A Sketch Method for Forecasting Intercity Passenger Rail Demand

Transcription

1 A Sketch Method for Forecasting Intercity Passenger Rail Demand A five-step sketchplanning approach for casts are challenging to develop when INTRODUCTION Intercity passenger rail demand fore- either of the following two conditions forecasting intercity is present: new service is proposed where no such service existed, or no statewide passenger rail demand origin-destination travel matrix exists to estimate the total demand for travel. was developed and Unfortunately, both conditions applied when Virginia s General Assembly asked applied to a proposed the Virginia Department of Rail and Public Transportation (DRPT) to forecast ridership for proposed passenger rail new passenger rail service between Washington, DC, USA, service. The approach Bristol, VA, USA and Richmond, VA. 1 This proposed service is referred to as the used select elements TransDominion Express (TDX). TDX has been studied five times over of traditional urban the past 10 years. 2 7 The studies had moderate differences in projected annual travel demand operating subsidies ($9 million versus $22 million) and rolling stock (modern forecasting including tilt technology versus conventional train sets). The most substantive difference was trip generation, trip the annual ridership (26,000 versus more than 500,000). 8 Although additional information was required by the General distribution and a Assembly, one of the necessary tasks was variation of mode to reconcile these forecasts. The methodology for developing the choice forecast had three constraints. First, the forecast needed to be completed within approximately two months, given the need to package the forecast with additional information concerning TDX revenue, costs and congestion-reduction benefits prior to a date set by the General Assembly. 9 Second, the methodology needed to use existing data to the extent By JOHN S. MILLER, PH.D. AND BARCLAY F. THORN practical. DRPT staff advised against the use of survey data (to avoid possible survey bias), and the authors wanted to avoid simply assuming a ridership number without justification. Finally, the methodology needed to be transparent so users could understand how the forecasts were developed and what assumptions were made. This feature reports on the methodology developed for forecasting TDX ridership under these constraints, illustrates how to apply the methodology and discusses its applicability and limitations in other situations. LITERATURE REVIEW Previous Virginia Forecasts Table 1 shows rail service times and projected ridership from select TDX studies. Differences in ridership are partially attributed to different service levels. The 1998 study suggested that modern tilt equipment was feasible, which would offer faster service. The 2000 study suggested such equipment could not be used and, thus, slower service times would result. 10 However, the studies appear to use different sensitivities of ridership to service times. For example, it seems unlikely that the 11-percent increase in travel time from the 1998 report to the 2000 report would result in a 93-percent ridership decrease. Forecasts from Other Locations The literature suggests two lessons that should temper any passenger demand forecast. First, the elasticity of intercity passenger rail demand to travel time varies because demand is influenced by other factors such as food/beverage service, on-board amenities, frequency of service and location For example, ridership on the Amtrak Cascades grew from 94,000 (in 1993) to 590,000 (in 2003), with seven factors explaining the increase: increased train frequency; reduced train travel times; increased traffic congestion; customer service improvements; smart, local marketing and promotion; custom-built Talgo trains; and station improvements. 14 The tripling of ridership on the Capitol Corridor line from 463,000 (in 1998) to 1.2 million (in 2004) may be attributed not only to an increase in frequency (8 daily trains in 1998 compared to 24 daily trains in 2005) but also to improved stations and better amenities ITE Journal on the web / January

2 Table 1. Detailed comparison of train schedules in previous studies (trains per day in parentheses, ridership forecasts in brackets). Segment Bristol to Lynchburg Lynchburg to Washington, DC Lynchburg to Richmond Richmond to Washington, DC such as a pair of outlets for every seat for increasingly prevalent laptop computer use, flip-down tray tables, a staffed food service car and a wider seat pitch Although a Boston, MA, USA to Portland, ME, USA, travel time reduction from 165 minutes to 150 minutes in May 2005 is suggested as the reason for a 47-percent increase in ridership on the Amtrak Downeaster line, a customer survey suggested that on-time performance, quality food service and friendliness were also important for attracting ridership. 