The Effect of ADR Listing on Underlying Equity Behavior
First Years Asset Management
Rebecca Mazza
Alexandre Rapaport
Marcelo Rosenburg
Massimiliano Rossi
Jose Zapata
Contents
Overview
Data
Methodology
Hypothesis 1
Hypothesis 2
Hypothesis 3
Empirical Results
Models 1 and 2
Models 3 and 4
Models 5 and 6
Volume Analysis
Volatility Analysis
Conclusions
Index of Tables
Table 1: Selected Companies
Table 2: Model 1 Output
Table 3: Model 2 Output
Table 4: Model 3 Output
Table 5: Model 4 Output
Table 6: Model 5 and Model 6 Output
Table of Figures
Figure 1: MSCI Local Index Coefficient
Figure 2: Sensitivity to MSCI Local Index
Figure 3: Sensitivity to MSCI U.S. Index
Figure 4: Sensitivity to Foreign Exchange Rate
1
Overview
In this paper, we examine the behavior of foreign equities that are cross-listed on a U.S. exchange as American Depositary Receipts (ADRs). When forecasting future equity returns using traditional asset pricing theories, it is unclear whether to expect that cross-listed equities’ behavior will more closely reflect the events in their local market or the U.S. market. Across a sample of seventeen cross-listed, large-cap companies from major industries in developed markets, we contrast the behavior of the underlying foreign equities before and after the ADR listing date to determine whether foreign equity behavior changes after being listed as an ADR and whether sensitivity to local market and U.S. market risk factors shifts.
Data
From the universe of companies that have an ADR program, we elected to focus on companies from three major industries: financial services, automotive and telecommunications. Within those industries, we selected companies with a large market capitalization from developed markets. From information available on Yahoo! Finance, we determined the ADR listing date for each company. From Datastream, we obtained monthly data on the split and dividend adjusted prices in local currency and volume traded of the underlying equities for each company. We eliminated all companies for which there were not at least sixty months of data available both before and after the ADR listing date. We then obtained monthly data on the foreign exchange rate for each currency represented in the selection of companies. From Morgan Stanley Capital International, we obtained the MSCI Index in terms of U.S. dollars for the U.S. and for the local market of each country represented in the selection of companies. The companies included in the final analysis are listed in Table 1.
Table 1: Selected Companies
Company / Ticker (ADR) / Industry / CountryABN Amro / ABN / Financial Services / Netherlands
Alcatel / ALA / Telecommunications / France
Bank of Tokyo / MBK / Financial Services / Japan
Deutsche Bank / DTBKY / Financial Services / Germany
Dresdner Bank / DRSDY / Financial Services / Germany
Ericsson / ERICY / Telecommunications / Sweden
Fiat / FIA / Automotive / Italy
Honda / HMC / Automotive / Japan
ING / ING / Financial Services / Netherlands
Nissan / NSANY / Automotive / Japan
Nokia / NOK / Telecommunications / Finland
NTT / NTT / Telecommunications / Japan
Peugeot / PEUGY / Automotive / France
Telecom Italia / TI / Telecommunications / Italy
Toyota / TM / Automotive / Japan
Volkswagon / VLKAY / Automotive / Germany
Volvo / VOLVY / Automotive / Sweden
For each company, we translated the underlying equity price into U.S. dollar terms using the foreign exchange rate for the local market currency, in order for the returns to be comparable to the MSCI country index, which is also in terms of U.S. dollars. We calculated the monthly returns on the underlying stock in U.S. dollar terms for each company and the relevant MSCI country index. We also calculated the monthly percent change in the foreign exchange rate for each currency.
We created a final set of data for each company that includes the time series of the returns on the underlying equity in terms of U.S. dollars for 121 months (sixty months before the month in which the ADR was listed, the ADR listing month and sixty months after the ADR was listed) and the corresponding time series’ of returns on the MSCI local country index, the MSCI U.S. index and the change in the foreign exchange rate. We created one final database that includes the full data set for all of the companies.
