Canadian Commodity Prices and Exchange Rate Dynamics: Evidence of Structural Change1. IntroductionCanada is a small open economy and a large exporter of goods, services, and commodities. Coined as a commodity currency in 2003 by a paper written by Yu-chin Chen and Kenneth Rogoff, Canada has long held a comparative advantage in the distribution of natural resources and commodities. Due to the dependence of output growth in Canada on exports of natural resources and commodities to the world market, the Canadian dollar has long been considered to be representative of the state of production and exporting of commodities in Canada. Ultimately when the price of commodity exports rises, often due to exogenous shocks on both the demand and supply side in the global economy, the Canadian dollar appreciates as net exports increase and Canada becomes a more attractive destination for investment.It has not always been the case, however, that the exchange rate between the Canadian dollar and the U.S. dollar were closely related to commodity prices. According to the report Intervention in the Foreign Exchange Market by the Bank of Canada, before September 1998 the Bank practised a policy of systematic intervention in the foreign exchange market to peg the value of the Canadian dollar close to the U.S. dollar. (Bank of Canada, 1) There were periods of time when a rise or fall in the value of the Canadian dollar was not matched with a corresponding change in commodity prices. By using intervention in the foreign exchange market, the Bank tried to keep the value of our dollar close to par with the U.S. dollar. In pursuit of autonomous monetary policy at the expense of a degree of control over the value of the Canadian dollar, the bank has not systematically intervened in the foreign exchange market since before the policy change. (Bank of Canada, 2)The objective of this paper is not to assign a normative value statement to whether inflation targeting or control over exchange rates is more desirable for an independent central bank. Instead, I aim to demonstrate that exogenous market forces have a stronger impact on the value of an economys currency especially for a large exporter such as Canada. First, I will provide background on relevant literature and topics of interest in relation to my empirical analysis. Next, I will define my formal hypotheses, empirical methodology and data, along with a consideration to improve future research on this topic. Following up, I will demonstrate the results and interpretation of the empirical tests. Finally, I will look at a current monetary policy implication in relation to current commodity prices and the Bank of Canadas projected overnight rate increase. From the analysis, we will see empirical evidence of a structural difference in the relationship between the USD/CAD exchange rate and commodity prices before and after the foreign exchange intervention policy change made by the Bank of Canada in September 1998. 2. Background & LiteratureInitially, I was drawn to the topic of the relationship between prices of commodities and goods with exchange rates by the paper: Exchange Rate Pass Through to Consumer Prices: Theory and Recent Evidence by Laurence Savoi-Chabot and Mikael Khan of the Bank of Canada. They demonstrate that exchange rate pass through, defined as the response in appreciation in prices of goods to a relative depreciation of an economys currency, has the largest effect in the Canadian energy sector as shown in the bar graph on the right. Generalizing to commodities, Savoi-Chabot and Khan make reference to the fact that Canada is a commodity currency, along with the relationship between commodities and inflation. Looking at changes in commodity prices impact on inflation and exchange rates rather than how exchange rates impact commodity prices, they note that when depreciation is caused by lower commodity prices, it will lower inflation because the lower levels of commodity prices affect commodity intensive consumer prices such as gasoline. (Savoi-Chabot and Khan, 3) (Savoi-Chabot and Khan, 4)Looking further into the dynamics of commodity prices and exchange rates,the paper: Power of Commodity Prices Over Exchange Rate Dynamics written by Nicholas Garmulewicz demonstrates the idea that commodity prices have an influence on the Canadian dollar relative to other currencies. It also initially provides reasoning for a structural change in dynamics between commodities and the Canadian dollar before and after the Bank of Canada changed their foreign exchange intervention policy from systematic to discretionary in September 1998. (Garmulewicz, 16) This paper begins by analyzing how commodity prices have an effect on the value of the CAD, and then switches to test whether the CAD contains information about the future prices of commodities.Due to the fact that commodity prices and exchange rates are considered to be very difficult to forecast, the lens of this paper will conduct a type of analysis on the former topic, rather than the later. Before defining the formal hypotheses from which I will conduct my empirical analysis around, I would like to provide some essential background information on the Bank of Canada foreign exchange policy change, the Bank Commodity Price Index (BCPI), and history of the USD/CAD exchange rate with the BCPI. Bank of Canada:The Bank of Canada has a broad, principle role from which all of the Banks activities stem from. As the nations central bank, it is responsible to promote the