Exchange rate pass-through in the Colombian car market

Carranza, González, Perez and Vélez

Intro

Motivation

Understanding pass-through is crucial for central banks and policy makers
Incomplete pass-through is a fact of life.
We delve into the micro determinants of incomplete pass-through.
Car industry is highly exposed to exchange rate shocks.

What we do

A model for the car market equilibrium

Estimate demand for new cars
Assume Bertrand-Nash
Recover marginal costs
Use model to run experiments
We find:
- more pass-through for high-tiered cars
- close to no pass-through for cars of Latin American origin
- firms absorb most of the cost shocks
- menu adjustment plays an important role

Theoretical Literature

Market structure: Dornbusch et al. (1987), Atkeson and Burstein (2008), Corsetti and Dedola (2005).
Curvature of demand: Knetter (1989), Bergin and Feenstra (2001).
Local costs: Sanyal and Jones (1982), Burstein, Neves, and Rebelo (2003)

Empirical literature

Timing: Parsley and Wei (2001), Goldberg and Campa (2010)
Consumer behavior: Gopinath et al. (2011), Amiti, Itskhoki, and Konings (2014), De Loecker et al. (2016), Fabra and Reguant (2014), Miller, Osborne, and Sheu (2017).
Firms response: Auer Raphael A and Schoenle (2016), Amiti, Itskhoki, and Konings (2019), Muehlegger and Sweeney (2021), Auer Raphael, Burstein, and Lein (2021)
Honorable mention: Nakamura and Zerom (2010)

The data and the depreciation

The data

Shares of specific models 2007-2016
- All new car sold in Colombia during those years
Prices and characteristics
- Average yearly prices from an association of insurers
Household characteristics
- Gran Encuesta Integrada de Hogares (CPS analogous)

Car data (I)

Car types

Year	Total	Imported	Type of Car		Models
			Automobile	Pickup
2007	192,993	58.90	70.22	14.23	543
2008	170,895	63.85	67.17	15.11	593
2009	148,432	66.54	69.09	14.51	568
2010	212,013	64.92	65.68	17.43	672
2011	267,342	66.78	67.97	18.05	770
2012	255,263	67.97	61.88	16.91	977
2013	245,825	68.88	60.08	9.31	1,008
2014	277,690	64.95	58.27	7.40	1,115
2015	242,536	64.28	60.98	6.16	1,117
2016	224,083	62.55	61.17	6.48	1,018

Car data (II)

Car sales

Year	Engine Displacement		Price (Million COP)				Sales
	S. Avg.	W. Avg.	S. Avg.	W. Avg.	Min.	Max.	(Billion COP)
2007	2.20	1.80	75.80	50.42	17.23	293.85	9,731.30
	(0.83 )	(0.66 )	(44.22 )	(27.07 )
2008	2.24	1.87	69.34	48.89	16.42	298.83	8,355.27
	(0.82)	(0.66)	(40.66)	(25.97)
2009	2.23	1.81	70.84	45.69	11.90	467.32	6,781.45
	(0.83)	(0.65)	(49.98)	(27.10)
2010	2.24	1.86	69.71	45.16	13.30	419.04	9,573.82
	(0.86)	(0.65)	(42.41)	(25.03)
2011	2.15	1.82	64.51	40.99	13.56	343.53	10,957.56
	(0.85)	(0.63)	(40.10)	(24.27)
2012	2.14	1.84	62.76	40.92	14.08	464.68	10,445.00
	(0.87)	(0.64)	(43.56)	(23.32)
2013	2.11	1.81	58.27	41.02	14.27	443.21	10,084.53
	(0.82)	(0.60)	(37.92)	(23.99)
2014	2.37	1.77	75.68	39.73	12.43	514.60	11,032.25
	(1.14)	(0.61)	(65.35)	(24.61)
2015	2.33	1.74	75.17	39.34	14.29	507.49	9,542.21
	(1.09)	(0.56)	(61.97)	(24.47)
2016	2.31	1.76	83.21	41.37	13.70	685.91	9,270.72
	(1.05)	(0.56)	(74.16)	(27.56)

Standard deviation in parenthesis

Same data focusing on the price and engine displacement dimensions. There are two things worth mentioning here:

Despite average displacement being relatively constant throughout the years, the weighted average price of cars constantly fell (except for 2016).
Even though the number of cars sold was 16% higher (previous table) in 2016 than it was in 2007, the value of total sales in constant pesos fell by almost 5% during the same period.