17 The limited data regarding commuter rail system elasticity suggest variability by location. A calculated 9.2-percent improvement in service (Philadelphia, PA, USA) yielded an 8.6-percent increase in ridership; a 77-percent service improvement in Boston yielded a 37.5-percent increase in ridership. 18 Second, previous forecasts have not necessarily been accurate. The Federal Railroad Administration forecast a ridership of 14.7 million passengers in the northeast corridor of the United States as a result of service improvements, but the growth was actually to 9.8 million. 19 Slower-than-expected travel times were partly to blame, but unforeseen modal competition, such as the low-cost People s Express airlines, was also a factor. More than a decade later, accurate forecasts are still a challenge. In 2006, the Federal Transit Administration did not adopt a proposal specifying that forecasting technical methods be certified because of concerns expressed by transit community members Frederic R. Harris (1998) 4:09 (2) [20,133]* 3:16 (2) [158,972] 2:25 (2) [16,617] 6:47 (2) [178,678] Amtrak Modified (2000) Amtrak Alternate (2000) 4:47 (2) 4:47 (1) 3:29 (2) 3:29 (2) 2:41 (2) 2:41 (1) 7:11 (2) 2:14 (1) Total ridership [374,400] [26,252] [40,750] * Note: For example, the 1998 study indicated that travel between Bristol and Lynchburg required 4 hours and 9 minutes. Based on this travel time and a frequency of two trains per day, the 1998 study predicted an annual ridership of 20,133 on this segment. that there was a lack of guidance regarding how to characterize forecast uncertainty. 20 METHODOLOGY A five-step methodology generated the TDX ridership forecast. These five steps were chosen because they met the three requirements mentioned previously: They were feasible within a relatively short time frame; they made use of existing data where applicable; and the computations themselves were transparent. They also addressed two lessons from the literature: The elasticity of ridership-to-service levels may vary substantially, thus large errors may result if the assumed relationship between service and ridership is wrong. Because Virginia does not have statewide origin-destination data for non-work trips, the first three steps estimated the potential non-work ridership between each station pair. Only a fraction of this potential ridership will use rail service as calculated from the sensitivity of demand to rail service levels in step five. The five steps were as follows: 1. Determine the number of potential non-business trips generated by each TDX station area. Three market segments most likely to use TDX were college students, tourists and nonvehicle households. It was assumed that a TDX station was accessible to any population, college, or tourist site in the same city or surrounding county in which the station was located. These three market segments were determined as follows: College trips. Virginia has 18 colleges along the TDX corridor. At each station, the number of in-state students was determined. For example, the percentage of in-state students at Sweet Briar College in Lynchburg, VA, is 39 percent compared with 92 percent at Radford University in Radford, VA. Because a TDX station would be accessible to 45 percent of the state s population, each station s in-state students was multiplied by 45 percent to estimate the number of students with residences along the TDX corridor. The resultant figure was multiplied by eight, assuming four round trips between home and college per student per year. Tourist trips. The total number of annual tourists in Virginia is 54.8 million, 21 percent of whom are from within Virginia Although tourist numbers for individual localities were not always available, tourist spending by locale was available. This relationship was used to estimate the total tourist trips assuming two trips per year. For example, a single TDX station could serve Charlottesville, VA and Albemarle County. These two jurisdictions accounted for 2.2 percent of the total tourism spending in Thus, 2.2 percent of the 54.8 million state visitors is approximately 1.2 million visitors per year. Visitors from inside the state comprise 21 percent of this amount and 45 percent, or 113,855, come from a TDX jurisdiction. Assuming two trips per year yields a potential 227,710 TDX tourist trips per year. Non-vehicle households. Non-vehicle households were assumed to make one round trip per year to another Virginia location. For example, multiplying the 3,787 non-vehicle households served by the Charlottesville station by 45 percent yields 1,704 non-vehicle households with a potential destination along the TDX corridor. 24 Assuming one round trip per year yields a potential 3,406 TDX non-vehicle household personal trips per year. 70 ITE Journal on the web / January 2009