economic and financial welfare in Canada, through the four areas of monetary policy, the financial system, currency, and funds management. (Bank of Canada Act, 1) Seemingly a clear objective at first thought, this objective has is open for interpretation. Over time, economists have grappled with what the promotion of economic and financial welfare specifically entails. These debates around the guiding principal have influenced the activities of the bank since its establishment in 1935. At this point in time, it is commonly known that the bank tries to achieve 2% annual inflation as usually measured by the total consumer price index (CPI). The main tool used by the bank of Canada while conducting monetary policy to target 2% inflation is by setting the overnight rate. The overnight rate, as defined by the Bank, is the interest rate at which major financial institutions borrow and lend one-day (or overnight) funds among themselves. (Monetary Policy, 2018)However, targeting 2% inflation is a recent objective that has only been officially agreed upon since the September 1998. Before this point in time, the Bank was concerned with other objectives in order to meet its principal role of promoting economic well-being There are some main reasons behind this chosen target level of inflation. First, it provides the mechanism for participants in the economy to drive their consumption and investment activities while keeping the purchasing power of a dollar from distorting too rapidly through time. People are incentivized to purchase goods and make investments that satisfy their needs or appreciate in value because of inflation. The other reason for a 2% inflation target is that it gives the Bank of Canada room for monetary stimulation before hitting the zero lower bound of the overnight rate.Before inflation targeting was considered to be the leading policy for promotion of economic welfare, the Bank of Canada used systematic intervention in foreign exchange markets to peg the value of the Canadian dollar close to the value of the U.S. dollar as outlined in their foreign exchange intervention article. (Bank of Canada, 1) The mechanism for this intervention was through the bank purchasing Canadian dollars in foreign exchange markets to boost demand for the CAD in an attempt to stimulate the market value of the Canadian dollar. At the same time, the Bank would re-deposit the same amount of money into the Canadian financial system to neutralize upward pressure on interest rates. (Bank of Canada, 2) This policy did have some merits during the course of its use, however, towards the end of the policy, there is evidence that market forces started becoming the dominant factor in US/CAD determination. Even though the bank was attempting to keep the value of the dollar relatively on par with the U.S., the nominal exchange rate had dropped below $0.70 U.S. at the time of the policy change.Ultimately, this shift in foreign exchange policy represents a move on the open economy trilemma. This trilemma is the theory that an open economy must choose between two of the following: free capital flow, exchange rate determination, and autonomous monetary policy. Before the change in policy, the Bank ultimately felt that the best way to achieve their principle role of promotion of economic welfare was through free capital flow and exchange rate determination. Since the Bank has not intervened in the foreign exchange market since before the policy change, autonomous monetary policy has been favoured over exchange rate determination. This now lends itself to the Bank being able to target inflation through the determination of the overnight rate and a floating exchange rate.Bank Commodity Price Index (BCPI):In order to track the relative prices of commodities relevant to activity in the Canadian economy, the Bank of Canada constructed the Bank of Canada Commodity Price Index (BCPI) which started in 1972Q1. This price index tracks the spot prices of twenty-six commodities mainly produced in Canada and sold on world markets. The components of the BCPI can be broken down into energy, metal and minerals, forestry, agriculture, and fisheries. In computing the index value, each component is assigned a weight using input-output data from Statistics Canada. A feature of these weights is that they are time varying. In particular, the energy component of the BCPI has increased in share since the Banks policy change. In 2018, 54% of the index is weighted by energy prices including the crude oils West Texas Intermediate, Western Canadian Select, and Brent along with natural gas and coal. The other 46% is weighted by all the other sectors relevant in Canadian commodities. (Commodity Price Index, 2018)USD/CAD Exchange Rate & BCPI: History has not always shown a positive correlation between BCPI movement and the USD/CAD exchange rate. Using data from the Bank of Canada and the St. Louis Federal Reserve Economic Database, I have plotted the BCPI and the exchange rate from January 1972 to August 2018. It is clear to see that before the policy change was implemented in September 1998, there were examples of the BCPI and the exchange rate not moving in the same direction. This can be due to many different factors, with one of them being that the exchange rate does not contain information about the current growth of the BCPI index. For example, in the late 1980s, we see that the Canadian dollar appreciated relative to the U.S. dollar even though the BCPI did not experience significant growth. After