Methodology
We set out to test three main hypotheses:
- There is a premium to the return on the underlying equity in the month that it is first cross-listed as an ADR.
- Sensitivity to local market factors decreases after the ADR listing date, while sensitivity to U.S. market factors increases.
- There is increased liquidity in the underlying stock after the ADR listing date; however, it is accompanied by higher volatility.
Hypothesis 1
To test the first hypothesis, we created a dummy variable, which we set equal to one in the month of the ADR listing and zero otherwise. For each company, we estimated the coefficients in Model 1:
Model 1
ri,t = 0 + i,LrL,t + i,USrUS,t + i,XrX,t + i,ddt + t
where ri,t is the return on the underlying equity, rL,t is the return on the MSCI local index, rUS,t is the return on the MSCI U.S. index, rX,t is the change in the foreign exchange rate and dt is the dummy variable. The return on the MSCI local index is orthogonalized to remove the portion that is dependent on the MSCI U.S. index, eliminating the problem of multicollinearity. The coefficient of the dummy variable is designed to indicate whether the return in the month of the ADR listing is different than the average return over the rest of the sample period. The other risk factors are included in the model in order to segregate their effect on the returns on the underlying stock from the effect of the ADR listing event.
We also estimated Model 2:
Model 2
ri,t = 0 + i,LrL,t + i,USrUS,t + i,XrX,t + t
where the dependent and independent variables are as in Model 1, excluding the dummy variable, in order to verify whether the dummy variable increased the explanatory power of the model.
Hypothesis 2
To test the second hypothesis, we divided the data set for each company into separate time series’ for before and after the ADR listing. We defined the month in which the ADR was listed as time t0, the before period as t0-1 through t0-60 and the after period as t0+1 through t0+60. We estimated the coefficients in Model 3 and Model 4:
Model 3
rib,t = 0 + i,LbrLb,t + i,USbrUSb,t + i,XbrXb,t + t
Model 4
ria,t = 0 + i,LarLa,t + i,USarUSa,t + i,XarXa,t + t
where, as before, ri,t is the return on the underlying stock, rL,t is the return on the MSCI local index (orthogonalized), rUS,t is the return on the MSCI U.S. index and rX,t is the change in the foreign exchange rate. The b and a subscripts refer to the before and after period, respectively. The coefficient of each of the risk factors indicates the degree of sensitivity of the returns on the underlying equity to movements in each risk factor. Comparing the coefficient for each risk factor in the before and after period indicates whether there was a change in the sensitivity of the returns on the underlying stock of each company to the risk factors after the ADR listing event.
To examine whether there was a systematic change across the sample set of companies in the relationship between the estimated coefficients and the expected returns after the ADR listing, we estimated Model 5 and Model 6:
Model 5
ri,AVGb = 0 + Lbhati,Lb + USbhati,USb + Xbhati,Xb + t
Model 6
ri,AVGa = 0 + Lahati,La + USahati,USa + Xahati,Xa + t
where ri,AVG is the average return for each company and hati,L, hati,US, and hati,X, are the coefficients of the risk factors estimated in Model 3 and Model 4. The b and a subscripts refer to the before and after periods, respectively. The coefficients on the hat factors indicate how well the risk factors included in the model explain the returns on the underlying equities. A comparison of the slope before and after the ADR listing date indicates whether the sensitivity to each of the risk factors changed after the ADR listing date.
Hypothesis 3
To test the third hypothesis, we plotted the trading volume of the underlying equity on its local market against time, to determine whether there was a systematic change in the volume traded after the ADR listing date. We also compared the average volume traded before and after the ADR listing date, again, to determine whether there was a systematic change in the volume traded after the ADR listing. An increase in the volume traded indicates increased liquidity in the underlying equity.
