becomes statistically insignificant even without controlling for savings (e.g., 0:749 (1:45) and 0:380 (0:75) in shown columns).
  - For emerging economies subsample (columns (5)–(8)): results broadly similar — no evidence that devaluations have stronger growth effects for emerging economies; Dev_RER effects become insignificant when savings are included.
- Sample sizes:
  - Examples from Table 3: N = 1879 (columns with full sample), N = 1247 (full sample excluding outliers), N = 1401 (emerging economies full), N = 897 (emerging economies excluding outliers).
  - # countries examples: 182, 178, 166, 158 per columns reported.
- Interpretation: Inclusion of savings (a plausible confounder) substantially reduces the magnitude and statistical significance of the Dev_RER effect on growth; excluding outliers also eliminates significance.

### Asymmetry: overvaluations versus undervaluations
- Alternative specification allowing asymmetric effects:
  - g_it = β + γ:lnY0_it + δ:Dev_RER_it + θ:P_it×Dev_RER_it + κ:Savings_it + μ_i + τ_t + ε_it
  - P_it = 1 if Dev_RER_it ≥ 0 (currency undervalued), 0 otherwise.
  - Marginal effect: overvalued currency captured by δ; undervalued currency by δ + θ.
- Selected estimates (Table 4, with savings included):
  - Full sample (columns (1)–(2)):
    - Dev_RER_it = 0:889 (1:20) and 0:015 (0:02) in two specifications; P_it×Dev_RER_it = −0:449 (−0:31) and 0:786 (0:66).
    - Savings_it = 2:959 (1:78) and 10:16 (6:41).
    - Conclusion: Dev_RER and interaction terms statistically insignificant for full sample.
  - Emerging economies (columns (3)–(4)):
    - Column (3): Dev_RER_it = 1:944 (1:76); P_it×Dev_RER_it = −3:184 (−2:06); Savings_it = 2:683 (1:06) and 9:446 (4:44) depending on whether outliers excluded.
    - However, when outliers are excluded (column (4)), the Dev_RER and interaction coefficients become insignificantly different from zero.
- Interpretation: No robust evidence that undervaluation spurs growth or that overvaluation harms growth once savings and outliers are accounted for.

### Conclusions and policy-relevant takeaways
- Core conclusion: Contrary to Rodrik (2008), this paper finds no robust evidence that real exchange rate misalignments matter for growth across a broad panel of countries or for an emerging-economies subsample once domestic savings are controlled for and outliers are handled.
- Mechanism highlighted: Savings are a potential confounder — higher savings can both increase growth and be associated with more depreciated currencies (positive correlation between Savings_it and Dev_RER_it), biasing upward estimates that omit savings.
- Robustness: Results hold across specifications that (i) include country and time fixed effects, (ii) control for initial GDP per capita (convergence), (iii) include the savings rate, and (iv) exclude outliers.
- Implication for policymakers and researchers: Policy emphasis on exchange rate undervaluation as a general tool for boosting growth is not supported by these panel estimates once savings behavior is accounted for; the role of domestic savings and investment should be considered when assessing exchange rate–growth links.

*WP/17/283 — Carlos Eduardo Goncalves and Mauro Rodrigues, November 24, 2017.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17283.pdf_