touching a then-record low of US$0.6913 on 4 February 1986, the dollar rebounded following a concerted strategy of aggressive intervention in the foreign exchange market, sharply higher interest rates, and the announcement of large foreign borrowings by the federal government. (Powell, 1999) Moving into the 1990s, the Bank of Canada lost the ability to significantly control the value of the Canadian dollar as the value had dropped to a new low of $0.65 U.S., matched by a simultaneous drop in the BCPI index. Clearly, the exchange rate was now containing more information about the state of the economy, as an example through the BCPI index.Moving into the era of discretionary foreign exchange intervention policy, the co-dependence of the growth in value of the Canadian dollar relative to the U.S. dollar along with growth of the BCPI index is visually compelling through the above graph. One of the most evident examples of this correlation is during the 2014 global crude oil price crash which caused the BCPI to drop over 200 points. At the same time, the CAD continued to drop in value down below $0.80 U.S., representing our dollar containing information about the BCPI.Post policy change, there is evidence of a structural break in the relationship between the exchange rate and the BCPI index with the exchange rate following the BCPI index much more closely after the break. 3. Formal HypothesesThere are two questions of consideration to test with empirical analysis. First, I will determine whether there is evidence of a structural break in the dynamics between the BCPI growth rate and the USD/CAD exchange rate in U.S. dollars before and after the Banks foreign exchange intervention policy change. To consider the idea of a structural change, I define the following hypothesis:H0: There is no structural change in the relationship between US/CAD growth and BCPI growth after the change in foreign exchange intervention policy in September 1998.H1: There is structural change in the relationship between US/CAD growth and BCPI growth after the change in foreign exchange intervention policy.Second, I will determine if there is evidence of a change in the sensitivity of the Banks overnight rate to changes in BCPI growth. In theory, an environment with a higher interest rate should lead to an appreciation to the relevant currency in the economy. Another reason this test is of interest is because after the policy change, the Bank began targeting inflation at 2% through the overnight rate. If higher commodity prices do act as upward pressure on Canadian inflation, the overnight rate might have a stronger positive relationship with BCPI growth. The second hypothesis to test is:H0: There is no structural change in the sensitivity of the Bank of Canada overnight rate to changes in BCPI growth after the change in foreign exchange intervention policy.H1: There is structural change in the sensitivity of the Bank of Canada overnight rate to changes in BCPI growth after the change in foreign exchange intervention policy.Ultimately, the second test for overnight rate sensitivity can be viewed as a potential path for the increase in value of the Canadian dollar relative to the U.S. dollar. 4. Empirical Methodology & DataIn testing my hypotheses, I have gathered data relating to the hypothesis tests of interest and look to employ it to test for two structural breaks. The time range of the data is from January 1972 to August 2018 with observations occurring at a monthly frequency. This yields 560 observations in the set for analysis. The variables of choice are defined as the following:Monthly USD/CAD growth rate: ln?USDCADtln?(USDCADt1)Monthly BCPI growth rate: ln?BCPItln?(BCPIt1)Monthly Canadian core CPI growth rate: ln?CPItln?(CPIt1)Bank of Canada overnight rate: Interest rate level at which the BoC lends to chartered banksFederal Funds rate: Interest rate level at which the Federal Reserve lends to bankswhere USD/CADtis the exchange rate between the Canadian and U.S. dollar in U.S. dollar terms, BCPIt is the level of the Banks commodity price index at time t, and CPIt is the level of the core consumer price index at time t to be used as a proxy for inflation in Canada. For USD/CAD, BCPI, and core CPI growth rates, I have approximated this by subtracting the natural logarithm of the level at time t-1 from the natural logarithm of the level at time t. This gives the advantage of making a comparison of growth rates, which is a consistent unit of comparison, rather than comparing levels. Comparing levels could lead to a scaling issue where a one-unit change in an independent variable could yield a change in the dependent variable that is not easily interpretable. Another variable that will be of interest in the analysis is the interest rate differential. I will define this variable as the Federal Funds rate subtracted from the Bank of Canada overnight rate. In theory a positive value would give an indication that money in the Canadian economy appreciates through time quicker and should be related to an appreciation of the Canadian dollar relative to the U.S. dollar. The models of consideration in the following sections are split-sample OLS and AR(1) processes.An important consideration before analyzing the data is to test if the variables are I(1) with a unit root, or are I(0) stationary and revert back to a consistent mean. To test this on the data, I have employed the Phillips-Perron test for a unit root. As outlined by Eric Zivot at the University of Washington, I am