Similarly, we calculated the twelve-month rolling standard deviation of returns on the underlying equity and plotted it against time, to determine whether there was a systematic change in the volatility of returns after the ADR listing date. We also compared the standard deviation of returns on the underlying equity in the before and after period, again, to determine whether there was a systematic change in the volatility of returns after the ADR listing.
Empirical Results
In all of the models, we elected to use a multivariate asset pricing framework, based on the prior research of Patro[1] (1999), which suggests that the univariate international Capital Asset Pricing Model (CAPM), where the market portfolio is the world market portfolio, does not sufficiently explain average returns. The risk factors we included reflect our a priori expectation that the underlying equities are subject to fluctuations in the local and U.S. markets, and perhaps the foreign exchange rate.
Models 1 and 2
When we estimated Model 1, the coefficient of the dummy variable for all but two companies was statistically insignificant. There was also no clear trend in the direction of the dummy variable coefficient – it was positive only half the time. This suggests that there was not a premium on the return on the underlying equity at the ADR listing date across our sample companies. The coefficients on each of the risk factors were statistically significant in the majority of cases, although the sign of the coefficients varies across the sample. Table2 summarizes the coefficients, t-ratios, and R2 estimated by Model 1 for each company.
Table 2: Model 1 Output
i,d / t-ratio / i,US / t-ratio / i,L / t-ratio / i,X / t-ratio / R2ABN
/ 0.01 / 0.10 / -0.44 / -2.33 / 0.16 / 0.79 / 0.52 / 1.74 / 0.09ALA / -0.02 / -0.20 / -0.04 / -0.18 / 0.26 / 1.46 / 0.45 / 1.82 / 0.05
DRSDY / 0.04 / 0.43 / -0.70 / -3.48 / -0.01 / -0.07 / -0.42 / -1.54 / 0.13
DTBKY / -0.01 / -0.12 / -0.33 / -1.69 / -0.04 / -0.20 / -0.62 / -2.36 / 0.08
ERICY / 0.26 / 2.21 / -0.03 / -0.12 / -0.26 / -1.51 / -0.37 / -0.96 / 0.07
FIA / -0.03 / -0.26 / 0.24 / 1.07 / 0.25 / 1.79 / -0.83 / -2.75 / 0.14
HMC / 0.07 / 0.79 / -0.20 / -1.14 / -0.26 / -1.86 / -0.76 / -2.41 / 0.08
ING / -0.01 / -0.19 / -0.31 / -1.76 / 0.37 / 1.82 / -0.22 / -0.88 / 0.08
MBK / 0.03 / 0.29 / 0.06 / 0.30 / 0.10 / 0.72 / -1.08 / -3.40 / 0.17
NOK / 0.01 / 0.10 / -0.12 / -0.51 / 1.32 / 11.86 / 0.44 / 1.40 / 0.58
NSANY / 0.07 / 0.79 / -0.06 / -0.30 / 0.38 / 3.42 / 0.19 / 0.85 / 0.10
NTT / 0.00 / -0.04 / -0.16 / -0.66 / -0.11 / -0.84 / -0.98 / -3.65 / 0.11
PEUGY / -0.05 / -0.55 / -0.38 / -1.82 / 0.42 / 2.09 / -0.91 / -3.24 / 0.19
TI / 0.00 / -0.02 / 0.63 / 3.15 / 0.98 / 8.44 / -0.55 / -2.27 / 0.40
TM / 0.11 / 1.64 / -0.35 / -2.06 / -0.04 / -0.34 / -0.91 / -4.05 / 0.23
VLKAY / 0.07 / 0.75 / -0.25 / -1.09 / 0.10 / 0.45 / -0.35 / -1.11 / 0.03
VOLVY / -0.02 / -0.22 / -0.03 / -0.15 / 0.12 / 0.86 / 0.09 / 0.28 / 0.01
When we estimated the model excluding the dummy variable, the R2 was slightly lower in each case and the coefficients on the remaining risk factors were slightly different. Again, the coefficients on each of the risk factors were statistically significant in the majority of cases, although the sign on the coefficients varies across the sample. Table 3 summarizes the coefficients, t-ratios, and R2 estimated by Model 2 for each company.