Important driving force behind this transformation of the market: exhange rate.

2007-2013 relatively constant fluctuating around 2000 COP
but it was over 3000 COP in 2016

Despite these large ER shocks:

Simple average price went from 58 million (2013) to 83 million (2016)
But weighted average car remained constant.

Implying that consumers substitution plays an important role.

The prices of dollars and cars

Between early 2014 and mid 2015 the USD went from 1800 COP to 3000 COP
Between 2013 and 2016 the average car went
- from 60 to 83 million COP (simple average).
- from 41.02 to 41.37 million COP (weighted average).
What causes that “discrepancy”?

Pass-through regressions

Fized effects specification

The basic specification we estimate is

\[ p_{jt}=\alpha_{0j}+\alpha_{1j} ER_t+\alpha_{2j}ER_{t-1}+u_{jt} \]

\(p_{jt}\): price of model \(j\) at time \(t\)
\(ER_t\): nominal COP/USD exchange rate at time \(t\)

Pass-through regressions


	Price

	(1)	(2)	(3)	(4)	(5)

\(ER_t\)	\(-\)0.075	\(-\)0.050	0.079	\(-\)0.074	\(-\)0.084
	(0.080)	(0.086)	(0.103)	(0.080)	(0.082)

\(ER_{t-1}\)	0.124\(^{***}\)	0.115\(^{***}\)	0.236\(^{***}\)	0.125\(^{***}\)	0.119\(^{***}\)
	(0.037)	(0.039)	(0.046)	(0.037)	(0.039)

\(BER_t\)		\(-\)0.246	0.812
		(0.334)	(0.657)

\(ER_t\times BER_t\)			\(-\)0.386
			(0.542)

\(ER_{t-1}\times BER_t\)			\(-\)0.998\(^{***}\)
			(0.300)

\(Share\)				0.040	\(-\)0.315
				(0.053)	(0.256)

\(ER_t\times Share\)					0.188
					(0.270)

\(ER_{t-1}\times Share\)					0.186
					(0.245)

Observations	2,894	2,894	2,894	2,894	2,894
R\(^{2}\)	0.016	0.016	0.059	0.016	0.019

Note: All regressions include fixed effect of car			\(^{}\)p\(<\)0.1; \(^{}\)p\(<\)0.05; \(^{**}\)p\(<\)0.01

All variables are standardized.
Column 1 no interactions, columns 2 and interaction with BER, columns 4 and 5 interaction with share.
Column 1: pass-through takes at least a year. An increase of one standard deviation in the exchange rate, leads to a .12 standard deviation increase in the price of cars a year later.
Columns 2-3 the pass-through was lower precisely for the cars imported from countries that experienced a relative real depreciation of their exchange rate with respect to the Colombian peso. Including this interaction increases the effect of the lagged nominal exchange rate.
Columns 4-5 the addition of the market shares to the regression does not have any effect on the estimated pass-through. In other words, there is no prima-facie evidence that the pass-through was lower for cars that became more popular after the depreciation.
These descriptive results show that price adjustment after the depreciation of 2014 was correlated with cost shifters, but not with crude demand shifters. This varation in the data will therefore allow us to estimate a structural model of demand and pricing that we can use to simulate counterfactual equilibria. We describe the model and its estimation in the following section.

A model for cars

The model

Utility
- \(u_{ijt}=\delta_{jt}(x_{jt},p_{jt},\xi_{jt})+\mu_{ijt}(x_{jt},p_{jt},\epsilon_i)+\varepsilon_{ijt}\)
Demand
- \(s_{jt}=s(x_{jt},p_{jt},\xi_{jt}) \equiv \int \text{Prob}(u_{ijt} \ge u_{irt})dF(\mu_{ijt},\varepsilon_{ijt})\)
- \(S_t=S(X_t,\Xi_t,P_t)\)
Pricing
- \(p_{j\in \Im_{bt}}^*=\arg\max \sum_{j \in \Im_{bt}}(p^*_{jt}-mc_{jt})s_{jt}M_{t}\)
- \(P_{t}^*=MC_t+\Upsilon(X_t,\Xi_t,dS_t/dP_t')\)

We are interested in understanding the adjustment of automobile prices after the large Colombian depreciation of 2014-2015. In particular, we want to estimate effect that the depreciation had on wholesale prices, markups and consumer behavior. Our strategy is to estimate a structural model of supply and demand, and use the model to simulate counterfactual scenarios to isolate the effect of the different factors that determine the observed retail prices.

