Panama: Selected Issues Paper

This paper presents empirical evidence on the effects of achieving investment grade on borrowing costs for the sovereign and the private sector. This study provides background information on sovereign credit ratings and compares Panama’s key macroeconomic and institutional characteristics with those of other emerging markets. Statistical evidence on the reduction in sovereign spreads associated with obtaining investment grade status and the impact of the sovereign’s upgrade on corporate financing costs were also discussed. The model is estimated using a variety of panel regression techniques.


This paper presents empirical evidence on the effects of achieving investment grade on borrowing costs for the sovereign and the private sector. This study provides background information on sovereign credit ratings and compares Panama’s key macroeconomic and institutional characteristics with those of other emerging markets. Statistical evidence on the reduction in sovereign spreads associated with obtaining investment grade status and the impact of the sovereign’s upgrade on corporate financing costs were also discussed. The model is estimated using a variety of panel regression techniques.

II. Macro-Financial Linkages in Panama 1

A. Introduction

1. The recent global crisis has given rise to a renewed interest in macro-financial linkages. The economic literature (e.g., Kiyotaki and Moore, 1997) has shown that shocks to economic activity can have quantitatively large adverse effects on the financial system, which then give rise to second-round effects on the real economy through reduced credit availability. Recent empirical support for this hypothesis using U.S. data is provided in Bayoumi and Melander (2008). While this type of credit cycles have not been extensively studied for emerging markets, they are highly relevant for Panama, given its high level of financial intermediation and the key role played by credit during the recent high-growth period.2

2. Feedback loops from credit to activity reflect both demand and supply effects. The linkages from credit to the real economy are commonly thought to be supply-driven, i.e. that reduced credit availability constrains spending by households and firms. However, changes in credit can also be demand-driven. While real demand is usually proxied by GDP growth, credit supply is often measured through surveys on lending standards—whenever available—and some measure of loanable funds, such as bank deposits. The effects of lending rates on credit are typically thought to represent a mix of supply and demand effects.

3. A Bayesian VAR (BVAR) model is used to study macro-financial linkages in Panama during 1999–2009. A central question of the paper is the importance of domestic feedback loops between the real and financial sectors, and the transmission of international financial and real shocks. The model is estimated using a methodology developed by Villani (2008), which allows for the specification of informative steady-state priors for the variables used. An important advantage of this methodology is that it substantially reduces the problem of degrees of freedom often associated with conventional VARs.

4. The paper finds clear evidence of domestic macro-financial linkages and confirms previous findings of strong external spillovers to real activity in Panama. The results show that GDP growth responds to domestic credit, and that the response is stronger than to other domestic macroeconomic variables. At the same time, credit growth is found to be mainly demand-driven. Finally, the paper confirms previous findings of important spillovers from U.S. GDP growth to economic activity in Panama (Swiston, 2010), with a positive one standard deviation shock to U.S. growth increasing the growth rate of real GDP in Panama by about 1½ percentage points.

5. The rest of the paper is organized as follows. Section B briefly describes the financial sector and the evolution of growth and lending during the crisis. Section C introduces the model. Section D presents the main results, and section E concludes.

B. Background

6. Panama showed an impressive growth record prior to the crisis, and avoided a recession following the global shock. Real GDP growth averaged 6 percent during 1999–2008, driven by rapid growth in private investment and exports. Growth slowed to 2.4 percent in 2009, but remained above the average for the region.

7. Panama experienced a healthy expansion of credit prior to the crisis.Panama has a very high level of financial intermediation compared to the region, with bank credit amounting to about 90 percent of GDP. Growth of domestic credit to the private sector averaged about 14 percent during 2004—08, against a backdrop of strong economic growth. It declined to about 1 percent in 2008 in the context of the global crisis and the domestic slowdown.


Domestic Credit of the Banking System

US$ millions

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002

8. The financial system held up well during the global crisis. Bank financial soundness indicators remained strong, while private deposits continued to grow at a solid pace, aided by inflows from the region. Given that retail funding is the dominant source of bank financing in Panama, funding to banks was affected relatively little by the international drought in interbank markets. This may be regarded as a priori evidence that lack of liquidity, or supply effects, was not the main contributor to the slowdown of domestic credit growth.

9. Domestic lending rates fell, but much less than U. S. rates. While the increases in lending rates before the global crisis were of similar magnitudes in Panama and the U. S., the decline in rates following the crises was more modest in Panama. One possible explanation for the asymmetry is that Panamanian banks adopted relatively more conservative lending standards following the financial crisis partly owing to the absence of lender of last resort. Sustained high deposit rates also likely limited the scope for decreasing lending rates.