3 The sum of these three market segments yields the total potential non-business trips generated by each station. 2. Select six stations for further analysis, with at least one station from each region. The six stations with the highest ridership and that spanned most of the TDX network were Alexandria, VA; Charlottesville; Lynchburg; Radford; Richmond; and Roanoke, VA (see Figure 1). These stations generated 90 percent of the total potential trips. 3. Estimate the total potential non-business ridership between each pair of stations. A singly constrained gravity model was used to estimate the total demand between each station pair. This gravity model was formulated as follows, using the trips from station 1 to station 2 as an example: Figure 1. Proposed TDX service in Virginia. (Eq. 1) Potential travel 12 = S 1 A 2 D 12 A 2 D 12 + A 3 D 13 + A 4 D 14 + A 5 D 15 + A 6 D 16 where: S = total trips generated at station (from step one) A = sum of population and employment for each station D = inverse of the distance between two station areas In Equation 1, the demand for travel from station 1 to 2 is based on the total trips generated by station 1 and the relative attractiveness of station area 2 to stations 3, 4, 5 and 6. This relative attractiveness depends on two factors: the closeness of station 2 to station 1 (reflected by D 12 ) and the number of opportunities at the station area 2 (reflected as A 2 ). A singly constrained gravity model was used because there was greater confidence in the total number of trips generated (S) and in the impedances (1/D) but less confidence in the measure of attraction (A). Thus, the literature s suggestion that a singly constrained gravity model may be preferable to a doubly constrained Figure 2. Four demand functions used to estimate travel demand as a function of rail service time. For example, suppose an automobile trip is 100 minutes and the same trip by rail is minutes. The percentage of rail trips (by Equation 2) is x = ( 0.228) = , or approximately 1.26 percent. By contrast, the percentage of rail trips (by Equation 5) is e ( 0.228) = 0.31 percent. gravity model in some situations was followed. 25 A doubly constrained gravity model was tested; such a model would have forced zero trips between all station pairs except those that included Alexandria. As this result did not seem reasonable, the decision to use the singly constrained gravity model appeared preferable. 4. Estimate potential business trips. Unlike non-business trips, a relatively good surrogate for a business trip distribution table existed in the form of journey-to-work travel available from the county-to-county worker flow files provided by the U.S. Census. 26 For example, tabulations showed a total of 316 workers traveling from the Lynchburg station area (Amherst County, Campbell County and Lynchburg) to the Charlottesville station area (the city of Charlottesville and Albemarle County). Assuming these workers make three round trips per week, 50 weeks per year, there may be 94,800 trips from the Lynchburg station to the Charlottesville station. 5. Determine the actual rail ridership as a function of travel time. Data from previous TDX studies suggest four different relationships between rail demand and travel time, as shown in Figure 2. The four demand functions ITE Journal on the web / January

4 Non-business trips Table 2. Total potential passenger trips per year between stations. Station Radford Roanoke Lynchburg Charlottesville Alexandria Richmond Radford 39,045* 9,819 7,423 89,261 30,472 Roanoke 39,045 30,855 16, ,163 65,366 Lynchburg 9,819 30,855 16, ,714 55,233 Charlottesville 7,423 16,579 16, , ,196 Alexandria 89, , , ,962 1,458,925 Richmond 30,472 65,366 55, ,196 1,458,925 Business trips Radford 604,200 16,500 12,450 34,500 23,100 Roanoke 604,200 74,400 5,850 4,500 15,600 Lynchburg 16,500 74,400 63,750 9,750 30,300 Charlottesville 12,450 5,850 63,750 30, ,200 Alexandria 34,500 4,500 9,750 30, ,300 Richmond 23,100 15,600 30, , ,300 * Note: For example, there are 39,045 total potential passenger trips from Radford to Roanoke. These trips could be made by any mode, not necessarily rail. Table 3. Total predicted train passenger trips between stations. Radford Roanoke Lynchburg Charlottesville Alexandria Richmond Radford 0 to 301* 0 to 29 0 to 14 0 to to 14 Roanoke 0 to to 1,849 0 to to 3,446 0 to 41 Lynchburg 0 to 34 0 to 2,290 1,235 to 3,228 1,919 to 5,639 0 to 1,051 Charlottesville 0 to 13 0 to to 2,785 3,737 to 10,244 0 Alexandria 0 to to 3,194 1,556 to 5,212 3,498 to 9,964 0 Richmond 0 to 17 0 to 51 0 to 1, * Note: For example, depending on which of the four demand functions from Figure 2 (Equations 2, 3, 4, or 5) is chosen, there may be as few as 0 or as many as 301 rail trips from Radford to Roanoke. shown in Figure 2 and Equations 2 through 5 were not given in previous studies; rather they were inferred from ridership forecasts and travel times that previous studies appeared to have used. 27 In the aggregate, usually an optimistic forecast results from Equation 2, a pessimistic forecast results from Equation 4 and a mid-range forecast results from Equation 5, which presumes captive riders regardless of service time. Equations 2 through 5 were used to estimate the upper and lower proportion of potential trips by rail for each station pair. For example, the automobile travel time from Lynchburg to Richmond is given as 142 minutes and the TDX travel time as 175 minutes. 