Table 3: Model 2 Output
i,US / t-ratio / i,L / t-ratio / i,X / t-ratio / R2ABN
/ -0.44 / -2.35 / 0.16 / 0.80 / 0.52 / 1.75 / 0.09ALA / -0.03 / -0.17 / 0.26 / 1.45 / 0.45 / 1.87 / 0.05
DRSDY / -0.69 / -3.47 / -0.01 / -0.08 / -0.41 / -1.54 / 0.13
DTBKY / -0.33 / -1.74 / -0.04 / -0.20 / -0.62 / -2.37 / 0.08
ERICY / -0.04 / -0.17 / -0.31 / -1.76 / -0.34 / -0.85 / 0.03
FIA / 0.24 / 1.08 / 0.25 / 1.78 / -0.83 / -2.76 / 0.14
HMC / -0.19 / -1.09 / -0.25 / -1.82 / -0.78 / -2.49 / 0.08
ING / -0.32 / -1.81 / 0.37 / 1.82 / -0.22 / -0.89 / 0.08
MBK / 0.05 / 0.26 / 0.10 / 0.72 / -1.08 / -3.41 / 0.17
NOK / -0.12 / -0.51 / 1.32 / 11.97 / 0.44 / 1.40 / 0.58
NSANY / -0.09 / -0.41 / 0.38 / 3.42 / 0.19 / 0.85 / 0.10
NTT / -0.16 / -0.66 / -0.11 / -0.85 / -0.98 / -3.69 / 0.11
PEUGY / -0.39 / -1.93 / 0.42 / 2.09 / -0.92 / -3.29 / 0.19
TI / 0.63 / 3.17 / 0.98 / 8.49 / -0.55 / -2.30 / 0.40
TM / -0.39 / -2.24 / 0.01 / 0.08 / -0.88 / -3.90 / 0.21
VLKAY / -0.23 / -0.99 / 0.09 / 0.43 / -0.34 / -1.10 / 0.03
VOLVY / -0.03 / -0.16 / 0.13 / 0.89 / 0.09 / 0.27 / 0.01
Models 3 and 4
In the estimation of Model 3 (for the before period), the coefficient on the MSCI U.S. index was statistically significant for only four of the seventeen companies, while the coefficients on the MSCI local index and the foreign exchange rate were significant for ten and thirteen companies, respectively. Table 4 summarizes the coefficients, t-ratios, and R2 estimated by Model 3 for each company.
Table 4: Model 3 Output
i,USb / t-ratio / i,Lb / t-ratio / i,Xb / t-ratio / R2ABN
/ 0.18 / 0.73 / 0.33 / 1.38 / 0.43 / 1.52 / 0.08ALA / -0.04 / -0.19 / 0.45 / 1.75 / 0.26 / 0.77 / 0.07
DRSDY / -0.09 / -0.42 / -0.06 / -0.36 / -0.70 / -3.58 / 0.20
DTBKY / -0.03 / -0.14 / -0.05 / -0.27 / -0.78 / -3.70 / 0.21
ERICY / 0.04 / 0.14 / -0.10 / -0.45 / -1.17 / -2.01 / 0.07
FIA / 0.01 / 0.02 / 0.57 / 2.98 / -0.59 / -1.26 / 0.24
HMC / -0.38 / -1.40 / -0.46 / -2.11 / -0.97 / -2.01 / 0.14
ING / 0.42 / 1.85 / 0.23 / 0.93 / -0.31 / -1.33 / 0.14
MBK / 0.00 / -0.02 / 0.25 / 1.06 / -0.79 / -1.94 / 0.20
NOK / -0.01 / -0.02 / 1.36 / 7.64 / 0.33 / 0.83 / 0.57
NSANY / 0.26 / 0.88 / 0.22 / 1.82 / -0.17 / -0.47 / 0.08
NTT / -0.08 / -0.16 / -0.17 / -0.90 / -1.26 / -2.22 / 0.09
PEUGY / -0.10 / -0.30 / 0.58 / 2.26 / -0.82 / -2.56 / 0.22
TI / 0.51 / 1.73 / 1.02 / 7.75 / -0.48 / -1.98 / 0.52
TM / -0.16 / -0.79 / 0.10 / 0.82 / -0.47 / -1.55 / 0.13