Panama Consumer Credit Rates (1-year) vs US Rates

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002

C. Model and Empirical Implementation

10. VAR models are useful for estimating and forecasting the dynamic responses of economic systems, but the short samples available are a challenge. VAR models have been widely used since they impose little restrictions on the system and perform well in forecasting, provided sufficiently long time series are available. The fact that samples are often small places an important constraint on the number of variables that can be included in the model due to the associated relatively few degrees of freedom.

11. Bayesian VAR modeling helps overcome small-sample shortcomings and forecasting. By including relevant prior information regarding the steady-state values of some variables, the degrees-of-freedom problem can be mitigated. As suggested by Villani (2008), using informative priors on the steady—state level of a variable makes the forecasts converge to a reasonable level. If the priors are appropriate, forecasting performance will be improved, as has been shown empirically (Adolfson et al, 2007, Österholm, 2008, Österholm and Zettelmeyer, 2008).

12. The model is given by


where G(L)= I- GtL-...-GpLp is a lag polynomial of order p, xt is an n x 1 vector of stationary macroeconomic variables and ηt is a n x 1 vector of iid terms with properties E(ηt) = 0 and Et ηt) = £. In this model, ψ represents the steady state, over which the researcher is assumed to have informative prior information.

The prior on Σ is given by p(Σ) | Σ|-(n+1)/2, the prior on vec(G), where G = (G1 ... Gp) ’, is given by vec(G)Npn2(θG,ΩG) and the prior on ψ is given by ΨNn(θψ,Ωψ). In practice, this implies that only the priors of the vector of dynamic coefficients vec(G) and the steady state parameters ψ will typically be informative.

13. Both domestic and external variables were included in the empirical specification. Following Österholm and Zettelmeyer (2008), we separate the vector of variables xt into a domestic and a foreign block:3


where the external block comprises:

  • ΔytUS, growth rate of U.S. real GDP,

  • iUSt, the nominal U.S. Federal Funds rate,

  • LOStUS, the U.S. Federal Reserve’s Senior Loan Officer Survey, a non-price measure of credit availability (Swiston, 2008, Bayoumi and Melander, 2008),

  • HYt, the U.S. high-yield corporate bond spread, included as a proxy for global risk aversion,

  • ΔFDItPAN, the ratio of Panama’s foreign direct investment to GDP.

The domestic block is given by:

  • ΔytPAN, Panama’s growth rate of real GDP,

  • itPAN, Panama’s nominal 3-month consumer lending rate in the onshore banking system,

  • credittPAN, Panama’s growth rate of real credit to the private sector by the onshore banking system,

  • deptPAN, Panama’s growth rate of real credit supply,

  • ΔgtPAN, Panama’s ratio of government spending to GDP.

14. The identification strategy for ordering the variables of the system was based on previous research. As in Villani (2008), it is assumed that the U.S. Federal Funds rate is the most exogenous variable, followed by the other variables in the external block. In the domestic block, the lending rate was assumed to be the least endogenous variable, followed by government spending and GDP growth; private sector credit and deposits were assumed to be the most endogenous. We assumed that no variables in the domestic block affect the external block, in line with the small size of the Panamanian economy compared to the U.S.

15. To better ascertain the role of financial factors in economic activity, the model specification separates demand and supply of credit. Following the literature, credit demand is proxied by real GDP growth. When available, survey-based lending standards have been found to be a useful measure of credit supply (Bayoumi and Melander, 2008, Calani et al, 2010, Lown and Morgan, 2000). However, in the absence of such surveys for Panama, we follow Daseking et al. (2003) and use total deposits in the onshore banking system as a proxy for credit availability.

16. The model is estimated with data for the last decade and allows for a structural break in growth rates in Panama. A sample at the quarterly frequency starting in the last quarter of 1999 was used. Evidence points to potential GDP growth increasing markedly in Panama starting in 2003. Data suggest a structural break also in other model variables around this time. Given that this could have potentially shifted the underlying relationships between the variables in the model, we included a dummy that takes on the value of one from that year onwards. This also means that separate priors were defined for the two subperiods. Finally, we set the lag length at 4.