So our model assumes that a consumer j’s utility from buying car j t time t is a deviation from the mean utility that all consumers would get from buying that car.

The demand then will be given by the share of consumers that choose ha t time t over any other alternative model r and since the demand for all cars would depend on own and rival price we can write the system determining the optimal prices as a matrix. As I said before, we assume that the firms will choose a price for the differentiated products on their menu so optimal prices should solve this problem.

In matrix for, MC is a model-level vector of marrginal costs, dsdp is a JxJ matrix of price derivatives for cars available at time t. So this vector contains the markups of each model as a function of a car characteristics and the derivative of demand with respect to all prices.

We estimate demand and supply jointly.

Estimation

We parametrize the components of the demand and the marginal cost as

\(\delta(x_{jt},p_{jt},\xi_{jt})=X_{jt}\beta+\xi_{jt}\)
\(\mu(x_{jt},p_{jt})=\sigma_d\epsilon_i^dx_{jt}^d+\alpha(\ln(y_{it})-\ln(p_{jt}))\)
\(\ln mc_{jt}=X^s_{jt}\Gamma+\omega_{jt}\)

The criterion function we optimize is

\(G(\theta)=[\{\xi(\theta_0):\omega(\theta_0)\}'Z]'\Omega[\{\xi(\theta_0):\omega(\theta_0)\}'Z]\)

Estimates

Estimates
	Coefficients	Std. Err.	t	p-value

Demand

\(\alpha\)	10,239	0,000	20575,14	0,000
CC	6,430	0,001	8882,891	0,000
Doors	-6,920	0,005	-1430,219	0,000
\(\sigma_{do}\)	5,058	0,003	1953,544	0,000
Air	-0,565	0,000	-3456,523	0,000

Supply

CC	0,502	0,000	135243,48	0,000
Doors	-0,006	0,000	-1873,482	0,000
BER	0,073	0,000	53691,355	0,000
Air	-0,029	0,000	-254459,06	0,000

Note: Estimation includes model, year, country of origin and type of vehicle fixed effects too. The standard errors for the coefficients of the non-linear part of the specification, \(\alpha\), \(\sigma_{cc}\) y \(\sigma_{do}\) are obtained from the variance-covariance matrix of the GMM and the standard errors for the estimates on displacement (CC), doors and AC are obtained via bootstrapping.

Goodness-of-fit

Baseline prices

Baseline markups

Baseline shares

Baseline engine size

The effect of traded costs

We start by obtaining counterfactual marginal costs

\[\begin{split} \ln mc_{jt}&= \gamma_{cc}x^{cc}_{jt}+\gamma_{doors}x^{doors}_{jt}+\gamma_{model}x^{model}_{jt}+\gamma_{AC}x^{AC}_{jt}+\\ &\gamma_{brand}x^{brand}_{jt}+\gamma_{type}x^{type}_{jt} +\color{red}{\gamma^{country}_{jt}} + \omega_{jt} \end{split}\]
We use the new marginal costs to obtain optimal prices and shares

Direct effect of exchange rate

So the result of doing that is shown here. The solid lines are the average prices of the baseline scenario. The blue one uses the shares of cars available in 2013 as weights while the red one uses the shares of each year. The dashed lines are the counterfactual prices.

There are three dashed lines:

Blue is the average price of cars calculated using the shares of only cars available in 2013.
Red is with contemporary shares.
Teal line forces the firms to keep the same markups (but allow consumers to decide). We interpret the fact that the blue and teal are so similar as the firms absorbing most of the external shock.

Regardless of the weights used to compute the average prices, fixing the traded component of the cost of selling cars at its 2013 level, implies a cheaper average car than under the baseline scenario in which the exchange rate is allowed to vary.

Car origin

With this figure we want to gain an insights into the differential effects that exchange rate shocks have on the price of cars from different origins.

The vertical axis: price of cars when the external costs are frozen in their 2013 levels
The horizontal axis we measure the baseline price of cars when the traded costs are allowed to vary after 2013.

Each dot corresponds to a type of car defined by the pair brand-displacement. Notice that the slope of the line connecting each dot to the origin is a measure of the product-specific pass-through. The steeper the slope the smaller the pass-through.