VLKAY / -0.38 / -1.06 / 0.12 / 0.41 / -0.61 / -1.82 / 0.09
VOLVY / 0.04 / 0.19 / 0.23 / 1.11 / 0.19 / 0.33 / 0.02
When we estimated Model 4 (for the after period), the incidence of significant coefficients of the MSCI U.S. index rose from four to seven companies, while the number of cases in which the coefficient on the MSCI local index was significant fell from ten to seven. Likewise, the coefficient on the foreign exchange rate was significant in thirteen companies before and only nine after the ADR listing. Although this seems to support our hypothesis that the sensitivity of the returns on the underlying equity to local market factors decreases after the ADR listing, while the sensitivity to U.S. market factors increases, the trend in the data is not uniform enough to be conclusive. Table 5 summarizes the coefficients, t-ratios, and R2 estimated by Model 4 for each company.
Table 5: Model 4 Output
i,USa / t-ratio / i,La / t-ratio / i,Xa / t-ratio / R2ABN
/ -0.68 / -2.38 / -0.01 / -0.02 / 0.78 / 1.33 / 0.15ALA / 0.10 / 0.23 / 0.06 / 0.22 / 0.62 / 1.64 / 0.05
DRSDY / -1.00 / -3.09 / 0.06 / 0.19 / 0.07 / 0.12 / 0.18
DTBKY / -0.48 / -1.52 / -0.02 / -0.05 / -0.41 / -0.72 / 0.07
ERICY / -0.17 / -0.34 / -0.43 / -1.56 / 0.06 / 0.11 / 0.05
FIA / 0.60 / 0.00 / -0.08 / 0.00 / -0.94 / 0.00 / 0.13
HMC / -0.18 / -0.76 / -0.23 / -1.25 / -0.70 / -1.67 / 0.07
ING / -0.59 / -2.21 / 0.41 / 1.33 / -0.08 / -0.16 / 0.14
MBK / 0.18 / 0.49 / -0.06 / -0.34 / -1.55 / -2.77 / 0.15
NOK / -0.09 / -0.33 / 1.32 / 9.09 / 0.84 / 1.46 / 0.60
NSANY / -0.26 / -0.84 / 0.73 / 3.22 / 0.22 / 0.71 / 0.18
NTT / -0.22 / -0.87 / 0.02 / 0.10 / -0.97 / -3.44 / 0.18
PEUGY / -0.48 / -1.71 / 0.22 / 0.69 / -1.11 / -2.16 / 0.20
TI / 0.71 / 2.41 / 0.89 / 4.29 / -0.77 / -1.34 / 0.29
TM / -0.71 / -2.59 / -0.22 / -1.25 / -1.32 / -4.22 / 0.34
VLKAY / -0.25 / -0.73 / 0.05 / 0.16 / 0.03 / 0.05 / 0.01
VOLVY / -0.14 / -0.34 / 0.06 / 0.26 / 0.08 / 0.19 / 0.00
We further contrasted the coefficients estimated before and after the ADR listing and found that in fourteen cases, the MSCI local index coefficient decreased after the ADR listing.[2] However, there was not a clear trend across the other risk factors; the MSCI U.S. index coefficient increased and the foreign exchange coefficient decreased for only about half of the companies. Figure 1 depicts the contrast of the MSCI local index coefficients before and after the ADR listing.