17. In view of the lack of conclusive evidence in previous research, numerical values for priors were mainly taken from the data. Priors for the U.S. interest rate was based on combing an inflation target of around two percent with a real interest rate of two percent as spelled out by the Fisher hypothesis for the earlier part of the period, while the dramatic interest rate changes during the latter part of the sample period led us to choose a wider distribution. Priors on growth rates for U.S. and Panamanian output were taken from desk estimates in the IMF’s Western Hemisphere department for both subperiods.4 For the rest of the variables, neither previous research nor theory could provide estimates, which is why relatively wide distributions were chosen.5

D. Results

Impulse Responses

18. The main variables of interest are GDP and private sector credit growth. Figure 1 (left and right panels) presents impulse responses for these two variables. The full set of impulse responses are found in Figures A1 in the Appendix. Impulse responses are generated in a standard fashion, i.e. reflect one-standard deviation shocks.6 Coefficients are often significant and generally of the expected sign, both for the external and domestic block.

Figure 1.
Figure 1.

Impulse Responses for Panama GDP Growth and Credit

Left panel: PAN GDP. Right panel: PAN credit

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002

Note: The shaded areas correspond to 95 and 68 percent confidence intervals.

19. Among the domestic variables, credit is more important than government spending for GDP growth. The results suggest that a one standard deviation increase in bank credit growth in Panama results in an increase of ½ percentage points in real GDP growth. The impact of fiscal policy is about a fifth that of credit, i.e. a one standard deviation increase in the ratio of government spending to GDP leads to 0.1 percentage points increase in real GDP growth. Credit supply (proxied by deposits) is found not to affect GDP growth in a significant fashion.

20. Spillovers to the real economy from U.S. growth shocks are large, but other external variables do not seem to have significant effects. Changes in U.S. real GDP growth have a strong and persistent impact on the Panamanian economy. A one standard deviation increase in U.S. growth leads to an increase of about 1½ percentage points in the rate of GDP growth in Panama. A one standard deviation increase in the FDI-to-GDP ratio increases economic growth by ¼ percentage point. Coefficients for the high-yield bond spread and U.S. interest rates have the expected negative sign, but the effects are not statistically significant.

21. Estimated spillovers from U.S. growth are broadly similar to those found in previous studies. Using a structural VAR (SVAR) framework, Swiston (2010) finds that a positive one standard deviation shock to U.S. GDP growth raises the growth rate of activity by about 1¼ percentage points in Panama. The impact of shocks from other advanced economies, as well as from the rest of Central America, on Panama’s GDP growth is found to be nil.

22. Credit supply and lending standards in the U.S. are not found to have significant effects on Panama’s credit growth. The impulse responses in Figure 1 (right panel) show that growth in total deposits in Panama has a positive effect on credit growth, but the coefficient is small and insignificant. U.S. lending standards affect credit in Panama within one year and in the expected direction: an easing in standards translates into higher willingness to lend and hence increases credit, but the results are also not significant. It is, however, possible that our measure of credit supply fails to take into account changes in credit standards (unrelated to the availability of loanable funds), which were tightened in Panama during the recent crisis.

23. Credit growth is to a large extent driven by developments in the domestic economy. A one standard deviation increase in domestic economic growth leads to an increase of about 0.9 percentage points in domestic credit growth. This suggests that the response of credit to changes in activity is about twice as large as the response of activity to changes in credit. No other domestic or external variables (including interest rates and risk aversion) are found to have significant effects on credit in Panama. Taken together, this provides support to the hypothesis that credit follows developments in the real economy.

Variance Decomposition

24. The estimated model can explain about half of the shocks to GDP growth in Panama over the last 10 years. As shown in Figure 2, own shocks to Panama’s GDP explain around 50 percent of the variance of output growth, which is a reasonable proportion for a VAR. Shocks to U.S. GDP explain around 25 percent at the ten-quarter horizon. Shocks to U.S. interest rates, lending standards and domestic credit explain the remaining 25 percent.

25. Credit developments are sensitive to several variables. Own shocks to credit also explain about 50 percent of the variance (Figure 2). Shocks to U.S. and Panama’s output explain about 15 percent each, while the remaining 20 percent can be attributed to shocks to real interest rates in the U.S. and Panama.

E. Concluding Remarks

26. Our estimated model finds important two-way linkages between the real economy and the financial sector in Panama. Real GDP in Panama is found to be strongly affected by credit to the private sector, FDI and government spending. However, the effect from all domestic variables is found to be less important than that from U.S. GDP growth. A one standard deviation shock to U.S. growth is found to increase the growth rate of Panama’s real GDP by about 1½ percentage points. We also find large effects of real activity on credit growth in Panama (twice the size of the effect of credit on real growth), suggesting that credit is mostly demand-driven. Measures of credit supply are not found to have significant effects on credit growth.