28 Because the TDX travel time is 23 percent greater than the automobile travel time, Equations 2 and 3 suggest that the TDX trips between Lynchburg and Richmond are between 0 percent and 1.23 percent of the total potential trips. (Eq. 2) Percent Trips by Rail = ( 0.23) = 1.23% (Eq. 3) Percent Trips by Rail = ( 0.23) = 0.7% (hence zero trips by rail) RESULTS Table 2 shows the result of steps one through three (potential non-business ridership) and step four (potential business ridership). Table 3 shows the forecast TDX ridership. The ranges in Table 3 arise because, based on previous studies, four different relationships between ridership and travel time are possible, as shown in Equations 2 through 5 in Figure 2. Equation 2 suggests the six stations could serve 51,985 annual passenger trips; Equation 4 suggests 12,806 passenger trips. Recognizing that the six stations captured approximately 90 percent of the riders, the range of annual passenger trips may be increased by approximately 10 percent: 14,000 to 58,000 with a midpoint value of 36,000. The results are traceable. For example, Table 2 suggested 55,233 non-business trips and 30,300 business trips from Lynchburg and Richmond, yielding a 72 ITE Journal on the web / January 2009

5 Forecast step Non-business trip generation (step 1) Non-business trip distribution (step 3) Business trip generation and distribution (step 4) Demand based on travel time (step 5) Table 4. Impact of TDX methodology assumptions on the passenger rail forecast. Assumption made Three market segments (tourists, students and zero-vehicle households) represent potential TDX ridership. Based on the surrounding city and county population data, TDX can serve 45 percent of the state s population. For colleges and universities located in the TDX corridor, about 45 percent of their instate students come from homes located in the TDX corridor. Population and employment are equally good attractors. Commuters travel to work on average three round trips per week, 50 weeks per year. The time required to access the station from the final origin or destination is negligible. No special bus service is provided to make access between the station and the final origin or destination. Example of a challenging assumption An additional market segment, recreational shopping, should have been included. Growth in certain portions of the state means TDX can potentially serve more of the state s population. Only a small portion of such students come from homes located in the TDX corridor. Employment is a better indicator of an area s attractiveness than population. Because of telecommuting, fewer long-distance trips are made. The access time is not minimal, thus the passenger rail time should be increased. Localities would be willing to provide such express bus service, making rail more appealing. Example of how to quantify the impact Increase non-business trips by 25 percent. Increase the percentage of the state having access to TDX from 45 to 55 percent. Decrease this percentage from 45 to 15 percent. Weight employment by a factor of four relative to population. Reduce potential business trips by 25 percent. Increase rail travel time by 15 minutes (increases average trip length by 9 percent). Modify demand slightly so that the most pessimistic demand function (Equation 4) is eliminated. Modify demand substantially such that only the most optimistic demand function (Equation 2) is used. How the challenging assumption affects ridership 19 percent * 17 percent 2 percent (decrease) 0.1 percent 6 percent (decrease) 34 percent (decrease) 8 percent 60 percent * Note: For example, if non-business trips are increased by 25 percent, the forecasted annual trips increase by about 19 percent based on averaging the forecasts of what is usually the most optimistic demand function (Equation 2) and what is usually the most pessimistic demand function (Equation 4). total of 85,433 trips by some mode between the two cities. Equations 2 and 3 suggested that the percentage of trips by rail might range from 1.23 percent of 85,533 to as low as 0. This range is shown in Table 3. These steps delineate the assumptions implicit in previous studies. For example, the disparity between previous forecasts could be attributed to two factors: differences in service times and differences in demand sensitivity to these travel times. In step five, a given set of service times those based on the assumption of conventional technology and those based on a relatively recent and detailed study was chosen. 