For instance, for European cars, the pass-through depends on the segment of the car. Relatively cheaper European cars exhibit little pass-through when compared to more expensive European cars. In other words, had the peso not depreciated with respect to the Euro, expensive cars from Germany or France would have become relatively more expensive than cheaper Euro cars. Something similar applies to Asian cars. Cars from Latin America, are bunched close to the 45 degree line which means that the depreciation did not change their prices much, consistent with the peso exhibiting a much smaller depreciation wrt those currencies. Analogously we see a much larger distance from that line for american cars as expected because the depreciation wrt the USD was much larger.

All this shows the many ways in which consumers can substitute when faced with a exchange rate shock.

The menus

Concluding remarks

Empirical exercise to advance our knowledge about the mechanisms and causes of incomplete exchange rate pass-through.
Firms react to increases in traded costs by:
- changing the origin of intermediate goods
- altering their offerings
- changing prices in reaction to their rivals
Consumers substitute away from expensive cars.
No effect of aggregate income.

Thanks

References

Amiti, Mary, Itskhoki Oleg, and Jozef Konings. 2014. “Importers, exporters, and exchange rate disconnect.” American Economic Review 104 (7): 1942–78.

———. 2019. “International shocks, variable markups, and domestic prices.” Review of Economic Studies 86 (6): 2356–2402.

Atkeson, Andrew, and Ariel Burstein. 2008. “Pricing-to-Market, Trade Costs, and International Relative Prices.” American Economic Review 98 (5): 1998–2031.

Auer, Raphael A, and Raphael S Schoenle. 2016. “Market structure and exchange rate pass-through.” Journal of International Economics 98: 60–77.

Auer, Raphael, Burstein Ariel, and Sarah M Lein. 2021. “Exchange rates and prices: evidence from the 2015 Swiss franc appreciation.” American Economic Review 111 (2): 652–86.

Bergin, Paul R, and Robert C Feenstra. 2001. “Pricing-to-Market, Staggered Contracts, and Real Exchange Rate Persistence.” Journal of International Economics 54 (2): 333–59.

Burstein, Ariel T, Joao C Neves, and Sergio Rebelo. 2003. “Distribution Costs and Real Exchange Rate Dynamics During Exchange-Rate-Based Stabilizations.” Journal of Monetary Economics 50 (6): 1189–1214.

Corsetti, Giancarlo, and Luca Dedola. 2005. “A Macroeconomic Model of International Price Discrimination.” Journal of International Economics 67 (1): 129–55.

De Loecker, Jan, Goldberg Pinelopi K, Khandelwal Amit K, and Nina Pavcnik. 2016. “Prices, markups, and trade reform.” Econometrica 84 (2): 445–510.

Dornbusch, Rudiger et al. 1987. “Exchange Rates and Prices.” American Economic Review 77 (1): 93–106.

Fabra, Natalia, and Mar Reguant. 2014. “Pass-through of emissions costs in electricity markets.” American Economic Review 104 (9): 2872–99.

Goldberg, Linda S, and José Manuel Campa. 2010. “The Sensitivity of the CPI to Exchange Rates: Distribution Margins, Imported Inputs, and Trade Exposure.” The Review of Economics and Statistics 92 (2): 392–407.

Gopinath, Gita, Gourinchas Pierre-Olivier, Hsieh Chang-Tai, and Nicholas Li. 2011. “International prices, costs, and markup differences.” American Economic Review 101 (6): 2450–86.

Knetter, Michael M. 1989. “Price Discrimination by US and German Exporters.” The American Economic Review 79 (1): 198–210.

Miller, Nathan H, Osborne Matthew, and Gloria Sheu. 2017. “Pass-through in a concentrated industry: empirical evidence and regulatory implications.” RAND Journal of Economics 48 (1): 69–93.

Muehlegger, Erich, and Richard L Sweeney. 2021. “Pass-Through of Own and Rival Cost Shocks: Evidence from the US Fracking Boom.” The Review of Economics and Statistics, 1–32.

Nakamura, Emi, and Dawit Zerom. 2010. “Accounting for incomplete pass-through.” Review of Economic Studies 77 (3): 1192–1230.

Parsley, David C, and Shang-Jin Wei. 2001. “Explaining the Border Effect: The Role of Exchange Rate Variability, Shipping Costs, and Geography.” Journal of International Economics 55 (1): 87–105.

Sanyal, Kalyan K, and Ronald W Jones. 1982. “The Theory of Trade in Middle Products.” The American Economic Review 72 (1): 16–31.