testing the null and alternative hypothesis as:H0: yt~I(1) H1: yt~I(0) (Zivot, 114)For all of the variables in outlined above, including the growth of the interest rate differential, p-values of the Phillips-Perron test are p<0.05, in support of a rejection of the null hypothesis at the 95% level. By taking the growth rates of core CPI, the BCPI, USD/CAD, and the interest rate differential as described above, we are able to proceed knowing that the data is I(0) stationary.A Consideration for Future Research:In future research, I would consider further testing the properties of my data and taking a more structured approach to developing my model for hypothesis testing. Before considering a model that provides the best fit for the data and allows for a proper test of a structural break, I did not consider the potential of simultaneity or joint-determination of the variables. After receiving feedback on the selection of variables, it is clear that there is a significant probability of a joint-determination problem. The most clear example is with the two variables of significant interest: USD/CAD and the BCPI. As Garmulewicz points out in his paper, interest rate differentials between countries can be viewed as endogenously or jointly determined with the exchange rate.(Garmulewicz, 7) The later part of Garmulewiczs paper also successfully shows that exchange rates contain information about commodity prices, and they can be used to predict future commodity prices. If I were conducting future research on this topic, I would consider employing methods to address the simultaneity bias problem. One way this could be handled is by estimating a two-stage least-squares model with instrumental variables. By estimating variables that are related to the independent variables but not jointly determined with the exchange rate (and overnight rate in the second hypothesis test) when testing my hypotheses, I could improve the ability for the model to determine direction of the relationships.5. Test for Structural Change in USD/CAD and BCPI DynamicsI now turn to the specification and results of the first hypothesis test. The formal model of consideration is: lnUSDCADt=?0+?1lnBCPIt+?2lnCPIt+?3BoCtFedt+?4lnBCPIt1+?5lnCPIt1+?6BoCtFedt1+?7lnUSDCADt1+ ?t (1)The equation can be re-written in terms of growth rates as the following:?lnUSDCAD=?0+?1?lnBCPI+?2?lnCPI+?3?BoCFed+?t (2)Equation (1) is a representation of the current factors affecting the exchange rate, along with their lags and the lagged value of the exchange rate. When regressing equation (1), which is not the formal test of the first question, I observe a high R-squared value (0.96). I hypothesize that the lagged exchange rate term has quite significant explanatory power over the current exchange rate. Thus, when the equation is re-arranged in the form of growth rates as equation (2) shows, I expected to see a reduction in the R-squared value of this regression due to the fact that the lagged exchange rate term is moved over to the dependent variable side of the equality. Because this lagged term has a large amount of explanatory power in determining the current exchange, I see a reduction in the R-squared term in the results for split-sample regressions of equation (2):(1)VARIABLESDeltalnCADUSDeltalnMBCPI0.199***(0.0226)DeltalnCPI-0.455(0.336)DeltaDiff0.0139**(0.00652)Constant0.000636(0.00116)Observations241R-squared0.275 January 1972 September 1998 September 1998 August 2018(1)VARIABLESDeltalnCADUSDeltalnMBCPI0.0680***(0.0261)DeltalnCPI0.0490(0.105)DeltaDiff-0.00224**(0.000894)Constant-0.00161**(0.000724)Observations317R-squared0.098 Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Comparing the BCPI growth rates before and after the Bank of Canada foreign exchange policy change in September 1998, both have statistically significant explanatory power at the 99% level. The most intriguing result is the fact that the BCPI growth rate appears to have over three-times more of an effect on exchange rate growth after the policy change. This matches with the prior discussed intuition that because the exchange rate is floating post policy change, it reacts more freely to market forces. BCPI growth is representative of a boost to Canadian exports and investment, ultimately acting a signal of the health of the Canadian economy in aggregate. There is more information about the Canadian economy captured in the exchange rate post policy change. This result provides evidence in support of the alternative hypothesis that there is a structural change in dynamics between USD/CAD growth and BCPI growth. 6. Test for Bank of Canada Overnight Rate MechanismThe second hypothesis test is trying to determine if the Bank of Canada overnight rate is sensitive to changes in BCPI growth. I now turn to the specification and results of the first hypothesis test. The formal model of consideration is: BoCt=?0+?1lnBCPIt+?2lnCPIt+?3lnBCPIt1+?4lnCPIt1+?5BoCt1+ ?t (3)The equation can be re-written in terms of growth rates as the following:?lnBoC=?0+?1?lnBCPI+?2?lnCPI+?t (4)Equation (1) is a representation of the current factors predicted to have an affect the overnight rate (namely BCPI growth and inflation growth), along with their lags and the lagged value of the overnight rate. Similar to the first hypothesis test, when regressing equation (3), I observe a high R-squared value (0.94). Again, I hypothesize that the lagged overnight rate term has quite significant explanatory power over the current overnight rate. Thus, when the equation