27. The results suggest that credit growth in Panama will increase as the recovery firms up. The increase in GDP growth starting in the last quarter of 2009 was likely aided by an improved external environment. With GDP growth expected to recover fast, the model predicts a significant rebound of bank credit to the private sector.

Figure 2.
Figure 2.

Variance Decompositions

Left panel: PAN GDP. Right panel: PAN credit

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002


Figure A1.
Figure A1.

Impulse Responses

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002

Figure A2.
Figure A2.

Variance Decompositions

Citation: IMF Staff Country Reports 2010, 315; 10.5089/9781455208647.002.A002


  • Adolfson, M., M.K. Anderson, J. Lindé, M. Villani, and A. Vredin, 2007, “Modern Forecasting Models in Action: Improving Macro Economic Analyses at Central Banks,” International Journal of Central Banking 3, pp. 11144.

    • Search Google Scholar
    • Export Citation
  • Bayoumi, T., Melander, O., 2008, “Credit Matters: Empirical Evidence on U.S. Macro-Financial Linkages”. IMF Working Paper 08/169.

  • Calani, M.C., Garcia, P.S., Oda, D.Z., 2010, “Supply and Demand Identification in the Credit Market”, Central Bank of Chile Working Papers no. 571

    • Search Google Scholar
    • Export Citation
  • Daseking, C., Ghosh, A.R., Thomas, A.H., Lane, T. D., 2003, “Lessons from the Crisis in Argentina”, IMF Occasional Paper no. 236.

  • Kiyotaki, N., Moore, J., 1997, “Credit Cycles”, Journal of Political Economy, Vol 105 (2), pp. 211248.

  • Litterman, R.B. 1986. “The limits of counter-cyclical monetary policy: an analysis based on optimal control theory and vector autoregressions,” Working Papers 297, Federal Reserve Bank of Minneapolis.

    • Search Google Scholar
    • Export Citation
  • Lown, C., Morgan, D., Rohatgi, S., 2000, “Listening to Loan Officers: The Impact of Commercial Credit on Lending and Output”, Federal Reserve Bank of New York Economic Policy Review (May), pp. 217235

    • Search Google Scholar
    • Export Citation
  • Swiston, A., 2010, “Spillovers to Central America in Light of the Crisis: What a Difference a Year Makes”. IMF Working Paper 10/35.

  • Swiston, A., 2008, “A Financial Conditions Index for the United States: Putting Credit Where Credit is Due”, IMF Working Paper 08/61.

  • Villani, M., 2009, “Steady State Priors for Vector Autoregressions,” Journal of Applied Econometrics, Vol. 24 Issue 4, pp. 630650

  • Österholm, P., Zettelmeyer, J., 2008. “The Effect of External Conditions on Growth in Latin America,” IMF Staff Papers, vol. 55(4), pp. 595623.

    • Crossref
    • Search Google Scholar
    • Export Citation

Prepared by Juliana Araujo and Kristin Magnusson.


Panama has no national currency or central bank and has the U.S. dollar as its only legal tender. This means that neither money supply nor credit are buffered or amplified by domestic monetary policy considerations.


Previous work (Swiston, 2010) has found trade in goods and services to be an important transmission channel of shocks from the United States to Panama, but we chose not to include it given the presence of U.S. GDP growth in the model and the paper’s focus on financial and domestic factors. Panama’s large off-shore financial center is also excluded from the model because the segment is largely de-linked from the rest of the economy.


For the first subperiod the prior on US GDP growth was between 1 and 3 percent and for the latter subperiod was -1 and 2 percent. The prior on Panama’s GDP growth rate was between 2 and 8 percent and 6 and 10 percent, respectively.


For the priors governing the dynamics of the model, we follow Litterman, 1986, in using a modified version of the Minnesota prior. If a variable is modified in levels, the prior mean on its first own lag is set to 0.9; if in growth rates, it is set to 0. The reason for modifying the traditional Minnesota prior in this fashion is that a prior mean on the first own lag equal to 1 is theoretically inconsistent with a mean-adjusted model, since a random walk does not have a well-specified mean.


One standard deviations shocks correspond to the following magnitudes for the included variables: 6.6 percent for PAN credit growth, 1.5 percent for PAN government spending, 4 percent for PAN FDI, 4 percent for PAN GDP, and 2 percent for U.S. growth.

Panama: Selected Issues Paper
Author: International Monetary Fund