29 By focusing on the four demand functions relating ridership to travel time, it became evident that ridership might vary by a factor of more than four (from 14,000 to 58,000) with the difference being solely attributable to sensitivity to travel time. Thus, although technology affects the variation in forecasts, technology is not the only or even the largest contributor to this variation. DISCUSSION: SOURCES OF FORECAST ERROR Although the TDX has not been implemented, a portion of the proposed service overlaps with existing passenger rail service. Amtrak service runs from Washington, DC to Lynchburg (as does the TDX). However, in Lynchburg, Amtrak continues south to other states while TDX continues west, terminating in Bristol. A rough comparison between TDX forecasts and observed Amtrak ridership may be made using the Lynchburg station. Amtrak showed 16,847 annual trips to and from the Lynchburg station in If it is assumed that half were between Lynchburg and points north, the midpoint estimate of 11,217 trips between Lynchburg and points north in Table 3 differs from the observed value ITE Journal on the web / January

6 by 33 percent. However, this comparison between existing and proposed service is not exact, partly because of assumptions about service and demand. The impact of such assumptions on total ridership may be considered with this methodology. For example, this effort assumed each station could serve the city and surrounding county in which it is located, meaning TDX could reach 45 percent of Virginia s population. However, if an assumption had been made that TDX could serve 55 percent of the state s population, the non-business trip generation step (step one in the methodology) would yield a 17-percent increase in ridership. Table 4 identifies this and other assumptions and quantifies the impact of changing them on the passenger forecasts included in Table 3. Assumptions that were not explicitly part of the methodology, such as amenities, may only be considered indirectly. In these cases, two or more tests may be used, as shown in the last row of Table 4. Ridership is most sensitive to changes in the demand functions that relate rail ridership to travel time (see Figure 2). Changes in the use of these functions may affect ridership by more than 60 percent in Table 4. Changes in the other assumptions, such as the number of potential non-business trips generated, have a lesser impact of 19 percent. These results suggest that the most productive area to refine in this methodology is the relationship between travel time and rail use (see Equations 2 through 5), including the presumed travel times for automobile and rail. This analysis also shows that the forecasts in Table 3 are useful as a rough order of magnitude estimate. Given that three earlier studies produced very different estimates, is forecasted ridership closest to the 1996 study (500,000 annual trips), the 1998 study (374,400 trips), or the 2000 study (40,750 or fewer trips)? 31 Although the range of trips shown in Table 3 initially appears large (14,000 to 58,000), this range clearly approaches the 2000 study. While the range may appear large when viewed in isolation, it is small when compared to the difference in forecasts from previous studies. CONCLUDING REMARKS AND IMPROVEMENTS FOR FUTURE WORK Previous TDX forecasts had varied by more than a factor of 10 (such as 500,000 in 1996 and 40,750 in 2000). 32 The forecast range given in this article (14,000 to 58,000) aligns with the 2000 forecast. Thus, one contribution of this methodology is that it yields a transparent approach to generate a rough order of magnitude forecast. When the report was presented to various audiences, one reaction was that TDX ridership was affected by the assumptions noted in Table 4, such as each locality s willingness to provide express bus service to TDX stations. This methodology may also be used to estimate the impact of various assumptions. For example, a 25-percent decrease in total non-business trips (by all modes) elicited a 6-percent decrease in rail ridership, while a smaller percentage increase in rail service times (9 percent) yielded a larger percentage decrease (34 percent) in rail ridership. Generally, assumptions about travel time and related demand functions have a greater impact on ridership than assumptions about trip generation and trip distribution. Another reaction to this work was that the focus was not on TDX ridership but rather what steps need to be taken to implement TDX in Virginia. Another contribution of the methodology is that it identifies, at a sketch-planning level, which portions of TDX are more promising should service be offered. Table 3 showed that particular segments account for a large portion of the ridership (70 percent of