is re-arranged in the form of growth rates as equation (4) shows, I see a reduction in the R-squared term in the results for split-sample regressions of equation (4) because the lagged overnight rate term is being included in the new dependent variable. The results of the split-sample regression are as follows: January 1972 September 1998 September 1998 August 2018(1)VARIABLESDeltaCanDeltalnMBCPI0.113(0.224)DeltalnCPI-3.188(6.917)Constant0.0210(0.0478)Observations317R-squared0.096(1)VARIABLESDeltaCanDeltalnMBCPI0.664***(0.229)DeltalnCPI0.903(3.403)Constant-0.0166(0.0117)Observations241R-squared0.136Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1Comparing the affect of BCPI growth rates on Bank of Canada overnight rate growth before and after the foreign exchange policy change in September 1998, the results are more inconclusive. After the policy change, the overnight rate is responsive to BCPI growth, however, a direct comparison to growth before the policy change cannot be made due to the fact that BCPI growth has an insignificant coefficient before the policy change. From this second hypothesis test, it can be determined that the overnight rate is responsive to BCPI growth after September 1998, however, it cannot be concluded that there has been a significant structural change compared to before the policy change. There is no evidence found from this split-sample regression in support of the alternative hypothesis. As a result, we fail to reject the null hypothesis for the second test. 7. Current Monetary Policy ImplicationCurrently, the bank of Canada is projected to continue raising the overnight rate moving into 2019 and 2020. As the article BoC raises its policy rate and opens door for faster policy normalization written by National Bank of Canada points out, the Bank of Canada has dropped the term gradual from its description of monetary policy tightening in the October 2018 monetary policy report. (National Bank of Canada, 1) This suggests that the Bank may consider continuing to raise the overnight interest rate moving into 2019.Given the recent decline in the BCPI index, this is a peculiar situation considering that there have also been two twenty-five basis point increases to the overnight rate in 2018. Looking at global commodity price growth, I plot the Canadian BCPI components and global crude oil prices above. I note that the energy component of the BCPI as shown on the left graph has been declining sharply in 2018. Highlighting the West Texas Intermediate and Western Canadian Select differential on the right graph, this differential has increased above 30 dollars compared to a historical mean of 17 dollars and 87 cents since 2005. As these crude oil prices make up approximately 45% of the banks commodity price index, I attribute the decline in Western Canadian Select (WCS) prices, along with a recent decline in West Texas Intermediate (WTI) prices (which make up 38.7% of the total BCPI in 2018) to be a major factor of the recent decline in the index overall. Given current infrastructure constraints in the Canadian oil and gas industry, the WTI/WCS differential has potential to remain close to the current level. The USD/CAD exchange rate has remained between $0.75 $0.80 USD in 2018 even with the increases in the overnight rate this year. Ultimately, this is consistent with my findings since the Banks 1998 policy change. The effect of the negative growth in the BCPI this year is outweighing the effect of interest rate increases as the Canadian dollar has depreciated this year. Again, I do not intend to interpret this through a normative lens, but rather demonstrate that this is consistent with historical data since 1998. The Bank of Canada does not focus on one sector or industry when considering how their policy relates to their objective of promoting economic welfare in Canada. It takes a holistic approach and a broad view of global and domestic developments in their entirety when making monetary policy decisions. 8. ConclusionThe results and conclusions of this paperA fitting afterthought of my research is described by Garmulewicz in his paper: While commodity prices go a long way in explaining movement in the real exchange rate, they are no deus ex machina.(Garmulewicz, 1) References: Chen, Y., Rogoff, K. Commodity Currencies. Journal of International Economics, 60, 133-160, (2003)Intervention in the Foreign Exchange Market. Bank of Canada, March 2011, https://www.bankofcanada.ca/wpcontent/uploads/2010/11/intervention_foreign_exchange.pdfSavoie-Chabot, L., Khan, M., Exchange Rate Pass-Through to Consumer Prices: Theory and Recent Evidence. Bank of Canada, Discussion Paper 2015-9, (2015)Garmulewicz, N., Power of Commodity Prices Over Exchange Rate Dynamics. University of Victoria Department of Economics, (2014)Bank of Canada Act. Bank of Canada, 1985, https://laws-lois.justice.gc.ca/PDF/B-2.pdfMonetary Policy. Bank of Canada, 2018, https://www.bankofcanada.ca/core-functions/monetary-policy/Commodity Price Index. Bank of Canada, 2018, https://www.bankofcanada.ca/rates/price-indexes/bcpi/A History of the Canadian Dollar. Powell, J., 1999, http://publications.gc.ca/collections/Collection/FB2-14-1999E.pdfUnit Root Tests. Zivot, E., 2018, https://faculty.washington.edu/ezivot/econ584/notes/unitroot.pdfBoC raises its policy rate and opens door for faster policy normalization. National Bank of Canada, 2018, https://www.nbc.ca/content/dam/bnc/en/rates-and-analysis/economic-analysis/boc-policy-monitor.pdf
Recent Comments