Figure 1: MSCI Local Index Coefficient
Models 5 and 6
Using the coefficients estimated in Model 3 and Model 4, we performed the cross-sectional regressions indicated in Model 5 and Model 6; Table 6 summarizes the output.
Table 6: Model 5 and Model 6 Output
Before / t-ratio / After / t-ratioUS / -0.023 / -1.232 / -0.004 / -0.527
L / 0.023 / 2.069 / 0.022 / 2.668
X / 0.001 / 0.108 / 0.000 / -0.058
R2 / 0.154 / 0.253
We found that the sensitivity of the returns on the underlying equities to each of the risk factors changed only slightly across our sample after the ADR listing; they did not follow a pattern consistent with our hypothesis. In addition, the data do not display a strong relationship between the size of the estimated coefficients and the average returns, suggesting that fluctuations in the risk factors do not entirely explain average returns on the underlying equities. This is more apparent in Figure 2, Figure 3 and Figure 4, which depict the sensitivity of the returns on the underlying equities to movements in the MSCI local index, the MSCI U.S. index and the foreign exchange rate, respectively, before and after the ADR listing.
Figure 2: Sensitivity to MSCI Local Index
Figure 3: Sensitivity to MSCI U.S. Index
Figure 4: Sensitivity to Foreign Exchange Rate
Volume Analysis
From the graphs that we plotted, we did not identify any single trend in the volume traded of the underlying equities on the local exchange after the ADR listing date. Because volume traded seemed to mirror the volume of the market as a whole, it is unclear whether changes near the ADR listing date are related to that event or broader market factors.
Contrasting the average volume traded over the course of the period before and after the ADR listing, we found that in twelve of the companies, volume increased after the ADR listing date.
Volatility Analysis
From the graphs that we plotted, we did not identify any single trend in the volatility of returns on the underlying equities after the ADR listing date. Because volatility seemed to mirror the volatility of the market as a whole, it is unclear whether changes near the ADR listing date are related to that event or broader market factors.
Contrasting the standard deviation of returns over the course of the period before and after the ADR listing, we found that in eleven of the companies, volatility increased after the ADR listing date.
Conclusions
Hypothesis 1: False
- Our data do not indicate that there was a premium on the return on the underlying equity at the ADR listing date across our sample companies.
Hypothesis 2: Inconclusive
- Although the data suggest that our hypothesis is valid - that the sensitivity of the returns on the underlying equity to local market factors decreases after the ADR listing, while the sensitivity to U.S. market factors increases - the trend in the data is not uniform enough to be conclusive.
- In addition, the data do not display a strong relationship between the size of the estimated coefficients and the average returns, suggesting that fluctuations in the risk factors do not entirely explain average returns on the underlying equities.
Hypothesis 3: Inconclusive
- It is unclear whether changes in the volume traded and volatility of returns on the underlying equities near the ADR listing date are related to that event or broader market factors.
Although we cannot confirm any of our hypotheses, because our data only span a small sample set of companies, we cannot draw any firm conclusions. It is probable that if this study were expanded to encompass a larger cross-section of the ADR market, it would likely result in very different conclusions. It is also likely that if the data included small cap companies or developing markets, the conclusions would also be very different.
[1] Patro, Dilip Kumar, “Return Behavior and Pricing of American Depositary Receipts” Journal of International Financial Markets, Institutions and Money, January 1999, p. 43-67.
[2] If the coefficient was negative in the before model and became more negative in the after model, it was counted as an increase. If the coefficient was negative in the before model and became less negative in the after model, it was counted as a decrease. If the coefficient was negative in the before model and became positive in the after model, it was counted as a decrease. If the coefficient was positive in the before model and became more positive in the after model, it was counted as an increase. If the coefficient was positive in the before model and became less positive, it was counted as a decrease. If the coefficient was positive in the before model and became negative, it was counted as a decrease.