Time zones, GDP & exports

ABSTRACT

Recent studies have established a negative effect of time zone differences on trade flows. We extend this literature by examining whether a country’s economic size is relevant in its’ response to an increase in time zone differences. We argue that the negative impact of time zone differences should be more important for low-income countries as these countries often face higher trade costs and have firms with lower productivity compared to its high-income counterparts. To examine this heterogeneous impact, we interact the time zone measure with various quartiles of GDP. We find that these low-income countries face a much higher negative impact of time zone differences on exports compared to high-income countries. Our results help explain why the small countries of Samoa and Tokelau changed time zones to closely align with their main trading partners, while high-income countries have not taken such steps.

KEYWORDS:

• Exports
• Time zone effects
• GDP
• quartile

JEL CLASSIFICATION:

• F10
• F13
• F15

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ The negative impact of distance on trade is a widely robust finding in the trade literature. Both distance and time zones are based on location, hence these two variables are closely related. Indeed, the time zone literature stemmed from the distance literature. This interconnectedness was examined by Stein and Daude (2007) and Bista and Tomasik (2015).

² This negative effect was confirmed in papers by Hattari and Rajan (2012), Anderson (2014) and Tomasik (2013). Furthermore, nonlinearities in the time zone measure have also been examined and found in multiple papers such as Stein and Daude (2007) and Anderson (2014).

³ Waugh (2010) mentions that to reconcile bilateral trade volumes and price data within a standard gravity model, the trade frictions between high and low-income countries must be systematically asymmetric (with of course, poor countries facing higher costs to export relative to rich countries).

⁴ Gullstrand (2011) found similar results when examining high-income countries trading with Sweden.

⁵ Tybout (2000) further mentions that business regulations are unusually dense and unpredictable in less developed countries, while Aymo Weder and Kisunko (1999) mention that firms in less developed countries generally consider the institutional obstacles to doing business more burdensome than their OECD counterparts.

⁶ The export data are retrieved from the World Integrated Trade Solution (WITS) database. Under the WITS, the data are from the United Nations Statistics Division (UNSD) Comtrade database. List of countries in the sample is available upon request.

⁷ For more information, see http://faculty.haas.berkeley.edu/arose/Andrew Rose’s website: http://faculty.haas.berkeley.edu/arose/.

⁸ We conduct a test for the presence of heteroscedasticity in our data utilizing the modified Wald test for group-wise heteroscedasticity for the residuals in our model. The null hypothesis of homoscedasticity is rejected by the data.

⁹ In the case of the United States, the more centrally located Chicago was used instead of Washington, D.C.

¹⁰ UTC stands for Coordinated Universal Time and is standard time zone measure that ranges from −12 to + 14. Differences in time zones can be found by subtracting one UTC from the other. Here, 10-(−6) yields the 16 h time zone difference. Absolute time zone difference is measured on a 12 h scale. To find the absolute time zone difference for time zone differences greater than 12, subtract the time zone difference from 24. Hence, 24–16 = 8 h of absolute time zone difference.

¹¹ It includes the natural logs of variables such as the bilateral distance, population, annual real GDP per capita and product of the areas of the countries. It also includes bilateral pair dummies such as country pairs using the same currency, sharing a common language, having a regional trade agreement, sharing a common land border or having a colonial relationship. A complete list of all the gravity variable definitions are available upon request.

¹² To calculate the marginal effect using PPML estimation the formula used is: $e^{-0.07}-1=-7\mathrm{\%}$ . Marginal effects are reported in this section. The control variables, when significant, have signs consistent with the previous literature.

¹³ RESET test provides no evidence against the specification for Table 1, Columns 1 and 2. Both model passes the RESET test at 5% (and 1%) significance level.

¹⁴ To calculate the overall effect with an interaction term, we sum the coefficients of the baseline effect and the interaction term. For example, to find the time zone effect for the twenty-fifth GDP quartile, we add −6% (the baseline effect) to −8% (the interaction term) to get the overall effect of −14%.

¹⁵ These results are available from the authors upon request.