the passengers are located at stations between Lynchburg and Alexandria). To that extent, there may be certain corridors for which pilot service might be productive. For such corridors, the sketch approach could be reapplied with a greater focus on the demand functions (see Equations 2 through 5). The literature suggested that other factors such as seat comfort influence demand. The last row of Table 4 showed that the impact of amenities such as express bus service is not yet fully understood. Therefore, an improvement to the methodology used for this study would be to incorporate such variables into the demand functions n References 1. HB 5002 Budget Bill: Appropriations for Biennium. Richmond, VA, USA: Virginia General Assembly, Accessible via leg1.state.va.us/cgi-bin/legp504. exe?062+bud Ibid. 3. House Document 51: Rail Passenger Service. Richmond: Virginia Department of Rail and Public Transportation, Accessible via leg2.state.va.us/dls/h&sdocs.nsf/by+year/ HD511996/$file/HD51_1996.pdf. 4. Bristol Rail Passenger Study Phase 2: Final Report. Accessible via studies/files/bristol2.pdf. 5. Rail Passenger Service Study: Richmond, VA/Washington, DC to Bristol, VA. Richmond: Virginia Department of Rail and Public Transportation, Phase II: Bristol, Roanoke, and Richmond Passenger Train Study. Richmond: Virginia Department of Rail and Public Transportation, House Document No. 37, Status of the Trans- Dominion Express Passenger Rail Service. Richmond: Virginia Department of Rail and Public Transportation, Accessible via virginia.gov/studies/files/tdx-report-final. pdf. 8. See notes 3, 4 and 5 above. 9. TransDominion Express Update Study, House Document No. 2. Accessible via leg2.state. va.us/dls/h&sdocs.nsf/by+year/hd22007/$file/ HD2.pdf. 10. See notes 4 and 5 above. 11. Perl, A. North American Intercity Rail Passenger Systems. Transportation in the New Millenium. Washington, DC, USA: Transportation Research Board, Accessible via onlinepubs.trb.org/onlinepubs/millennium/ pdf. 12. Christiansen, D.L. Planning for Intercity Rail Passenger Service at the State Level. Transportation Engineering, Vol. 48, No. 2 (1978): Evans, J.E. Transit Cooperative Research Program Report 95: Traveler Response to Transportation System Changes. Washington, DC: Transportation Research Board, Amtrak Cascades Ridership and Station On-Off Information. Olympia, WA, USA: Washington State Department of Transportation, Accessible via rail/cascades/ridership.cfm#ridership. 15. The Capitol Corridor 2005 Performance Report. Oakland, CA, USA: Capitol Corridor 74 ITE Journal on the web / January 2009

7 Joint Powers Authority, Accessible via perfreport05.pdf. 16. Eugene Skoropowski, personal communication, October 20, Downeaster Achieves Record Ridership. The Downeaster. Portland, ME, USA: Northern New England Passenger Rail Authority, Accessible via See note 13 above. 19. Vaca, E. Intercity Rail Ridership Forecasting and the Implementation of High-Speed Rail in California. Berkeley, CA: University of California Transportation Center, Public Transportation: New Starts Program is in a Period of Transition. Report GAO Washington, DC: Government Accountability Office, Visitation Estimates: Virginia Total Travel. Richmond: Virginia Tourism Corporation, Accessible via DKSAVisitationEstimates2005.pdf. 22. Where Do Virginia s Visitors Come From? Richmond: Virginia Tourism Corporation, Accessible via WheredoVAsVisitorscomefrom.pdf. 23. Virginia Locality Economic Impact Data. Richmond: Virginia Tourism Corporation, Accessible via virginiascan.yesvirginia. org/localspending/. 24. Census Transportation Planning Package (CTPP) 2000: Profiles for Virginia. Washington, DC: American Association of State Highway and Transportation Officials. Accessible via ctpp. transportation.org/home/va.htm. 25. Virginia Travel Demand Modeling Policies and Procedures Manual, Draft Version Richmond: Virginia Department of Transportation, Residence County to Workplace County Flows for Virginia. Washington, DC: U.S. Census Bureau, Accssible via cen2000/commuting/2kresco_va.xls. 27. See note 9 above. 28. See note 6 above. 29. See note 5 above. 30. Amtrak Fact Sheet, Fiscal Year 2006: Commonwealth of Virginia. Amtrak Government Affairs. Accessible via pdf/factsheets/virginia06.pdf. 31. See notes 3, 4 and 5 above. 32. See notes 3 and 5 above. 33. See note 11 above. 34. See note 14 above. 35. See note 17 above. JOHN S. MILLER, Ph.D., is a licensed professional engineer and a research scientist with the Virginia Transportation Research Council. He holds a bachelor of science in electrical engineering and a Ph.D. in civil engineering from the University of Virginia. BARCLAY F. THORN holds a bachelor of art in history from Dartmouth College and a J.D. from the University of Virginia School of Law. He currently practices law as an associate with Schiff Hardin LLP in Chicago, IL, USA. ITE Journal on the web / January