Total Non-Farm Employment

Comparing Arkansas’ total non-farm employment to the U.S. indicates the State is performing better relative to the U.S.  The chart below graphs the natural logs of the monthly values for each series normalized to January 2015. Normalization to this date reflects the relative performance beginning when Governor Hutchinson took office.

Note that the slope of the Arkansas line exceeds that for the U.S. beginning April 2015 until April 2016. The Arkansas slope dips a bit until June 2016 after which the growth rate of Arkansas exceeds the U.S. until April 2017.  Note the slopes of both curves is about the same since mid-2017.

Arkansas non-farm employment growth rate dipped a bit in mid 2017, remained flat until April 2018 and is now growing slightly better that the U.S.


The Bureau of Labor Statistics updates the State’s employment mid-month following the end of the previous month.

Updated 16 November 2019.  The last data point is October 2019.

SOURCE: FRED, US – PAYEMS, ARKANSAS – ARNA and the author’s calculations

Lottery Revenues

The graph reveals that lottery revenues peaked in May 2011 (not shown) followed by a long decline reaching a minimum in December 2014. A major revamp of the lottery yielded a strong upward thrust which peaked in August 2016 followed by a gradual decline. The data suggests when new games are added lottery revenues increase. Lottery revenues vary seasonally by month of the year.

Recent lottery revenue suggests total sales peaked in October 2018.  The October revenues reflect an increase in sales resulting from the large lottery payout.  The statistical forecast identified a pulse variable in October equal to +$18,510,000. The forecasts include a slight increase by the end of the fiscal year in June 2019.  This forecast for total lottery revenue for the 12 months ending June 2019 = $516,269,700 reflecting the historic pattern of the series since December 2010, including the seasonal.  This updated forecast reflects a flat outlook beginning in October 2018 for gross lottery sales for the year ending June 2019.

The official forecast from the Director’s office for fiscal year ending June 2019 is $482,900,000.  Actual total revenues for the year was $515,005,000.

Using monthly total monthly sales thru September 2019, the forecast for total 12 month sales thru June 2020 = $515,337,000.  The updated forecasts are flat thru June 2020.


The statistical analysis of the lottery revenue series reveals that three significant events (all unknown) affected the revenues.  In January 2016, some event occurred that had the effect of increasing revenues $11,240,000 above the expected value, in January 2017 another event occurred that decreased revenues by $9,354,000 from the expected, and in August 2017 another event occurred that increased revenues by $6,323,000.

An investigation by the lottery staff would reveal the most likely cause of these three events.  After their identification, management could take action to negate the negative event occurring in January 2017 and replicate the positive events in January 2016 and August 2017.  This assumes, of course, that the three events were the result of management decisions and not events beyond the control of management.

For those readers who are interested in the final ARIMA model, the R² = 98.3%, and all assumptions about the error terms are satisfied.

The Summary of Lottery Forecasts recaps the monthly forecasts for the past 12 months.  The forecast dated June – 17 uses data available thru June 2017 and represents forecast for the 12 months ending June 2018.

Note the upper and lower 80% confidence intervals decreases as the year ending June 2018 approaches.  This reflects that the equation used to calculate the upper and lower limits includes a value for the number of periods ahead the forecast is calculated, i.e., forecasting 12 periods ahead includes 12 in the equation while forecasting four periods ahead uses four in the calculation.  Hence, the upper and lower limits converge (become more accurate) as terminal date approaches.


Generally, the forecasts for the 12 months ending June 2018 began at $454,752,700 on June 2017 and gradually increased to the actual 12 months ending June 2018 at $500,010,000.  The difference in the monthly data for total sales for all counties used in this analysis and the reported total annual sales in the Director’s Annual Report ($449,141,028) is attributed to biannual audits of the lottery’s monthly reported sales not reflected in the monthly disclosure reports.

SOURCE: Lottery Commission, monthly disclosure reports, all counties total sales The Commission updates its revenue collections approximately eight – ten days after the end of the month.

Forecasts updated on 15 October using the data ending in September 2019.


The graph shows the actual net revenues for the ending 12 months from January 2017 thru November 2019.  The forecast for the net available revenues for the 12 months ending 30 June 2020 = $6,049,139.  As an additional month of actual net revenues is reported, a new 12 month total is calculated. The new 12 month total is used, along with all the previous actual month’s totals, to calculate an updated forecast for the 12 months ending June 2020. In essence, the forecast for the 12 months ending June 2020 is revised monthly as actual monthly net revenues are reported.

The forecast for the 12 months ending June 2020 = $6,049,138. This is an updated forecast from the previous $6,035,086.  The State’s official forecast issued 15 May 2019 for the 12 months ending June 2020 = $5,537,100.


