Devalue Currency To Augment Demand

Taka loses fast its value against dollar,
Reasons put forward for paying the bills of importer.
Increase in money supply depreciates local currency,
Devalued taka augments demand, output and local vacancy.
Overshooting exchange rate may approach stable value,
Lowering interest rate is a must amid pandemic flu.

A recent news report (May 20)says there has been a dollar crisis. To meet the growing import bill amid sluggish export, demand for dollar surged. Taka against dollar has been depreciated to Tk 88.50 at the banks. This is happening when government doled out Tk 5000 crore credit to RMG owners to clear dues of workers. In addition, government assured cash benefits to rural poor and other incentives to various sectors. By the time I am writing this piece, remittances in the month of Ramzan have reached $1.09 billion.

I embarked upon to see what impact an increase in M2 would leave on exchange rate. M2 comprises of currency outside bank, demand deposits, narrow money supply and time deposits. Data were taken from Bangladesh Economic Review 2018 for the period 1996-2018.

Autocorrelation check for 23 observations and 1 explanatory variable reported positive correlation (d = 0.374). I did not transform the data. It was assumed that during the unit root test inclusion of lagged residuals will take care the autocorrelation.

Then I went for unit root test to see whether exchange rate and M2 were stationary. To bare eyes, it appeared that both the variables wandered around a trend. So I constructed the following regression equations:

🔺 Exch_t = a + bt + c Exch_t-1 + d 🔺 Exch_t-1
🔺 M2_t = a + bt + c M2_t-1 + d 🔺 M2_t-1
Where 🔺 Exch_t= Differences in exchange rate at t,
Exch_t-1 = exchange rate at t-1,
🔺 M2_t= Differences in M2 at t,
M2_t-1= M2 at t-1,
🔺 M2_t-1= Differences in M2 at t-1,
t = a time trend variable, here year.

After the regression run , I obtained the following result: 🔺 Exch_t = -1778.34 + 0.902t -0.501Exch_t-1 + 0.455 🔺 Exch_t-1
(t=-2.52, p=0.022, se=706.22) (t = 2.53, p=0.022, se=0.36) (t=-2.79, p=0.012, se=0.18) (t=1.97, p=0.065, se=0.230)
(F=3.39, p=0.042) 🔺 M2_t = -7176664 + 3590.12t + 0.0028M2_t-1 + 0.373 🔺 M2_t-1
(t= -2.59, p=0.019, se=2774933) (t=2.59, p=0.019, se=1386.81) ( t= 0.088, p= 0.93, se=0.0313) (t=1.27, p=0.22, se=0.29)
(F=67.54, p=0.00)

Huge standard errors put question mark on the intercept and trend coefficient of 🔺 M2 function. Tau statistics of slope coefficients of lagged exchange rate and M2 , -2.79 and 0.88 , in absolute terms were smaller than MacKinnon critical tau statistics at 5% level, -3.4620. So I did not throw away the null hypothesis that c=0 or exchange rate or M2 are nonstationary.

As the first differences of these two variables appeared to be nonstationary, it was assumed , for the sake of simplicity that they were integrated on order d,I(d). Regressing exchange rate on M2 , I obtained the residuals for cointegration test. Then I ran the following regression:

🔺 resid_t = b resid_t-1 + c 🔺 resid_t-1

And the result was: 🔺 resid_t = -0.160 resid_t-1 + 0.51 🔺 resid_t-1

(t=-2.012, p=0.058,se=0.079) (t= 2.84, p=0.010,se=0.18)

(F=5.85,p=0.011)

The computed tau statistic -2.012 was greater than the critical value -3.37% at the 5% level of significance. I did not reject the null hypothesis that least squares residuals are not cointegrated. Cointegrated Regression Durbin Watson (CRDW) test also validated the claim . The computed d = 0.374 turned out to be smaller than the critical value 0.386 at 5% level of significance. So I did not reject the hypothesis that exchange rate and M2 are not cointegrated.

In this particular situation, exchange rate and M2 were I(d) series and not cointegrated. So I went for a VAR model:

🔺 Exch_t = b₁🔺 Exch_t-1 + b₂ 🔺 Exch_t-2+ b₃🔺 M2_t-1 + b₄ 🔺 M2_t-2+ v^🔺Exch_t

🔺 M2_t = c₁ 🔺Exch_t-1 + c₂ 🔺 Exch_t-2+ c₃ 🔺 M2_t-1 + c₄ 🔺 M2_t-2+ v^🔺M2_t

VAR model did not fit well (🔺 Exch chi² =6.88, p= 0.144, and for 🔺 M2 chi²= 142.85, p=0.00). Nevertheless, I wanted to see the Impulse Response Function (IRF) that shows effect of a shock of endogenous variable on itself and other endogenous variables. An increase in orthogonalized shock to M2 resulted in a short decrease ( depreciation of Taka ) in the exchange rate that withers away 1 period later.

Though the VAR model is to be accepted with a dollop of salt, this is pretty much in line with theory found in economic text book. Temporary drop in global demand shifts the DD schedule, which shows mixes of output and exchange rate for keeping output market in equilibrium in the short-spell, to the left. This in turn reduces full employment-level output to a lesser level, provoking depreciation of currency. A currency depreciation augments both aggregate demand and output at home. Meanwhile, increase in money supply in the domestic market depreciates exchange rate and causes AA schedule, which links exchange rates and output levels to keep the money and foreign exchange markets in equilibrium, to shift upward. Domestic goods become more competitive in global market , triggering a rise in domestic output and employment. For a given level of output, an increase in money supply can cause exchange rate to overshoot its long-term exchange rate for a while. One may argue that since our import surpasses our export and in this time of falling export earnings a depreciation may erode our current account balance. Point is that economic theory says for a brief period there may be a dent in the current account balance (ours a negative) but in the long run it will definitely improve.

Point is currency depreciation is good for our economy and wild fall in Taka may approach its long-run value with the course of time. To revive the falling demand, government can do more apart from doling out incentives. One step can be to lower the domestic interest rate in a bid to increase the money supply.