An upward trend in net revenues existed until April 2016 (not shown).  Beginning in April 2016, net revenues exhibited little changes as noted in the flat portion of the graph above.  The flat points in the above graph ended in November 2017.  The upward trend resumes here.


The following graph shows the June 2020 forecasts for net revenues by month and the upper and lower 95% confidence intervals around each forecast.

Note that the upper and lower limits are wider surrounding the June 2019 forecast made in July 2018 compared to the much narrower limits surrounding the June 2019 forecast made in May 2019. The wider vs. narrow limits reflect that the equation for the confidence intervals for the forecasts 12 months ahead made in July 2018 includes the number 12 while the formula for one period ahead made in May 2019 includes the number one.  Hence, the farther ahead the forecast the greater uncertainty as evidenced by the wider confidence limits.


The graph above shows the monthly forecasts hovering around $5,650,000 until February 2019 when it began to increase.

The State’s fiscal year ends in June.  The updated monthly forecasts for the fiscal year ending June 2020 is shown below.


The forecast for the 12 months net revenue collections ending June 2020 is $6,049,138.

The monthly updates for the forecast net revenue collections using data ending the previous month are:

Forecast for 12  months ending June 2020:

Origin                                                                     Forecast

July                                                                          $5,873,700

August                                                                     $5,953,580

September                                                              $5,944,460

October                                                                   $5,998,230

November                                                              $6,035,090

December                                                               $6,049,138

Updated 4 December 2019 using net revenue collected thru November 2019.

Source:  Arkansas Dept. of Revenue and Finance

Housing Permits

The graph shows the New Private Housing Units Authorized by Building Permits: 1 – Unit Structures for Arkansas, not seasonally adjusted. The series exhibits significant variation and is highly seasonal with November and December permits significantly lower that the other ten months.

The updated forecasts thru March 2020 are shown below.

Note that the autumn seasonal downswing is beginning.  Note further that the peak of the seasonal upswing is lower than the upswings beginning in December of  2016 and 2017.  The lower peak in the 2019 seasonal upswing is cause for concern.


The U.S. Bureau of Census updates this data series approximately 45 days after the end of the month.

Updated 16 November with last data point September 2019.

Source: FRED:  Series = ARBP1FH, NSA


The Federal Reserve Bank of Philadelphia produces leading indexes for each of the 50 states. The indexes are calculated monthly and are usually released a week after the release of the coincident indexes. The Bank issues a release each month describing the current and future economic situation of the 50 states.

The leading index for each state predicts the six-month growth rate of the state’s coincident index. In addition to the coincident index, the models include other variables that lead the economy: state-level housing permits (1 to 4 units), state initial unemployment insurance claims, delivery times from the Institute for Supply Management (ISM) manufacturing survey, and the interest rate spread between the 10-year Treasury bond and the 3-month Treasury bill.

A time-series model (vector autoregression) is used to construct the leading index. Current and prior values of the forecast variables are used to determine the future values of the index.

The following chart graphs the index for AR and the U.S.  Note that most of the time the AR index is lower that the U.S.  Furthermore, the AR index exhibits a downward slope since it peaked in September 2015. A slight uptick in the AR index is evident beginning in early 2019.


The chart below shows the monthly indices for the U.S. and Arkansas in natural logs normalized to January 2015. The data are normalized to this date to reflect Governor Hutchinson taking office.

The forecast values for AR reflect a simple (1 – B) model with three pulse values. One is for April 2011, another for June 2011 and the last is for March 2015.  The normalized AR forecast for the leading indicator is flat.

The forecast values for the U.S. reflect a simple AR(1) model with three intervention variables . There are two time trends one beginning in September 2009 and another in January 2011. There is one pulse variable in September 2010.

The important conclusion is the normalized leading indicators suggest a modest increase in the U.S. economic activity for the months thru March 2020.  The AR index is declining until early 2019 when it begins to increase.  While the actual AR index reveals an uptrend, the forecast is flat.


Forecasts updated 9 November 2019 reflecting data ending September 2019.

Source: FRED: Arkansas data = ARSLIND, US data = USSLIND and author’s calculations.

Arkansas Total Non-Farm Employment

Arkansas total non-farm employment is closely aligned with U.S. payroll employment.  In an analysis (not shown) 99.9% of the variation in Arkansas employment is explained by the variation in U.S. payroll employment and the historical values of Arkansas’ non-farm employment. Conclusion: As U.S. payroll employment moves up and down, so goes Arkansas’ employment.

Forecasts of Arkansas’ expected total payroll employment can be shown as the 12 month % change.  The graphic pattern shows a steady decline in the 12 month percentage change from January 2016 thru August 2018.  Since then, the 12 month percentage change has trended up until September 2019.  The forecast for the next six months is a constant increase of about 1,200.


Forecasts updated on 17 November 2019

The analysis uses data thru September 2019.

SOURCE: FRED, series = ARNA, seasonally adjusted